/brz/remove-bazaar

To get this branch, use:
bzr branch http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2007-03-05 03:10:21 UTC
  • mto: (2255.11.4 dirstate)
  • mto: This revision was merged to the branch mainline in revision 2322.
  • Revision ID: robertc@robertcollins.net-20070305031021-ypbakvagbivyw5mu
Teach delta.report_changes about unversioned files, removing all inventory access during status --short.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 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
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
38
 
    debug,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
39
47
    dirstate,
40
48
    errors,
41
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
 
    revision as _mod_revision,
44
55
    revisiontree,
45
 
    trace,
 
56
    textui,
46
57
    transform,
47
 
    views,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
48
61
    )
49
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
50
64
import bzrlib.ui
51
65
""")
52
66
 
 
67
from bzrlib import symbol_versioning
53
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
 
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
56
73
from bzrlib.mutabletree import needs_tree_write_lock
57
74
from bzrlib.osutils import (
 
75
    compact_date,
58
76
    file_kind,
59
77
    isdir,
 
78
    normpath,
60
79
    pathjoin,
 
80
    rand_chars,
61
81
    realpath,
62
82
    safe_unicode,
 
83
    splitpath,
 
84
    supports_executable,
63
85
    )
64
 
from bzrlib.trace import mutter
 
86
from bzrlib.trace import mutter, note
65
87
from bzrlib.transport.local import LocalTransport
66
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        zero_eight,
 
97
        zero_eleven,
 
98
        zero_thirteen,
 
99
        )
67
100
from bzrlib.tree import Tree
68
101
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
102
 
70
103
 
71
 
class DirStateWorkingTree(WorkingTree3):
 
104
class WorkingTree4(WorkingTree3):
 
105
    """This is the Format 4 working tree.
 
106
 
 
107
    This differs from WorkingTree3 by:
 
108
     - having a consolidated internal dirstate.
 
109
     - not having a regular inventory attribute.
 
110
 
 
111
    This is new in bzr TODO FIXME SETMEBEFORE MERGE.
 
112
    """
 
113
 
72
114
    def __init__(self, basedir,
73
115
                 branch,
74
116
                 _control_files=None,
83
125
        """
84
126
        self._format = _format
85
127
        self.bzrdir = _bzrdir
 
128
        from bzrlib.trace import note, mutter
 
129
        assert isinstance(basedir, basestring), \
 
130
            "base directory %r is not a string" % basedir
86
131
        basedir = safe_unicode(basedir)
87
132
        mutter("opening working tree %r", basedir)
88
133
        self._branch = branch
 
134
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
135
            "branch %r is not a Branch" % self.branch
89
136
        self.basedir = realpath(basedir)
90
137
        # if branch is at our basedir and is a format 6 or less
91
138
        # assume all other formats have their own control files.
 
139
        assert isinstance(_control_files, LockableFiles), \
 
140
            "_control_files must be a LockableFiles, not %r" % _control_files
92
141
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
142
        self._dirty = None
95
143
        #-------------
96
144
        # during a read or write lock these objects are set, and are
98
146
        self._dirstate = None
99
147
        self._inventory = None
100
148
        #-------------
101
 
        self._setup_directory_is_tree_reference()
102
 
        self._detect_case_handling()
103
 
        self._rules_searcher = None
104
 
        self.views = self._make_views()
105
 
        #--- allow tests to select the dirstate iter_changes implementation
106
 
        self._iter_changes = dirstate._process_entry
107
149
 
108
150
    @needs_tree_write_lock
109
151
    def _add(self, files, ids, kinds):
111
153
        state = self.current_dirstate()
112
154
        for f, file_id, kind in zip(files, ids, kinds):
113
155
            f = f.strip('/')
 
156
            assert '//' not in f
 
157
            assert '..' not in f
114
158
            if self.path2id(f):
115
159
                # special case tree root handling.
116
160
                if f == '' and self.path2id(f) == ROOT_ID:
134
178
        if reset_inventory and self._inventory is not None:
135
179
            self._inventory = None
136
180
 
137
 
    @needs_tree_write_lock
138
 
    def add_reference(self, sub_tree):
139
 
        # use standard implementation, which calls back to self._add
140
 
        #
141
 
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
143
 
        self._add_reference(sub_tree)
144
 
 
145
181
    def break_lock(self):
146
182
        """Break a lock if one is present from another instance.
147
183
 
179
215
        self._control_files.break_lock()
180
216
        self.branch.break_lock()
181
217
 
182
 
    def _comparison_data(self, entry, path):
183
 
        kind, executable, stat_value = \
184
 
            WorkingTree3._comparison_data(self, entry, path)
185
 
        # it looks like a plain directory, but it's really a reference -- see
186
 
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
189
 
            kind = 'tree-reference'
190
 
        return kind, executable, stat_value
191
 
 
192
218
    @needs_write_lock
193
219
    def commit(self, message=None, revprops=None, *args, **kwargs):
194
220
        # mark the tree as dirty post commit - commit
205
231
 
206
232
        :raises errors.NotWriteLocked: when not in a lock.
207
233
        """
208
 
        self._must_be_locked()
 
234
        if not self._control_files._lock_count:
 
235
            raise errors.ObjectNotLocked(self)
209
236
        return self._current_dirstate()
210
237
 
211
238
    def _current_dirstate(self):
217
244
            return self._dirstate
218
245
        local_path = self.bzrdir.get_workingtree_transport(None
219
246
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
247
        self._dirstate = dirstate.DirState.on_file(local_path)
222
248
        return self._dirstate
223
249
 
224
 
    def _sha1_provider(self):
225
 
        """A function that returns a SHA1Provider suitable for this tree.
226
 
 
227
 
        :return: None if content filtering is not supported by this tree.
228
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
229
 
          form of files, i.e. after read filters are applied.
230
 
        """
231
 
        if self.supports_content_filtering():
232
 
            return ContentFilterAwareSHA1Provider(self)
233
 
        else:
234
 
            return None
235
 
 
236
250
    def filter_unversioned_files(self, paths):
237
251
        """Filter out paths that are versioned.
238
252
 
261
275
        self._inventory = None
262
276
        self._dirty = False
263
277
 
264
 
    @needs_tree_write_lock
265
 
    def _gather_kinds(self, files, kinds):
266
 
        """See MutableTree._gather_kinds."""
267
 
        for pos, f in enumerate(files):
268
 
            if kinds[pos] is None:
269
 
                kinds[pos] = self._kind(f)
270
 
 
271
278
    def _generate_inventory(self):
272
279
        """Create and set self.inventory from the dirstate object.
273
 
 
 
280
        
274
281
        This is relatively expensive: we have to walk the entire dirstate.
275
282
        Ideally we would not, and can deprecate this function.
276
283
        """
280
287
        state._read_dirblocks_if_needed()
281
288
        root_key, current_entry = self._get_entry(path='')
282
289
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
290
        assert current_entry[0][0] == 'd' # directory
285
291
        inv = Inventory(root_id=current_id)
286
292
        # Turn some things into local variables
287
293
        minikind_to_kind = dirstate.DirState._minikind_to_kind
310
316
                inv_entry = factory[kind](file_id, name_unicode,
311
317
                                          parent_ie.file_id)
312
318
                if kind == 'file':
313
 
                    # This is only needed on win32, where this is the only way
314
 
                    # we know the executable bit.
315
 
                    inv_entry.executable = executable
316
319
                    # not strictly needed: working tree
317
 
                    #inv_entry.text_size = size
318
 
                    #inv_entry.text_sha1 = sha1
 
320
                    #entry.executable = executable
 
321
                    #entry.text_size = size
 
322
                    #entry.text_sha1 = sha1
 
323
                    pass
319
324
                elif kind == 'directory':
320
325
                    # add this entry to the parent map.
321
326
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
 
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
327
 
                    inv_entry.reference_revision = link_or_sha1 or None
328
 
                elif kind != 'symlink':
329
 
                    raise AssertionError("unknown kind %r" % kind)
330
327
                # These checks cost us around 40ms on a 55k entry tree
331
 
                if file_id in inv_byid:
332
 
                    raise AssertionError('file_id %s already in'
333
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
334
 
                if name_unicode in parent_ie.children:
335
 
                    raise AssertionError('name %r already in parent'
336
 
                        % (name_unicode,))
 
328
                assert file_id not in inv_byid, ('file_id %s already in'
 
329
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
330
                assert name_unicode not in parent_ie.children
337
331
                inv_byid[file_id] = inv_entry
338
332
                parent_ie.children[name_unicode] = inv_entry
339
333
        self._inventory = inv
344
338
        If either file_id or path is supplied, it is used as the key to lookup.
345
339
        If both are supplied, the fastest lookup is used, and an error is
346
340
        raised if they do not both point at the same row.
347
 
 
 
341
        
348
342
        :param file_id: An optional unicode file_id to be looked up.
349
343
        :param path: An optional unicode path to be looked up.
350
344
        :return: The dirstate row tuple for path/file_id, or (None, None)
359
353
    def get_file_sha1(self, file_id, path=None, stat_value=None):
360
354
        # check file id is valid unconditionally.
361
355
        entry = self._get_entry(file_id=file_id, path=path)
362
 
        if entry[0] is None:
363
 
            raise errors.NoSuchId(self, file_id)
 
356
        assert entry[0] is not None, 'what error should this raise'
 
357
        # TODO:
 
358
        # if row stat is valid, use cached sha1, else, get a new sha1.
364
359
        if path is None:
365
360
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
366
361
 
367
362
        file_abspath = self.abspath(path)
368
363
        state = self.current_dirstate()
369
 
        if stat_value is None:
370
 
            try:
371
 
                stat_value = os.lstat(file_abspath)
372
 
            except OSError, e:
373
 
                if e.errno == errno.ENOENT:
374
 
                    return None
375
 
                else:
376
 
                    raise
377
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
378
 
            stat_value=stat_value)
 
364
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
365
                                          stat_value=stat_value)
379
366
        if entry[1][0][0] == 'f':
380
 
            if link_or_sha1 is None:
381
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
382
 
                try:
383
 
                    sha1 = osutils.sha_file(file_obj)
384
 
                finally:
385
 
                    file_obj.close()
386
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
387
 
                return sha1
388
 
            else:
389
 
                return link_or_sha1
 
367
            return link_or_sha1
390
368
        return None
391
369
 
392
370
    def _get_inventory(self):
393
371
        """Get the inventory for the tree. This is only valid within a lock."""
394
 
        if 'evil' in debug.debug_flags:
395
 
            trace.mutter_callsite(2,
396
 
                "accessing .inventory forces a size of tree translation.")
397
372
        if self._inventory is not None:
398
373
            return self._inventory
399
 
        self._must_be_locked()
400
374
        self._generate_inventory()
401
375
        return self._inventory
402
376
 
406
380
    @needs_read_lock
407
381
    def get_parent_ids(self):
408
382
        """See Tree.get_parent_ids.
409
 
 
 
383
        
410
384
        This implementation requests the ids list from the dirstate file.
411
385
        """
412
386
        return self.current_dirstate().get_parent_ids()
413
387
 
414
 
    def get_reference_revision(self, file_id, path=None):
415
 
        # referenced tree's revision is whatever's currently there
416
 
        return self.get_nested_tree(file_id, path).last_revision()
417
 
 
418
 
    def get_nested_tree(self, file_id, path=None):
419
 
        if path is None:
420
 
            path = self.id2path(file_id)
421
 
        # else: check file_id is at path?
422
 
        return WorkingTree.open(self.abspath(path))
423
 
 
424
388
    @needs_read_lock
425
389
    def get_root_id(self):
426
390
        """Return the id of this trees root"""
428
392
 
429
393
    def has_id(self, file_id):
430
394
        state = self.current_dirstate()
 
395
        file_id = osutils.safe_file_id(file_id)
431
396
        row, parents = self._get_entry(file_id=file_id)
432
397
        if row is None:
433
398
            return False
434
399
        return osutils.lexists(pathjoin(
435
400
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
401
 
437
 
    def has_or_had_id(self, file_id):
438
 
        state = self.current_dirstate()
439
 
        row, parents = self._get_entry(file_id=file_id)
440
 
        return row is not None
441
 
 
442
402
    @needs_read_lock
443
403
    def id2path(self, file_id):
444
 
        "Convert a file-id to a path."
 
404
        file_id = osutils.safe_file_id(file_id)
445
405
        state = self.current_dirstate()
446
406
        entry = self._get_entry(file_id=file_id)
447
407
        if entry == (None, None):
448
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
408
            return None
449
409
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
450
410
        return path_utf8.decode('utf8')
451
411
 
452
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
453
 
        entry = self._get_entry(path=path)
454
 
        if entry == (None, None):
455
 
            return False # Missing entries are not executable
456
 
        return entry[1][0][3] # Executable?
457
 
 
458
 
    if not osutils.supports_executable():
459
 
        def is_executable(self, file_id, path=None):
460
 
            """Test if a file is executable or not.
461
 
 
462
 
            Note: The caller is expected to take a read-lock before calling this.
463
 
            """
464
 
            entry = self._get_entry(file_id=file_id, path=path)
465
 
            if entry == (None, None):
466
 
                return False
467
 
            return entry[1][0][3]
468
 
 
469
 
        _is_executable_from_path_and_stat = \
470
 
            _is_executable_from_path_and_stat_from_basis
471
 
    else:
472
 
        def is_executable(self, file_id, path=None):
473
 
            """Test if a file is executable or not.
474
 
 
475
 
            Note: The caller is expected to take a read-lock before calling this.
476
 
            """
477
 
            self._must_be_locked()
478
 
            if not path:
479
 
                path = self.id2path(file_id)
480
 
            mode = os.lstat(self.abspath(path)).st_mode
481
 
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
 
 
483
 
    def all_file_ids(self):
484
 
        """See Tree.iter_all_file_ids"""
485
 
        self._must_be_locked()
486
 
        result = set()
487
 
        for key, tree_details in self.current_dirstate()._iter_entries():
488
 
            if tree_details[0][0] in ('a', 'r'): # relocated
489
 
                continue
490
 
            result.add(key[2])
491
 
        return result
492
 
 
493
412
    @needs_read_lock
494
413
    def __iter__(self):
495
414
        """Iterate through file_ids for this tree.
507
426
                result.append(key[2])
508
427
        return iter(result)
509
428
 
510
 
    def iter_references(self):
511
 
        if not self._repo_supports_tree_reference:
512
 
            # When the repo doesn't support references, we will have nothing to
513
 
            # return
514
 
            return
515
 
        for key, tree_details in self.current_dirstate()._iter_entries():
516
 
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
 
                # not relevant to the working tree
518
 
                continue
519
 
            if not key[1]:
520
 
                # the root is not a reference.
521
 
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
523
 
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
526
 
            except errors.NoSuchFile:
527
 
                # path is missing on disk.
528
 
                continue
529
 
 
530
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
531
 
        """See MutableTree._observed_sha1."""
532
 
        state = self.current_dirstate()
533
 
        entry = self._get_entry(file_id=file_id, path=path)
534
 
        state._observed_sha1(entry, sha1, statvalue)
535
 
 
536
 
    def kind(self, file_id):
537
 
        """Return the kind of a file.
538
 
 
539
 
        This is always the actual kind that's on disk, regardless of what it
540
 
        was added as.
541
 
 
542
 
        Note: The caller is expected to take a read-lock before calling this.
543
 
        """
544
 
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
548
 
        return self._kind(relpath)
549
 
 
550
 
    def _kind(self, relpath):
551
 
        abspath = self.abspath(relpath)
552
 
        kind = file_kind(abspath)
553
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
554
 
            entry = self._get_entry(path=relpath)
555
 
            if entry[1] is not None:
556
 
                if entry[1][0][0] == 't':
557
 
                    kind = 'tree-reference'
558
 
        return kind
559
 
 
560
429
    @needs_read_lock
561
430
    def _last_revision(self):
562
431
        """See Mutable.last_revision."""
564
433
        if parent_ids:
565
434
            return parent_ids[0]
566
435
        else:
567
 
            return _mod_revision.NULL_REVISION
 
436
            return None
568
437
 
569
438
    def lock_read(self):
570
 
        """See Branch.lock_read, and WorkingTree.unlock.
571
 
 
572
 
        :return: An object with an unlock method which will release the lock
573
 
            obtained.
574
 
        """
 
439
        """See Branch.lock_read, and WorkingTree.unlock."""
575
440
        self.branch.lock_read()
576
441
        try:
577
442
            self._control_files.lock_read()
579
444
                state = self.current_dirstate()
580
445
                if not state._lock_token:
581
446
                    state.lock_read()
582
 
                # set our support for tree references from the repository in
583
 
                # use.
584
 
                self._repo_supports_tree_reference = getattr(
585
 
                    self.branch.repository._format, "supports_tree_reference",
586
 
                    False)
587
447
            except:
588
448
                self._control_files.unlock()
589
449
                raise
590
450
        except:
591
451
            self.branch.unlock()
592
452
            raise
593
 
        return self
594
453
 
595
454
    def _lock_self_write(self):
596
455
        """This should be called after the branch is locked."""
600
459
                state = self.current_dirstate()
601
460
                if not state._lock_token:
602
461
                    state.lock_write()
603
 
                # set our support for tree references from the repository in
604
 
                # use.
605
 
                self._repo_supports_tree_reference = getattr(
606
 
                    self.branch.repository._format, "supports_tree_reference",
607
 
                    False)
608
462
            except:
609
463
                self._control_files.unlock()
610
464
                raise
611
465
        except:
612
466
            self.branch.unlock()
613
467
            raise
614
 
        return self
615
468
 
616
469
    def lock_tree_write(self):
617
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
618
 
 
619
 
        :return: An object with an unlock method which will release the lock
620
 
            obtained.
621
 
        """
 
470
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
622
471
        self.branch.lock_read()
623
 
        return self._lock_self_write()
 
472
        self._lock_self_write()
624
473
 
625
474
    def lock_write(self):
626
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
627
 
 
628
 
        :return: An object with an unlock method which will release the lock
629
 
            obtained.
630
 
        """
 
475
        """See MutableTree.lock_write, and WorkingTree.unlock."""
631
476
        self.branch.lock_write()
632
 
        return self._lock_self_write()
 
477
        self._lock_self_write()
633
478
 
634
479
    @needs_tree_write_lock
635
480
    def move(self, from_paths, to_dir, after=False):
637
482
        result = []
638
483
        if not from_paths:
639
484
            return result
 
485
 
640
486
        state = self.current_dirstate()
641
 
        if isinstance(from_paths, basestring):
642
 
            raise ValueError()
 
487
 
 
488
        assert not isinstance(from_paths, basestring)
643
489
        to_dir_utf8 = to_dir.encode('utf8')
644
490
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
645
491
        id_index = state._get_id_index()
667
513
        if self._inventory is not None:
668
514
            update_inventory = True
669
515
            inv = self.inventory
 
516
            to_dir_ie = inv[to_dir_id]
670
517
            to_dir_id = to_entry[0][2]
671
 
            to_dir_ie = inv[to_dir_id]
672
518
        else:
673
519
            update_inventory = False
674
520
 
697
543
            new_entry = to_block[1][added_entry_index]
698
544
            rollbacks.append(lambda:state._make_absent(new_entry))
699
545
 
 
546
        # create rename entries and tuples
700
547
        for from_rel in from_paths:
701
548
            # from_rel is 'pathinroot/foo/bar'
702
549
            from_rel_utf8 = from_rel.encode('utf8')
705
552
            from_entry = self._get_entry(path=from_rel)
706
553
            if from_entry == (None, None):
707
554
                raise errors.BzrMoveFailedError(from_rel,to_dir,
708
 
                    errors.NotVersionedError(path=from_rel))
 
555
                    errors.NotVersionedError(path=str(from_rel)))
709
556
 
710
557
            from_id = from_entry[0][2]
711
558
            to_rel = pathjoin(to_dir, from_tail)
738
585
                if from_missing: # implicitly just update our path mapping
739
586
                    move_file = False
740
587
                elif not after:
741
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
588
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
589
                        extra="(Use --after to update the Bazaar id)")
742
590
 
743
591
            rollbacks = []
744
592
            def rollback_rename():
745
593
                """A single rename has failed, roll it back."""
746
 
                # roll back everything, even if we encounter trouble doing one
747
 
                # of them.
748
 
                #
749
 
                # TODO: at least log the other exceptions rather than just
750
 
                # losing them mbp 20070307
751
594
                exc_info = None
752
595
                for rollback in reversed(rollbacks):
753
596
                    try:
754
597
                        rollback()
755
598
                    except Exception, e:
 
599
                        import pdb;pdb.set_trace()
756
600
                        exc_info = sys.exc_info()
757
601
                if exc_info:
758
602
                    raise exc_info[0], exc_info[1], exc_info[2]
799
643
 
800
644
                if minikind == 'd':
801
645
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
802
 
                        """Recursively update all entries in this dirblock."""
803
 
                        if from_dir == '':
804
 
                            raise AssertionError("renaming root not supported")
 
646
                        """all entries in this block need updating.
 
647
 
 
648
                        TODO: This is pretty ugly, and doesn't support
 
649
                        reverting, but it works.
 
650
                        """
 
651
                        assert from_dir != '', "renaming root not supported"
805
652
                        from_key = (from_dir, '')
806
653
                        from_block_idx, present = \
807
654
                            state._find_block_index_from_key(from_key)
817
664
                        to_block_index = state._ensure_block(
818
665
                            to_block_index, to_entry_index, to_dir_utf8)
819
666
                        to_block = state._dirblocks[to_block_index]
820
 
 
821
 
                        # Grab a copy since move_one may update the list.
822
 
                        for entry in from_block[1][:]:
823
 
                            if not (entry[0][0] == from_dir):
824
 
                                raise AssertionError()
 
667
                        for entry in from_block[1]:
 
668
                            assert entry[0][0] == from_dir
825
669
                            cur_details = entry[1][0]
826
670
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
827
671
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
828
672
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
829
673
                            minikind = cur_details[0]
830
 
                            if minikind in 'ar':
831
 
                                # Deleted children of a renamed directory
832
 
                                # Do not need to be updated.
833
 
                                # Children that have been renamed out of this
834
 
                                # directory should also not be updated
835
 
                                continue
836
674
                            move_one(entry, from_path_utf8=from_path_utf8,
837
675
                                     minikind=minikind,
838
676
                                     executable=cur_details[3],
841
679
                                     size=cur_details[2],
842
680
                                     to_block=to_block,
843
681
                                     to_key=to_key,
844
 
                                     to_path_utf8=to_path_utf8)
 
682
                                     to_path_utf8=to_rel_utf8)
845
683
                            if minikind == 'd':
846
684
                                # We need to move all the children of this
847
685
                                # entry
857
695
 
858
696
        return result
859
697
 
860
 
    def _must_be_locked(self):
861
 
        if not self._control_files._lock_count:
862
 
            raise errors.ObjectNotLocked(self)
863
 
 
864
698
    def _new_tree(self):
865
699
        """Initialize the state in this tree to be a new tree."""
866
700
        self._dirty = True
886
720
        for tree in trees:
887
721
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
888
722
                parents):
889
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
890
 
                    trees, require_versioned)
 
723
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
891
724
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
892
725
        # -- make all paths utf8 --
893
726
        paths_utf8 = set()
953
786
            if not all_versioned:
954
787
                raise errors.PathsNotVersionedError(paths)
955
788
        # -- remove redundancy in supplied paths to prevent over-scanning --
956
 
        search_paths = osutils.minimum_path_selection(paths)
957
 
        # sketch:
 
789
        search_paths = set()
 
790
        for path in paths:
 
791
            other_paths = paths.difference(set([path]))
 
792
            if not osutils.is_inside_any(other_paths, path):
 
793
                # this is a top level path, we must check it.
 
794
                search_paths.add(path)
 
795
        # sketch: 
958
796
        # for all search_indexs in each path at or under each element of
959
797
        # search_paths, if the detail is relocated: add the id, and add the
960
798
        # relocated path as one to search if its not searched already. If the
1016
854
 
1017
855
    def read_working_inventory(self):
1018
856
        """Read the working inventory.
1019
 
 
 
857
        
1020
858
        This is a meaningless operation for dirstate, but we obey it anyhow.
1021
859
        """
1022
860
        return self.inventory
1027
865
 
1028
866
        WorkingTree4 supplies revision_trees for any basis tree.
1029
867
        """
 
868
        revision_id = osutils.safe_revision_id(revision_id)
1030
869
        dirstate = self.current_dirstate()
1031
870
        parent_ids = dirstate.get_parent_ids()
1032
871
        if revision_id not in parent_ids:
1039
878
    @needs_tree_write_lock
1040
879
    def set_last_revision(self, new_revision):
1041
880
        """Change the last revision in the working tree."""
 
881
        new_revision = osutils.safe_revision_id(new_revision)
1042
882
        parents = self.get_parent_ids()
1043
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1044
 
            if len(parents) >= 2:
1045
 
                raise AssertionError(
1046
 
                    "setting the last parent to none with a pending merge is "
1047
 
                    "unsupported.")
 
883
        if new_revision in (NULL_REVISION, None):
 
884
            assert len(parents) < 2, (
 
885
                "setting the last parent to none with a pending merge is "
 
886
                "unsupported.")
1048
887
            self.set_parent_ids([])
1049
888
        else:
1050
889
            self.set_parent_ids([new_revision] + parents[1:],
1053
892
    @needs_tree_write_lock
1054
893
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1055
894
        """Set the parent ids to revision_ids.
1056
 
 
 
895
        
1057
896
        See also set_parent_trees. This api will try to retrieve the tree data
1058
897
        for each element of revision_ids from the trees repository. If you have
1059
898
        tree data already available, it is more efficient to use
1063
902
        :param revision_ids: The revision_ids to set as the parent ids of this
1064
903
            working tree. Any of these may be ghosts.
1065
904
        """
 
905
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1066
906
        trees = []
1067
907
        for revision_id in revision_ids:
1068
908
            try:
1091
931
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1092
932
        real_trees = []
1093
933
        ghosts = []
1094
 
 
1095
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1096
 
        graph = self.branch.repository.get_graph()
1097
 
        heads = graph.heads(parent_ids)
1098
 
        accepted_revisions = set()
1099
 
 
1100
934
        # convert absent trees to the null tree, which we convert back to
1101
935
        # missing on access.
1102
936
        for rev_id, tree in parents_list:
1103
 
            if len(accepted_revisions) > 0:
1104
 
                # we always accept the first tree
1105
 
                if rev_id in accepted_revisions or rev_id not in heads:
1106
 
                    # We have already included either this tree, or its
1107
 
                    # descendent, so we skip it.
1108
 
                    continue
1109
 
            _mod_revision.check_not_reserved_id(rev_id)
 
937
            rev_id = osutils.safe_revision_id(rev_id)
1110
938
            if tree is not None:
1111
939
                real_trees.append((rev_id, tree))
1112
940
            else:
1113
941
                real_trees.append((rev_id,
1114
 
                    self.branch.repository.revision_tree(
1115
 
                        _mod_revision.NULL_REVISION)))
 
942
                    self.branch.repository.revision_tree(None)))
1116
943
                ghosts.append(rev_id)
1117
 
            accepted_revisions.add(rev_id)
1118
944
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1119
945
        self._make_dirty(reset_inventory=False)
1120
946
 
1125
951
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1126
952
            self._make_dirty(reset_inventory=True)
1127
953
 
1128
 
    def _sha_from_stat(self, path, stat_result):
1129
 
        """Get a sha digest from the tree's stat cache.
1130
 
 
1131
 
        The default implementation assumes no stat cache is present.
1132
 
 
1133
 
        :param path: The path.
1134
 
        :param stat_result: The stat result being looked up.
1135
 
        """
1136
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1137
 
 
1138
 
    @needs_read_lock
1139
 
    def supports_tree_reference(self):
1140
 
        return self._repo_supports_tree_reference
1141
 
 
1142
954
    def unlock(self):
1143
955
        """Unlock in format 4 trees needs to write the entire dirstate."""
1144
 
        # do non-implementation specific cleanup
1145
 
        self._cleanup()
1146
 
 
1147
956
        if self._control_files._lock_count == 1:
1148
957
            # eventually we should do signature checking during read locks for
1149
958
            # dirstate updates.
1180
989
            return
1181
990
        state = self.current_dirstate()
1182
991
        state._read_dirblocks_if_needed()
1183
 
        ids_to_unversion = set(file_ids)
 
992
        ids_to_unversion = set()
 
993
        for file_id in file_ids:
 
994
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1184
995
        paths_to_unversion = set()
1185
996
        # sketch:
1186
997
        # check if the root is to be unversioned, if so, assert for now.
1213
1024
                # just forget the whole block.
1214
1025
                entry_index = 0
1215
1026
                while entry_index < len(block[1]):
1216
 
                    entry = block[1][entry_index]
1217
 
                    if entry[1][0][0] in 'ar':
1218
 
                        # don't remove absent or renamed entries
 
1027
                    # Mark this file id as having been removed
 
1028
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1029
                    if not state._make_absent(block[1][entry_index]):
1219
1030
                        entry_index += 1
1220
 
                    else:
1221
 
                        # Mark this file id as having been removed
1222
 
                        ids_to_unversion.discard(entry[0][2])
1223
 
                        if not state._make_absent(entry):
1224
 
                            # The block has not shrunk.
1225
 
                            entry_index += 1
1226
1031
                # go to the next block. (At the moment we dont delete empty
1227
1032
                # dirblocks)
1228
1033
                block_index += 1
1252
1057
                self._inventory.remove_recursive_id(file_id)
1253
1058
 
1254
1059
    @needs_tree_write_lock
1255
 
    def rename_one(self, from_rel, to_rel, after=False):
1256
 
        """See WorkingTree.rename_one"""
1257
 
        self.flush()
1258
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1259
 
 
1260
 
    @needs_tree_write_lock
1261
 
    def apply_inventory_delta(self, changes):
1262
 
        """See MutableTree.apply_inventory_delta"""
1263
 
        state = self.current_dirstate()
1264
 
        state.update_by_delta(changes)
1265
 
        self._make_dirty(reset_inventory=True)
1266
 
 
1267
 
    def update_basis_by_delta(self, new_revid, delta):
1268
 
        """See MutableTree.update_basis_by_delta."""
1269
 
        if self.last_revision() == new_revid:
1270
 
            raise AssertionError()
1271
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1272
 
 
1273
 
    @needs_read_lock
1274
 
    def _validate(self):
1275
 
        self._dirstate._validate()
1276
 
 
1277
 
    @needs_tree_write_lock
1278
1060
    def _write_inventory(self, inv):
1279
1061
        """Write inventory as the current inventory."""
1280
 
        if self._dirty:
1281
 
            raise AssertionError("attempting to write an inventory when the "
1282
 
                "dirstate is dirty will lose pending changes")
1283
 
        had_inventory = self._inventory is not None
1284
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1285
 
        # working inventory. We do this because self.inventory may be inv, or
1286
 
        # may have been modified, and either case would prevent a clean delta
1287
 
        # being created.
1288
 
        self._inventory = None
1289
 
        # generate a delta,
1290
 
        delta = inv._make_delta(self.inventory)
1291
 
        # and apply it.
1292
 
        self.apply_inventory_delta(delta)
1293
 
        if had_inventory:
 
1062
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1063
        self.current_dirstate().set_state_from_inventory(inv)
 
1064
        self._make_dirty(reset_inventory=False)
 
1065
        if self._inventory is not None:
1294
1066
            self._inventory = inv
1295
1067
        self.flush()
1296
1068
 
1297
1069
 
1298
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1299
 
 
1300
 
    def __init__(self, tree):
1301
 
        self.tree = tree
1302
 
 
1303
 
    def sha1(self, abspath):
1304
 
        """See dirstate.SHA1Provider.sha1()."""
1305
 
        filters = self.tree._content_filter_stack(
1306
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1307
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1308
 
 
1309
 
    def stat_and_sha1(self, abspath):
1310
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1311
 
        filters = self.tree._content_filter_stack(
1312
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1313
 
        file_obj = file(abspath, 'rb', 65000)
1314
 
        try:
1315
 
            statvalue = os.fstat(file_obj.fileno())
1316
 
            if filters:
1317
 
                file_obj = filtered_input_file(file_obj, filters)
1318
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1319
 
        finally:
1320
 
            file_obj.close()
1321
 
        return statvalue, sha1
1322
 
 
1323
 
 
1324
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1325
 
    """Dirstate working tree that supports content filtering.
1326
 
 
1327
 
    The dirstate holds the hash and size of the canonical form of the file, 
1328
 
    and most methods must return that.
1329
 
    """
1330
 
 
1331
 
    def _file_content_summary(self, path, stat_result):
1332
 
        # This is to support the somewhat obsolete path_content_summary method
1333
 
        # with content filtering: see
1334
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1335
 
        #
1336
 
        # If the dirstate cache is up to date and knows the hash and size,
1337
 
        # return that.
1338
 
        # Otherwise if there are no content filters, return the on-disk size
1339
 
        # and leave the hash blank.
1340
 
        # Otherwise, read and filter the on-disk file and use its size and
1341
 
        # hash.
1342
 
        #
1343
 
        # The dirstate doesn't store the size of the canonical form so we
1344
 
        # can't trust it for content-filtered trees.  We just return None.
1345
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1346
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1347
 
        return ('file', None, executable, dirstate_sha1)
1348
 
 
1349
 
 
1350
 
class WorkingTree4(DirStateWorkingTree):
1351
 
    """This is the Format 4 working tree.
1352
 
 
1353
 
    This differs from WorkingTree3 by:
1354
 
     - Having a consolidated internal dirstate, stored in a
1355
 
       randomly-accessible sorted file on disk.
1356
 
     - Not having a regular inventory attribute.  One can be synthesized
1357
 
       on demand but this is expensive and should be avoided.
1358
 
 
1359
 
    This is new in bzr 0.15.
1360
 
    """
1361
 
 
1362
 
 
1363
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1364
 
    """This is the Format 5 working tree.
1365
 
 
1366
 
    This differs from WorkingTree4 by:
1367
 
     - Supporting content filtering.
1368
 
 
1369
 
    This is new in bzr 1.11.
1370
 
    """
1371
 
 
1372
 
 
1373
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1374
 
    """This is the Format 6 working tree.
1375
 
 
1376
 
    This differs from WorkingTree5 by:
1377
 
     - Supporting a current view that may mask the set of files in a tree
1378
 
       impacted by most user operations.
1379
 
 
1380
 
    This is new in bzr 1.14.
1381
 
    """
1382
 
 
1383
 
    def _make_views(self):
1384
 
        return views.PathBasedViews(self)
1385
 
 
1386
 
 
1387
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1388
 
 
1389
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1390
 
                   accelerator_tree=None, hardlink=False):
 
1070
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1071
    """The first consolidated dirstate working tree format.
 
1072
 
 
1073
    This format:
 
1074
        - exists within a metadir controlling .bzr
 
1075
        - includes an explicit version marker for the workingtree control
 
1076
          files, separate from the BzrDir format
 
1077
        - modifies the hash cache format
 
1078
        - is new in bzr TODO FIXME SETBEFOREMERGE
 
1079
        - uses a LockDir to guard access to it.
 
1080
    """
 
1081
 
 
1082
    def get_format_string(self):
 
1083
        """See WorkingTreeFormat.get_format_string()."""
 
1084
        return "Bazaar Working Tree format 4\n"
 
1085
 
 
1086
    def get_format_description(self):
 
1087
        """See WorkingTreeFormat.get_format_description()."""
 
1088
        return "Working tree format 4"
 
1089
 
 
1090
    def initialize(self, a_bzrdir, revision_id=None):
1391
1091
        """See WorkingTreeFormat.initialize().
1392
1092
 
1393
 
        :param revision_id: allows creating a working tree at a different
 
1093
        revision_id allows creating a working tree at a different
1394
1094
        revision than the branch is at.
1395
 
        :param accelerator_tree: A tree which can be used for retrieving file
1396
 
            contents more quickly than the revision tree, i.e. a workingtree.
1397
 
            The revision tree will be used for cases where accelerator_tree's
1398
 
            content is different.
1399
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1400
 
            where possible.
1401
 
 
1402
 
        These trees get an initial random root id, if their repository supports
1403
 
        rich root data, TREE_ROOT otherwise.
1404
1095
        """
 
1096
        revision_id = osutils.safe_revision_id(revision_id)
1405
1097
        if not isinstance(a_bzrdir.transport, LocalTransport):
1406
1098
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1407
1099
        transport = a_bzrdir.get_workingtree_transport(self)
1408
1100
        control_files = self._open_control_files(a_bzrdir)
1409
1101
        control_files.create_lock()
1410
1102
        control_files.lock_write()
1411
 
        transport.put_bytes('format', self.get_format_string(),
1412
 
            mode=a_bzrdir._get_file_mode())
1413
 
        if from_branch is not None:
1414
 
            branch = from_branch
1415
 
        else:
1416
 
            branch = a_bzrdir.open_branch()
 
1103
        control_files.put_utf8('format', self.get_format_string())
 
1104
        branch = a_bzrdir.open_branch()
1417
1105
        if revision_id is None:
1418
1106
            revision_id = branch.last_revision()
1419
1107
        local_path = transport.local_abspath('dirstate')
1420
 
        # write out new dirstate (must exist when we create the tree)
1421
1108
        state = dirstate.DirState.initialize(local_path)
1422
1109
        state.unlock()
1423
 
        del state
1424
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1110
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1425
1111
                         branch,
1426
1112
                         _format=self,
1427
1113
                         _bzrdir=a_bzrdir,
1429
1115
        wt._new_tree()
1430
1116
        wt.lock_tree_write()
1431
1117
        try:
1432
 
            self._init_custom_control_files(wt)
1433
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
1434
 
                if branch.repository.supports_rich_root():
1435
 
                    wt._set_root_id(generate_ids.gen_root_id())
1436
 
                else:
1437
 
                    wt._set_root_id(ROOT_ID)
1438
 
                wt.flush()
1439
 
            basis = None
1440
 
            # frequently, we will get here due to branching.  The accelerator
1441
 
            # tree will be the tree from the branch, so the desired basis
1442
 
            # tree will often be a parent of the accelerator tree.
1443
 
            if accelerator_tree is not None:
1444
 
                try:
1445
 
                    basis = accelerator_tree.revision_tree(revision_id)
1446
 
                except errors.NoSuchRevision:
1447
 
                    pass
1448
 
            if basis is None:
1449
 
                basis = branch.repository.revision_tree(revision_id)
1450
 
            if revision_id == _mod_revision.NULL_REVISION:
1451
 
                parents_list = []
1452
 
            else:
1453
 
                parents_list = [(revision_id, basis)]
 
1118
            wt.set_last_revision(revision_id)
 
1119
            wt.flush()
 
1120
            basis = wt.basis_tree()
1454
1121
            basis.lock_read()
1455
 
            try:
1456
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
1457
 
                wt.flush()
1458
 
                # if the basis has a root id we have to use that; otherwise we
1459
 
                # use a new random one
1460
 
                basis_root_id = basis.get_root_id()
1461
 
                if basis_root_id is not None:
1462
 
                    wt._set_root_id(basis_root_id)
1463
 
                    wt.flush()
1464
 
                if wt.supports_content_filtering():
1465
 
                    # The original tree may not have the same content filters
1466
 
                    # applied so we can't safely build the inventory delta from
1467
 
                    # the source tree.
1468
 
                    delta_from_tree = False
1469
 
                else:
1470
 
                    delta_from_tree = True
1471
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1472
 
                # because wt4.apply_inventory_delta does not mutate the input
1473
 
                # inventory entries.
1474
 
                transform.build_tree(basis, wt, accelerator_tree,
1475
 
                                     hardlink=hardlink,
1476
 
                                     delta_from_tree=delta_from_tree)
1477
 
            finally:
1478
 
                basis.unlock()
 
1122
            transform.build_tree(basis, wt)
 
1123
            basis.unlock()
1479
1124
        finally:
1480
1125
            control_files.unlock()
1481
1126
            wt.unlock()
1482
1127
        return wt
1483
1128
 
1484
 
    def _init_custom_control_files(self, wt):
1485
 
        """Subclasses with custom control files should override this method.
1486
 
 
1487
 
        The working tree and control files are locked for writing when this
1488
 
        method is called.
1489
 
 
1490
 
        :param wt: the WorkingTree object
1491
 
        """
1492
1129
 
1493
1130
    def _open(self, a_bzrdir, control_files):
1494
1131
        """Open the tree itself.
1496
1133
        :param a_bzrdir: the dir for the tree.
1497
1134
        :param control_files: the control files for the tree.
1498
1135
        """
1499
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1136
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1500
1137
                           branch=a_bzrdir.open_branch(),
1501
1138
                           _format=self,
1502
1139
                           _bzrdir=a_bzrdir,
1503
1140
                           _control_files=control_files)
1504
1141
 
1505
 
    def __get_matchingbzrdir(self):
1506
 
        return self._get_matchingbzrdir()
1507
 
 
1508
 
    def _get_matchingbzrdir(self):
1509
 
        """Overrideable method to get a bzrdir for testing."""
1510
 
        # please test against something that will let us do tree references
1511
 
        return bzrdir.format_registry.make_bzrdir(
1512
 
            'dirstate-with-subtree')
1513
 
 
1514
 
    _matchingbzrdir = property(__get_matchingbzrdir)
1515
 
 
1516
 
 
1517
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1518
 
    """The first consolidated dirstate working tree format.
1519
 
 
1520
 
    This format:
1521
 
        - exists within a metadir controlling .bzr
1522
 
        - includes an explicit version marker for the workingtree control
1523
 
          files, separate from the BzrDir format
1524
 
        - modifies the hash cache format
1525
 
        - is new in bzr 0.15
1526
 
        - uses a LockDir to guard access to it.
1527
 
    """
1528
 
 
1529
 
    upgrade_recommended = False
1530
 
 
1531
 
    _tree_class = WorkingTree4
1532
 
 
1533
 
    def get_format_string(self):
1534
 
        """See WorkingTreeFormat.get_format_string()."""
1535
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1536
 
 
1537
 
    def get_format_description(self):
1538
 
        """See WorkingTreeFormat.get_format_description()."""
1539
 
        return "Working tree format 4"
1540
 
 
1541
 
 
1542
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1543
 
    """WorkingTree format supporting content filtering.
1544
 
    """
1545
 
 
1546
 
    upgrade_recommended = False
1547
 
 
1548
 
    _tree_class = WorkingTree5
1549
 
 
1550
 
    def get_format_string(self):
1551
 
        """See WorkingTreeFormat.get_format_string()."""
1552
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1553
 
 
1554
 
    def get_format_description(self):
1555
 
        """See WorkingTreeFormat.get_format_description()."""
1556
 
        return "Working tree format 5"
1557
 
 
1558
 
    def supports_content_filtering(self):
1559
 
        return True
1560
 
 
1561
 
 
1562
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1563
 
    """WorkingTree format supporting views.
1564
 
    """
1565
 
 
1566
 
    upgrade_recommended = False
1567
 
 
1568
 
    _tree_class = WorkingTree6
1569
 
 
1570
 
    def get_format_string(self):
1571
 
        """See WorkingTreeFormat.get_format_string()."""
1572
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1573
 
 
1574
 
    def get_format_description(self):
1575
 
        """See WorkingTreeFormat.get_format_description()."""
1576
 
        return "Working tree format 6"
1577
 
 
1578
 
    def _init_custom_control_files(self, wt):
1579
 
        """Subclasses with custom control files should override this method."""
1580
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1581
 
 
1582
 
    def supports_content_filtering(self):
1583
 
        return True
1584
 
 
1585
 
    def supports_views(self):
1586
 
        return True
1587
 
 
1588
1142
 
1589
1143
class DirStateRevisionTree(Tree):
1590
 
    """A revision tree pulling the inventory from a dirstate.
1591
 
    
1592
 
    Note that this is one of the historical (ie revision) trees cached in the
1593
 
    dirstate for easy access, not the workingtree.
1594
 
    """
 
1144
    """A revision tree pulling the inventory from a dirstate."""
1595
1145
 
1596
1146
    def __init__(self, dirstate, revision_id, repository):
1597
1147
        self._dirstate = dirstate
1598
 
        self._revision_id = revision_id
 
1148
        self._revision_id = osutils.safe_revision_id(revision_id)
1599
1149
        self._repository = repository
1600
1150
        self._inventory = None
1601
1151
        self._locked = 0
1602
1152
        self._dirstate_locked = False
1603
 
        self._repo_supports_tree_reference = getattr(
1604
 
            repository._format, "supports_tree_reference",
1605
 
            False)
1606
 
 
1607
 
    def __repr__(self):
1608
 
        return "<%s of %s in %s>" % \
1609
 
            (self.__class__.__name__, self._revision_id, self._dirstate)
1610
 
 
1611
 
    def annotate_iter(self, file_id,
1612
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1153
 
 
1154
    def annotate_iter(self, file_id):
1613
1155
        """See Tree.annotate_iter"""
1614
 
        text_key = (file_id, self.inventory[file_id].revision)
1615
 
        annotations = self._repository.texts.annotate(text_key)
1616
 
        return [(key[-1], line) for (key, line) in annotations]
 
1156
        w = self._repository.weave_store.get_weave(file_id,
 
1157
                           self._repository.get_transaction())
 
1158
        return w.annotate_iter(self.inventory[file_id].revision)
1617
1159
 
1618
 
    def _get_ancestors(self, default_revision):
1619
 
        return set(self._repository.get_ancestry(self._revision_id,
1620
 
                                                 topo_sorted=False))
1621
1160
    def _comparison_data(self, entry, path):
1622
1161
        """See Tree._comparison_data."""
1623
1162
        if entry is None:
1637
1176
        pred = self.has_filename
1638
1177
        return set((p for p in paths if not pred(p)))
1639
1178
 
1640
 
    def get_root_id(self):
1641
 
        return self.path2id('')
1642
 
 
1643
 
    def id2path(self, file_id):
1644
 
        "Convert a file-id to a path."
1645
 
        entry = self._get_entry(file_id=file_id)
1646
 
        if entry == (None, None):
1647
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1648
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1649
 
        return path_utf8.decode('utf8')
1650
 
 
1651
 
    def iter_references(self):
1652
 
        if not self._repo_supports_tree_reference:
1653
 
            # When the repo doesn't support references, we will have nothing to
1654
 
            # return
1655
 
            return iter([])
1656
 
        # Otherwise, fall back to the default implementation
1657
 
        return super(DirStateRevisionTree, self).iter_references()
1658
 
 
1659
1179
    def _get_parent_index(self):
1660
1180
        """Return the index in the dirstate referenced by this tree."""
1661
1181
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1666
1186
        If either file_id or path is supplied, it is used as the key to lookup.
1667
1187
        If both are supplied, the fastest lookup is used, and an error is
1668
1188
        raised if they do not both point at the same row.
1669
 
 
 
1189
        
1670
1190
        :param file_id: An optional unicode file_id to be looked up.
1671
1191
        :param path: An optional unicode path to be looked up.
1672
1192
        :return: The dirstate row tuple for path/file_id, or (None, None)
1673
1193
        """
1674
1194
        if file_id is None and path is None:
1675
1195
            raise errors.BzrError('must supply file_id or path')
 
1196
        file_id = osutils.safe_file_id(file_id)
1676
1197
        if path is not None:
1677
1198
            path = path.encode('utf8')
1678
1199
        parent_index = self._get_parent_index()
1681
1202
    def _generate_inventory(self):
1682
1203
        """Create and set self.inventory from the dirstate object.
1683
1204
 
1684
 
        (So this is only called the first time the inventory is requested for
1685
 
        this tree; it then remains in memory until it's out of date.)
1686
 
 
1687
1205
        This is relatively expensive: we have to walk the entire dirstate.
1688
 
        """
1689
 
        if not self._locked:
1690
 
            raise AssertionError(
1691
 
                'cannot generate inventory of an unlocked '
1692
 
                'dirstate revision tree')
 
1206
        Ideally we would not, and instead would """
 
1207
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1208
            'dirstate revision tree'
1693
1209
        # separate call for profiling - makes it clear where the costs are.
1694
1210
        self._dirstate._read_dirblocks_if_needed()
1695
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1696
 
            raise AssertionError(
1697
 
                'parent %s has disappeared from %s' % (
1698
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1211
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1212
            'parent %s has disappeared from %s' % (
 
1213
            self._revision_id, self._dirstate.get_parent_ids())
1699
1214
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1700
1215
        # This is identical now to the WorkingTree _generate_inventory except
1701
1216
        # for the tree index use.
1702
1217
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1703
1218
        current_id = root_key[2]
1704
 
        if current_entry[parent_index][0] != 'd':
1705
 
            raise AssertionError()
 
1219
        assert current_entry[parent_index][0] == 'd'
1706
1220
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1707
1221
        inv.root.revision = current_entry[parent_index][4]
1708
1222
        # Turn some things into local variables
1721
1235
                # all the paths in this block are not versioned in this tree
1722
1236
                continue
1723
1237
            for key, entry in block[1]:
1724
 
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1238
                minikind, link_or_sha1, size, executable, revid = entry[parent_index]
1725
1239
                if minikind in ('a', 'r'): # absent, relocated
1726
1240
                    # not this tree
1727
1241
                    continue
1735
1249
                if kind == 'file':
1736
1250
                    inv_entry.executable = executable
1737
1251
                    inv_entry.text_size = size
1738
 
                    inv_entry.text_sha1 = fingerprint
 
1252
                    inv_entry.text_sha1 = link_or_sha1
1739
1253
                elif kind == 'directory':
1740
1254
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1741
1255
                elif kind == 'symlink':
1742
1256
                    inv_entry.executable = False
1743
 
                    inv_entry.text_size = None
1744
 
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1745
 
                elif kind == 'tree-reference':
1746
 
                    inv_entry.reference_revision = fingerprint or None
 
1257
                    inv_entry.text_size = size
 
1258
                    inv_entry.symlink_target = utf8_decode(link_or_sha1)[0]
1747
1259
                else:
1748
 
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1749
 
                            % entry)
 
1260
                    raise Exception, kind
1750
1261
                # These checks cost us around 40ms on a 55k entry tree
1751
 
                if file_id in inv_byid:
1752
 
                    raise AssertionError('file_id %s already in'
1753
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1754
 
                if name_unicode in parent_ie.children:
1755
 
                    raise AssertionError('name %r already in parent'
1756
 
                        % (name_unicode,))
 
1262
                assert file_id not in inv_byid
 
1263
                assert name_unicode not in parent_ie.children
1757
1264
                inv_byid[file_id] = inv_entry
1758
1265
                parent_ie.children[name_unicode] = inv_entry
1759
1266
        self._inventory = inv
1769
1276
            return None
1770
1277
        parent_index = self._get_parent_index()
1771
1278
        last_changed_revision = entry[1][parent_index][4]
1772
 
        try:
1773
 
            rev = self._repository.get_revision(last_changed_revision)
1774
 
        except errors.NoSuchRevision:
1775
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1776
 
        return rev.timestamp
 
1279
        return self._repository.get_revision(last_changed_revision).timestamp
1777
1280
 
1778
1281
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1779
1282
        entry = self._get_entry(file_id=file_id, path=path)
1783
1286
            return parent_details[1]
1784
1287
        return None
1785
1288
 
1786
 
    def get_file(self, file_id, path=None):
 
1289
    def get_file(self, file_id):
1787
1290
        return StringIO(self.get_file_text(file_id))
1788
1291
 
 
1292
    def get_file_lines(self, file_id):
 
1293
        ie = self.inventory[file_id]
 
1294
        return self._repository.weave_store.get_weave(file_id,
 
1295
                self._repository.get_transaction()).get_lines(ie.revision)
 
1296
 
1789
1297
    def get_file_size(self, file_id):
1790
 
        """See Tree.get_file_size"""
1791
1298
        return self.inventory[file_id].text_size
1792
1299
 
1793
 
    def get_file_text(self, file_id, path=None):
1794
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1795
 
        return ''.join(content)
1796
 
 
1797
 
    def get_reference_revision(self, file_id, path=None):
1798
 
        return self.inventory[file_id].reference_revision
1799
 
 
1800
 
    def iter_files_bytes(self, desired_files):
1801
 
        """See Tree.iter_files_bytes.
1802
 
 
1803
 
        This version is implemented on top of Repository.iter_files_bytes"""
1804
 
        parent_index = self._get_parent_index()
1805
 
        repo_desired_files = []
1806
 
        for file_id, identifier in desired_files:
1807
 
            entry = self._get_entry(file_id)
1808
 
            if entry == (None, None):
1809
 
                raise errors.NoSuchId(self, file_id)
1810
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1811
 
                                       identifier))
1812
 
        return self._repository.iter_files_bytes(repo_desired_files)
 
1300
    def get_file_text(self, file_id):
 
1301
        return ''.join(self.get_file_lines(file_id))
1813
1302
 
1814
1303
    def get_symlink_target(self, file_id):
1815
1304
        entry = self._get_entry(file_id=file_id)
1817
1306
        if entry[1][parent_index][0] != 'l':
1818
1307
            return None
1819
1308
        else:
1820
 
            target = entry[1][parent_index][1]
1821
 
            target = target.decode('utf8')
1822
 
            return target
 
1309
            # At present, none of the tree implementations supports non-ascii
 
1310
            # symlink targets. So we will just assume that the dirstate path is
 
1311
            # correct.
 
1312
            return entry[1][parent_index][1]
1823
1313
 
1824
1314
    def get_revision_id(self):
1825
1315
        """Return the revision id for this tree."""
1828
1318
    def _get_inventory(self):
1829
1319
        if self._inventory is not None:
1830
1320
            return self._inventory
1831
 
        self._must_be_locked()
1832
1321
        self._generate_inventory()
1833
1322
        return self._inventory
1834
1323
 
1843
1332
        return bool(self.path2id(filename))
1844
1333
 
1845
1334
    def kind(self, file_id):
1846
 
        entry = self._get_entry(file_id=file_id)[1]
1847
 
        if entry is None:
1848
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1849
 
        parent_index = self._get_parent_index()
1850
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1851
 
 
1852
 
    def stored_kind(self, file_id):
1853
 
        """See Tree.stored_kind"""
1854
 
        return self.kind(file_id)
1855
 
 
1856
 
    def path_content_summary(self, path):
1857
 
        """See Tree.path_content_summary."""
1858
 
        id = self.inventory.path2id(path)
1859
 
        if id is None:
1860
 
            return ('missing', None, None, None)
1861
 
        entry = self._inventory[id]
1862
 
        kind = entry.kind
1863
 
        if kind == 'file':
1864
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1865
 
        elif kind == 'symlink':
1866
 
            return (kind, None, None, entry.symlink_target)
1867
 
        else:
1868
 
            return (kind, None, None, None)
 
1335
        return self.inventory[file_id].kind
1869
1336
 
1870
1337
    def is_executable(self, file_id, path=None):
1871
1338
        ie = self.inventory[file_id]
1873
1340
            return None
1874
1341
        return ie.executable
1875
1342
 
1876
 
    def is_locked(self):
1877
 
        return self._locked
1878
 
 
1879
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1343
    def list_files(self, include_root=False):
1880
1344
        # We use a standard implementation, because DirStateRevisionTree is
1881
1345
        # dealing with one of the parents of the current state
1882
1346
        inv = self._get_inventory()
1883
 
        if from_dir is None:
1884
 
            from_dir_id = None
1885
 
        else:
1886
 
            from_dir_id = inv.path2id(from_dir)
1887
 
            if from_dir_id is None:
1888
 
                # Directory not versioned
1889
 
                return
1890
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1891
 
        if inv.root is not None and not include_root and from_dir is None:
 
1347
        entries = inv.iter_entries()
 
1348
        if self.inventory.root is not None and not include_root:
1892
1349
            entries.next()
1893
1350
        for path, entry in entries:
1894
1351
            yield path, 'V', entry.kind, entry.file_id, entry
1895
1352
 
1896
1353
    def lock_read(self):
1897
 
        """Lock the tree for a set of operations.
1898
 
 
1899
 
        :return: An object with an unlock method which will release the lock
1900
 
            obtained.
1901
 
        """
 
1354
        """Lock the tree for a set of operations."""
1902
1355
        if not self._locked:
1903
1356
            self._repository.lock_read()
1904
1357
            if self._dirstate._lock_token is None:
1905
1358
                self._dirstate.lock_read()
1906
1359
                self._dirstate_locked = True
1907
1360
        self._locked += 1
1908
 
        return self
1909
 
 
1910
 
    def _must_be_locked(self):
1911
 
        if not self._locked:
1912
 
            raise errors.ObjectNotLocked(self)
1913
1361
 
1914
1362
    @needs_read_lock
1915
1363
    def path2id(self, path):
1932
1380
                self._dirstate_locked = False
1933
1381
            self._repository.unlock()
1934
1382
 
1935
 
    @needs_read_lock
1936
 
    def supports_tree_reference(self):
1937
 
        return self._repo_supports_tree_reference
1938
 
 
1939
1383
    def walkdirs(self, prefix=""):
1940
 
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1941
 
        # implementation based on an inventory.
1942
 
        # This should be cleaned up to use the much faster Dirstate code
1943
 
        # So for now, we just build up the parent inventory, and extract
1944
 
        # it the same way RevisionTree does.
 
1384
        # TODO: jam 20070215 This is the cheap way by cheating and using the
 
1385
        #       RevisionTree implementation.
 
1386
        #       This should be cleaned up to use the much faster Dirstate code
 
1387
        #       This is a little tricky, though, because the dirstate is
 
1388
        #       indexed by current path, not by parent path.
 
1389
        #       So for now, we just build up the parent inventory, and extract
 
1390
        #       it the same way RevisionTree does.
1945
1391
        _directory = 'directory'
1946
1392
        inv = self._get_inventory()
1947
1393
        top_id = inv.path2id(prefix)
1972
1418
 
1973
1419
 
1974
1420
class InterDirStateTree(InterTree):
1975
 
    """Fast path optimiser for changes_from with dirstate trees.
1976
 
 
1977
 
    This is used only when both trees are in the dirstate working file, and
1978
 
    the source is any parent within the dirstate, and the destination is
1979
 
    the current working tree of the same dirstate.
1980
 
    """
1981
 
    # this could be generalized to allow comparisons between any trees in the
1982
 
    # dirstate, and possibly between trees stored in different dirstates.
 
1421
    """Fast path optimiser for changes_from with dirstate trees."""
1983
1422
 
1984
1423
    def __init__(self, source, target):
1985
1424
        super(InterDirStateTree, self).__init__(source, target)
1994
1433
        target.set_parent_ids([revid])
1995
1434
        return target.basis_tree(), target
1996
1435
 
1997
 
    @classmethod
1998
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1999
 
        result = klass.make_source_parent_tree(source, target)
2000
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2001
 
        return result
2002
 
 
2003
 
    @classmethod
2004
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2005
 
                                                  target):
2006
 
        from bzrlib.tests.test__dirstate_helpers import \
2007
 
            compiled_dirstate_helpers_feature
2008
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2009
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2010
 
        result = klass.make_source_parent_tree(source, target)
2011
 
        result[1]._iter_changes = ProcessEntryC
2012
 
        return result
2013
 
 
2014
1436
    _matching_from_tree_format = WorkingTreeFormat4()
2015
1437
    _matching_to_tree_format = WorkingTreeFormat4()
2016
 
 
2017
 
    @classmethod
2018
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2019
 
        # This method shouldn't be called, because we have python and C
2020
 
        # specific flavours.
2021
 
        raise NotImplementedError
2022
 
 
2023
 
    def iter_changes(self, include_unchanged=False,
 
1438
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1439
 
 
1440
    def _iter_changes(self, include_unchanged=False,
2024
1441
                      specific_files=None, pb=None, extra_trees=[],
2025
1442
                      require_versioned=True, want_unversioned=False):
2026
1443
        """Return the changes from source to target.
2027
1444
 
2028
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1445
        :return: An iterator that yields tuples. See InterTree._iter_changes
2029
1446
            for details.
2030
1447
        :param specific_files: An optional list of file paths to restrict the
2031
1448
            comparison to. When mapping filenames to ids, all matches in all
2042
1459
            output. An unversioned file is defined as one with (False, False)
2043
1460
            for the versioned pair.
2044
1461
        """
2045
 
        # TODO: handle extra trees in the dirstate.
2046
 
        if (extra_trees or specific_files == []):
2047
 
            # we can't fast-path these cases (yet)
2048
 
            return super(InterDirStateTree, self).iter_changes(
 
1462
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1463
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1464
        # NB: show_status depends on being able to pass in non-versioned files
 
1465
        # and report them as unknown
 
1466
            # TODO: handle extra trees in the dirstate.
 
1467
        if extra_trees:
 
1468
            for f in super(InterDirStateTree, self)._iter_changes(
2049
1469
                include_unchanged, specific_files, pb, extra_trees,
2050
 
                require_versioned, want_unversioned=want_unversioned)
 
1470
                require_versioned):
 
1471
                yield f
 
1472
            return
2051
1473
        parent_ids = self.target.get_parent_ids()
2052
 
        if not (self.source._revision_id in parent_ids
2053
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2054
 
            raise AssertionError(
2055
 
                "revision {%s} is not stored in {%s}, but %s "
2056
 
                "can only be used for trees stored in the dirstate"
2057
 
                % (self.source._revision_id, self.target, self.iter_changes))
2058
1474
        target_index = 0
2059
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1475
        if self.source._revision_id == NULL_REVISION:
2060
1476
            source_index = None
2061
1477
            indices = (target_index,)
2062
1478
        else:
2063
 
            if not (self.source._revision_id in parent_ids):
2064
 
                raise AssertionError(
2065
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2066
 
                    self.source._revision_id, parent_ids))
 
1479
            assert (self.source._revision_id in parent_ids), \
 
1480
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1481
                self.source._revision_id, parent_ids)
2067
1482
            source_index = 1 + parent_ids.index(self.source._revision_id)
2068
 
            indices = (source_index, target_index)
 
1483
            indices = (source_index,target_index)
2069
1484
        # -- make all specific_files utf8 --
2070
1485
        if specific_files:
2071
1486
            specific_files_utf8 = set()
2072
1487
            for path in specific_files:
2073
 
                # Note, if there are many specific files, using cache_utf8
2074
 
                # would be good here.
2075
1488
                specific_files_utf8.add(path.encode('utf8'))
2076
1489
            specific_files = specific_files_utf8
2077
1490
        else:
2078
1491
            specific_files = set([''])
2079
1492
        # -- specific_files is now a utf8 path set --
2080
 
 
2081
1493
        # -- get the state object and prepare it.
2082
1494
        state = self.target.current_dirstate()
2083
1495
        state._read_dirblocks_if_needed()
 
1496
        def _entries_for_path(path):
 
1497
            """Return a list with all the entries that match path for all ids.
 
1498
            """
 
1499
            dirname, basename = os.path.split(path)
 
1500
            key = (dirname, basename, '')
 
1501
            block_index, present = state._find_block_index_from_key(key)
 
1502
            if not present:
 
1503
                # the block which should contain path is absent.
 
1504
                return []
 
1505
            result = []
 
1506
            block = state._dirblocks[block_index][1]
 
1507
            entry_index, _ = state._find_entry_index(key, block)
 
1508
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1509
            while (entry_index < len(block) and
 
1510
                block[entry_index][0][0:2] == key[0:2]):
 
1511
                result.append(block[entry_index])
 
1512
                entry_index += 1
 
1513
            return result
2084
1514
        if require_versioned:
2085
1515
            # -- check all supplied paths are versioned in a search tree. --
2086
 
            not_versioned = []
 
1516
            all_versioned = True
2087
1517
            for path in specific_files:
2088
 
                path_entries = state._entries_for_path(path)
 
1518
                path_entries = _entries_for_path(path)
2089
1519
                if not path_entries:
2090
1520
                    # this specified path is not present at all: error
2091
 
                    not_versioned.append(path)
2092
 
                    continue
 
1521
                    all_versioned = False
 
1522
                    break
2093
1523
                found_versioned = False
2094
1524
                # for each id at this path
2095
1525
                for entry in path_entries:
2102
1532
                if not found_versioned:
2103
1533
                    # none of the indexes was not 'absent' at all ids for this
2104
1534
                    # path.
2105
 
                    not_versioned.append(path)
2106
 
            if len(not_versioned) > 0:
2107
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1535
                    all_versioned = False
 
1536
                    break
 
1537
            if not all_versioned:
 
1538
                raise errors.PathsNotVersionedError(specific_files)
2108
1539
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2109
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
2110
 
 
2111
 
        use_filesystem_for_exec = (sys.platform != 'win32')
2112
 
        iter_changes = self.target._iter_changes(include_unchanged,
2113
 
            use_filesystem_for_exec, search_specific_files, state,
2114
 
            source_index, target_index, want_unversioned, self.target)
2115
 
        return iter_changes.iter_changes()
 
1540
        search_specific_files = set()
 
1541
        for path in specific_files:
 
1542
            other_specific_files = specific_files.difference(set([path]))
 
1543
            if not osutils.is_inside_any(other_specific_files, path):
 
1544
                # this is a top level path, we must check it.
 
1545
                search_specific_files.add(path)
 
1546
        # sketch: 
 
1547
        # compare source_index and target_index at or under each element of search_specific_files.
 
1548
        # follow the following comparison table. Note that we only want to do diff operations when
 
1549
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1550
        # for the target.
 
1551
        # cases:
 
1552
        # 
 
1553
        # Source | Target | disk | action
 
1554
        #   r    | fdl    |      | add source to search, add id path move and perform
 
1555
        #        |        |      | diff check on source-target
 
1556
        #   r    | fdl    |  a   | dangling file that was present in the basis. 
 
1557
        #        |        |      | ???
 
1558
        #   r    |  a     |      | add source to search
 
1559
        #   r    |  a     |  a   | 
 
1560
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1561
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1562
        #   a    | fdl    |      | add new id
 
1563
        #   a    | fdl    |  a   | dangling locally added file, skip
 
1564
        #   a    |  a     |      | not present in either tree, skip
 
1565
        #   a    |  a     |  a   | not present in any tree, skip
 
1566
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1567
        #        |        |      | may not be selected by the users list of paths.
 
1568
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1569
        #        |        |      | may not be selected by the users list of paths.
 
1570
        #  fdl   | fdl    |      | content in both: diff them
 
1571
        #  fdl   | fdl    |  a   | deleted locally, but not unversioned - show as deleted ?
 
1572
        #  fdl   |  a     |      | unversioned: output deleted id for now
 
1573
        #  fdl   |  a     |  a   | unversioned and deleted: output deleted id
 
1574
        #  fdl   |  r     |      | relocated in this tree, so add target to search.
 
1575
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1576
        #        |        |      | this id at the other path.
 
1577
        #  fdl   |  r     |  a   | relocated in this tree, so add target to search.
 
1578
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1579
        #        |        |      | this id at the other path.
 
1580
 
 
1581
        # for all search_indexs in each path at or under each element of
 
1582
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1583
        # relocated path as one to search if its not searched already. If the
 
1584
        # detail is not relocated, add the id.
 
1585
        searched_specific_files = set()
 
1586
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1587
        # Using a list so that we can access the values and change them in
 
1588
        # nested scope. Each one is [path, file_id, entry]
 
1589
        last_source_parent = [None, None, None]
 
1590
        last_target_parent = [None, None, None]
 
1591
 
 
1592
        def _process_entry(entry, path_info):
 
1593
            """Compare an entry and real disk to generate delta information.
 
1594
 
 
1595
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1596
                the path of entry. If None, then the path is considered absent.
 
1597
                (Perhaps we should pass in a concrete entry for this ?)
 
1598
                Basename is returned as a utf8 string because we expect this
 
1599
                tuple will be ignored, and don't want to take the time to
 
1600
                decode.
 
1601
            """
 
1602
            # TODO: when a parent has been renamed, dont emit path renames for children,
 
1603
            if source_index is None:
 
1604
                source_details = NULL_PARENT_DETAILS
 
1605
            else:
 
1606
                source_details = entry[1][source_index]
 
1607
            target_details = entry[1][target_index]
 
1608
            target_minikind = target_details[0]
 
1609
            if path_info is not None and target_minikind in 'fdl':
 
1610
                assert target_index == 0
 
1611
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1612
                                                  stat_value=path_info[3])
 
1613
                # The entry may have been modified by update_entry
 
1614
                target_details = entry[1][target_index]
 
1615
                target_minikind = target_details[0]
 
1616
            else:
 
1617
                link_or_sha1 = None
 
1618
            source_minikind = source_details[0]
 
1619
            if source_minikind in 'fdlr' and target_minikind in 'fdl':
 
1620
                # claimed content in both: diff
 
1621
                #   r    | fdl    |      | add source to search, add id path move and perform
 
1622
                #        |        |      | diff check on source-target
 
1623
                #   r    | fdl    |  a   | dangling file that was present in the basis.
 
1624
                #        |        |      | ???
 
1625
                if source_minikind in 'r':
 
1626
                    # add the source to the search path to find any children it
 
1627
                    # has.  TODO ? : only add if it is a container ?
 
1628
                    if not osutils.is_inside_any(searched_specific_files,
 
1629
                                                 source_details[1]):
 
1630
                        search_specific_files.add(source_details[1])
 
1631
                    # generate the old path; this is needed for stating later
 
1632
                    # as well.
 
1633
                    old_path = source_details[1]
 
1634
                    old_dirname, old_basename = os.path.split(old_path)
 
1635
                    path = pathjoin(entry[0][0], entry[0][1])
 
1636
                    old_entry = state._get_entry(source_index,
 
1637
                                                 path_utf8=old_path)
 
1638
                    # update the source details variable to be the real
 
1639
                    # location.
 
1640
                    source_details = old_entry[1][source_index]
 
1641
                    source_minikind = source_details[0]
 
1642
                else:
 
1643
                    old_dirname = entry[0][0]
 
1644
                    old_basename = entry[0][1]
 
1645
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1646
                if path_info is None:
 
1647
                    # the file is missing on disk, show as removed.
 
1648
                    content_change = True
 
1649
                    target_kind = None
 
1650
                    target_exec = False
 
1651
                else:
 
1652
                    # source and target are both versioned and disk file is present.
 
1653
                    target_kind = path_info[2]
 
1654
                    if target_kind == 'directory':
 
1655
                        if source_minikind != 'd':
 
1656
                            content_change = True
 
1657
                        else:
 
1658
                            # directories have no fingerprint
 
1659
                            content_change = False
 
1660
                        target_exec = False
 
1661
                    elif target_kind == 'file':
 
1662
                        if source_minikind != 'f':
 
1663
                            content_change = True
 
1664
                        else:
 
1665
                            # We could check the size, but we already have the
 
1666
                            # sha1 hash.
 
1667
                            content_change = (link_or_sha1 != source_details[1])
 
1668
                        # Target details is updated at update_entry time
 
1669
                        target_exec = bool(
 
1670
                            stat.S_ISREG(path_info[3].st_mode)
 
1671
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1672
                    elif target_kind == 'symlink':
 
1673
                        if source_minikind != 'l':
 
1674
                            content_change = True
 
1675
                        else:
 
1676
                            content_change = (link_or_sha1 != source_details[1])
 
1677
                        target_exec = False
 
1678
                    else:
 
1679
                        raise Exception, "unknown kind %s" % path_info[2]
 
1680
                # parent id is the entry for the path in the target tree
 
1681
                if old_dirname == last_source_parent[0]:
 
1682
                    source_parent_id = last_source_parent[1]
 
1683
                else:
 
1684
                    source_parent_entry = state._get_entry(source_index,
 
1685
                                                           path_utf8=old_dirname)
 
1686
                    source_parent_id = source_parent_entry[0][2]
 
1687
                    if source_parent_id == entry[0][2]:
 
1688
                        # This is the root, so the parent is None
 
1689
                        source_parent_id = None
 
1690
                    else:
 
1691
                        last_source_parent[0] = old_dirname
 
1692
                        last_source_parent[1] = source_parent_id
 
1693
                        last_source_parent[2] = source_parent_entry
 
1694
 
 
1695
                new_dirname = entry[0][0]
 
1696
                if new_dirname == last_target_parent[0]:
 
1697
                    target_parent_id = last_target_parent[1]
 
1698
                else:
 
1699
                    # TODO: We don't always need to do the lookup, because the
 
1700
                    #       parent entry will be the same as the source entry.
 
1701
                    target_parent_entry = state._get_entry(target_index,
 
1702
                                                           path_utf8=new_dirname)
 
1703
                    target_parent_id = target_parent_entry[0][2]
 
1704
                    if target_parent_id == entry[0][2]:
 
1705
                        # This is the root, so the parent is None
 
1706
                        target_parent_id = None
 
1707
                    else:
 
1708
                        last_target_parent[0] = new_dirname
 
1709
                        last_target_parent[1] = target_parent_id
 
1710
                        last_target_parent[2] = target_parent_entry
 
1711
 
 
1712
                source_exec = source_details[3]
 
1713
                return ((entry[0][2], (old_path, path), content_change,
 
1714
                        (True, True),
 
1715
                        (source_parent_id, target_parent_id),
 
1716
                        (old_basename, entry[0][1]),
 
1717
                        (_minikind_to_kind[source_minikind], target_kind),
 
1718
                        (source_exec, target_exec)),)
 
1719
            elif source_minikind in 'a' and target_minikind in 'fdl':
 
1720
                # looks like a new file
 
1721
                if path_info is not None:
 
1722
                    path = pathjoin(entry[0][0], entry[0][1])
 
1723
                    # parent id is the entry for the path in the target tree
 
1724
                    # TODO: these are the same for an entire directory: cache em.
 
1725
                    parent_id = state._get_entry(target_index,
 
1726
                                                 path_utf8=entry[0][0])[0][2]
 
1727
                    if parent_id == entry[0][2]:
 
1728
                        parent_id = None
 
1729
                    target_exec = bool(
 
1730
                        stat.S_ISREG(path_info[3].st_mode)
 
1731
                        and stat.S_IEXEC & path_info[3].st_mode)
 
1732
                    return ((entry[0][2], (None, path), True,
 
1733
                            (False, True),
 
1734
                            (None, parent_id),
 
1735
                            (None, entry[0][1]),
 
1736
                            (None, path_info[2]),
 
1737
                            (None, target_exec)),)
 
1738
                else:
 
1739
                    # but its not on disk: we deliberately treat this as just
 
1740
                    # never-present. (Why ?! - RBC 20070224)
 
1741
                    pass
 
1742
            elif source_minikind in 'fdl' and target_minikind in 'a':
 
1743
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1744
                # if its still on disk, *and* theres no other entry at this
 
1745
                # path [we dont know this in this routine at the moment -
 
1746
                # perhaps we should change this - then it would be an unknown.
 
1747
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1748
                # parent id is the entry for the path in the target tree
 
1749
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1750
                if parent_id == entry[0][2]:
 
1751
                    parent_id = None
 
1752
                return ((entry[0][2], (old_path, None), True,
 
1753
                        (True, False),
 
1754
                        (parent_id, None),
 
1755
                        (entry[0][1], None),
 
1756
                        (_minikind_to_kind[source_minikind], None),
 
1757
                        (source_details[3], None)),)
 
1758
            elif source_minikind in 'fdl' and target_minikind in 'r':
 
1759
                # a rename; could be a true rename, or a rename inherited from
 
1760
                # a renamed parent. TODO: handle this efficiently. Its not
 
1761
                # common case to rename dirs though, so a correct but slow
 
1762
                # implementation will do.
 
1763
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1764
                    search_specific_files.add(target_details[1])
 
1765
            elif source_minikind in 'r' and target_minikind in 'r':
 
1766
                # neither of the selected trees contain this file,
 
1767
                # so skip over it. This is not currently directly tested, but
 
1768
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1769
                pass
 
1770
            else:
 
1771
                print "*******", source_minikind, target_minikind
 
1772
                import pdb;pdb.set_trace()
 
1773
            return ()
 
1774
        while search_specific_files:
 
1775
            # TODO: the pending list should be lexically sorted?
 
1776
            current_root = search_specific_files.pop()
 
1777
            searched_specific_files.add(current_root)
 
1778
            # process the entries for this containing directory: the rest will be
 
1779
            # found by their parents recursively.
 
1780
            root_entries = _entries_for_path(current_root)
 
1781
            root_abspath = self.target.abspath(current_root)
 
1782
            try:
 
1783
                root_stat = os.lstat(root_abspath)
 
1784
            except OSError, e:
 
1785
                if e.errno == errno.ENOENT:
 
1786
                    # the path does not exist: let _process_entry know that.
 
1787
                    root_dir_info = None
 
1788
                else:
 
1789
                    # some other random error: hand it up.
 
1790
                    raise
 
1791
            else:
 
1792
                root_dir_info = ('', current_root,
 
1793
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1794
                    root_abspath)
 
1795
            if not root_entries and not root_dir_info:
 
1796
                # this specified path is not present at all, skip it.
 
1797
                continue
 
1798
            path_handled = False
 
1799
            for entry in root_entries:
 
1800
                for result in _process_entry(entry, root_dir_info):
 
1801
                    # this check should probably be outside the loop: one
 
1802
                    # 'iterate two trees' api, and then _iter_changes filters
 
1803
                    # unchanged pairs. - RBC 20070226
 
1804
                    path_handled = True
 
1805
                    if (include_unchanged
 
1806
                        or result[2]                    # content change
 
1807
                        or result[3][0] != result[3][1] # versioned status
 
1808
                        or result[4][0] != result[4][1] # parent id
 
1809
                        or result[5][0] != result[5][1] # name
 
1810
                        or result[6][0] != result[6][1] # kind
 
1811
                        or result[7][0] != result[7][1] # executable
 
1812
                        ):
 
1813
                        result = (result[0],
 
1814
                            ((utf8_decode(result[1][0])[0]),
 
1815
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
1816
                        yield result
 
1817
            if want_unversioned and not path_handled:
 
1818
                new_executable = bool(
 
1819
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
1820
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
1821
                yield (None, (None, current_root), True, (False, False),
 
1822
                    (None, None),
 
1823
                    (None, splitpath(current_root)[-1]),
 
1824
                    (None, root_dir_info[2]), (None, new_executable))
 
1825
            dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
1826
            initial_key = (current_root, '', '')
 
1827
            block_index, _ = state._find_block_index_from_key(initial_key)
 
1828
            if block_index == 0:
 
1829
                # we have processed the total root already, but because the
 
1830
                # initial key matched it we should skip it here.
 
1831
                block_index +=1
 
1832
            try:
 
1833
                current_dir_info = dir_iterator.next()
 
1834
            except OSError, e:
 
1835
                if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
1836
                    # there may be directories in the inventory even though
 
1837
                    # this path is not a file on disk: so mark it as end of
 
1838
                    # iterator
 
1839
                    current_dir_info = None
 
1840
                else:
 
1841
                    raise
 
1842
            else:
 
1843
                if current_dir_info[0][0] == '':
 
1844
                    # remove .bzr from iteration
 
1845
                    bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
1846
                    assert current_dir_info[1][bzr_index][0] == '.bzr'
 
1847
                    del current_dir_info[1][bzr_index]
 
1848
            # walk until both the directory listing and the versioned metadata
 
1849
            # are exhausted. TODO: reevaluate this, perhaps we should stop when
 
1850
            # the versioned data runs out.
 
1851
            if (block_index < len(state._dirblocks) and
 
1852
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
1853
                current_block = state._dirblocks[block_index]
 
1854
            else:
 
1855
                current_block = None
 
1856
            while (current_dir_info is not None or
 
1857
                   current_block is not None):
 
1858
                if (current_dir_info and current_block
 
1859
                    and current_dir_info[0][0] != current_block[0]):
 
1860
                    if current_dir_info[0][0] < current_block[0] :
 
1861
                        # import pdb; pdb.set_trace()
 
1862
                        # print 'unversioned dir'
 
1863
                        # filesystem data refers to paths not covered by the dirblock.
 
1864
                        # this has two possibilities:
 
1865
                        # A) it is versioned but empty, so there is no block for it
 
1866
                        # B) it is not versioned.
 
1867
                        # in either case it was processed by the containing directories walk:
 
1868
                        # if it is root/foo, when we walked root we emitted it,
 
1869
                        # or if we ere given root/foo to walk specifically, we
 
1870
                        # emitted it when checking the walk-root entries
 
1871
                        # advance the iterator and loop - we dont need to emit it.
 
1872
                        try:
 
1873
                            current_dir_info = dir_iterator.next()
 
1874
                        except StopIteration:
 
1875
                            current_dir_info = None
 
1876
                    else:
 
1877
                        # We have a dirblock entry for this location, but there
 
1878
                        # is no filesystem path for this. This is most likely
 
1879
                        # because a directory was removed from the disk.
 
1880
                        # We don't have to report the missing directory,
 
1881
                        # because that should have already been handled, but we
 
1882
                        # need to handle all of the files that are contained
 
1883
                        # within.
 
1884
                        for current_entry in current_block[1]:
 
1885
                            # entry referring to file not present on disk.
 
1886
                            # advance the entry only, after processing.
 
1887
                            for result in _process_entry(current_entry, None):
 
1888
                                # this check should probably be outside the loop: one
 
1889
                                # 'iterate two trees' api, and then _iter_changes filters
 
1890
                                # unchanged pairs. - RBC 20070226
 
1891
                                if (include_unchanged
 
1892
                                    or result[2]                    # content change
 
1893
                                    or result[3][0] != result[3][1] # versioned status
 
1894
                                    or result[4][0] != result[4][1] # parent id
 
1895
                                    or result[5][0] != result[5][1] # name
 
1896
                                    or result[6][0] != result[6][1] # kind
 
1897
                                    or result[7][0] != result[7][1] # executable
 
1898
                                    ):
 
1899
                                    result = (result[0],
 
1900
                                        ((utf8_decode(result[1][0])[0]),
 
1901
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
1902
                                    yield result
 
1903
                        block_index +=1
 
1904
                        if (block_index < len(state._dirblocks) and
 
1905
                            osutils.is_inside(current_root,
 
1906
                                              state._dirblocks[block_index][0])):
 
1907
                            current_block = state._dirblocks[block_index]
 
1908
                        else:
 
1909
                            current_block = None
 
1910
                    continue
 
1911
                entry_index = 0
 
1912
                if current_block and entry_index < len(current_block[1]):
 
1913
                    current_entry = current_block[1][entry_index]
 
1914
                else:
 
1915
                    current_entry = None
 
1916
                advance_entry = True
 
1917
                path_index = 0
 
1918
                if current_dir_info and path_index < len(current_dir_info[1]):
 
1919
                    current_path_info = current_dir_info[1][path_index]
 
1920
                else:
 
1921
                    current_path_info = None
 
1922
                advance_path = True
 
1923
                path_handled = False
 
1924
                while (current_entry is not None or
 
1925
                    current_path_info is not None):
 
1926
                    if current_entry is None:
 
1927
                        # the check for path_handled when the path is adnvaced
 
1928
                        # will yield this path if needed.
 
1929
                        pass
 
1930
                    elif current_path_info is None:
 
1931
                        # no path is fine: the per entry code will handle it.
 
1932
                        for result in _process_entry(current_entry, current_path_info):
 
1933
                            # this check should probably be outside the loop: one
 
1934
                            # 'iterate two trees' api, and then _iter_changes filters
 
1935
                            # unchanged pairs. - RBC 20070226
 
1936
                            if (include_unchanged
 
1937
                                or result[2]                    # content change
 
1938
                                or result[3][0] != result[3][1] # versioned status
 
1939
                                or result[4][0] != result[4][1] # parent id
 
1940
                                or result[5][0] != result[5][1] # name
 
1941
                                or result[6][0] != result[6][1] # kind
 
1942
                                or result[7][0] != result[7][1] # executable
 
1943
                                ):
 
1944
                                result = (result[0],
 
1945
                                    ((utf8_decode(result[1][0])[0]),
 
1946
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
1947
                                yield result
 
1948
                    elif current_entry[0][1] != current_path_info[1]:
 
1949
                        if current_path_info[1] < current_entry[0][1]:
 
1950
                            # extra file on disk: pass for now, but only
 
1951
                            # increment the path, not the entry
 
1952
                            # import pdb; pdb.set_trace()
 
1953
                            # print 'unversioned file'
 
1954
                            advance_entry = False
 
1955
                        else:
 
1956
                            # entry referring to file not present on disk.
 
1957
                            # advance the entry only, after processing.
 
1958
                            for result in _process_entry(current_entry, None):
 
1959
                                # this check should probably be outside the loop: one
 
1960
                                # 'iterate two trees' api, and then _iter_changes filters
 
1961
                                # unchanged pairs. - RBC 20070226
 
1962
                                path_handled = True
 
1963
                                if (include_unchanged
 
1964
                                    or result[2]                    # content change
 
1965
                                    or result[3][0] != result[3][1] # versioned status
 
1966
                                    or result[4][0] != result[4][1] # parent id
 
1967
                                    or result[5][0] != result[5][1] # name
 
1968
                                    or result[6][0] != result[6][1] # kind
 
1969
                                    or result[7][0] != result[7][1] # executable
 
1970
                                    ):
 
1971
                                    result = (result[0],
 
1972
                                        ((utf8_decode(result[1][0])[0]),
 
1973
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
1974
                                    yield result
 
1975
                            advance_path = False
 
1976
                    else:
 
1977
                        for result in _process_entry(current_entry, current_path_info):
 
1978
                            # this check should probably be outside the loop: one
 
1979
                            # 'iterate two trees' api, and then _iter_changes filters
 
1980
                            # unchanged pairs. - RBC 20070226
 
1981
                            path_handled = True
 
1982
                            if (include_unchanged
 
1983
                                or result[2]                    # content change
 
1984
                                or result[3][0] != result[3][1] # versioned status
 
1985
                                or result[4][0] != result[4][1] # parent id
 
1986
                                or result[5][0] != result[5][1] # name
 
1987
                                or result[6][0] != result[6][1] # kind
 
1988
                                or result[7][0] != result[7][1] # executable
 
1989
                                ):
 
1990
                                result = (result[0],
 
1991
                                    ((utf8_decode(result[1][0])[0]),
 
1992
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
1993
                                yield result
 
1994
                    if advance_entry and current_entry is not None:
 
1995
                        entry_index += 1
 
1996
                        if entry_index < len(current_block[1]):
 
1997
                            current_entry = current_block[1][entry_index]
 
1998
                        else:
 
1999
                            current_entry = None
 
2000
                    else:
 
2001
                        advance_entry = True # reset the advance flaga
 
2002
                    if advance_path and current_path_info is not None:
 
2003
                        if not path_handled:
 
2004
                            # unversioned in all regards
 
2005
                            if want_unversioned:
 
2006
                                new_executable = bool(
 
2007
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2008
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2009
                                if want_unversioned:
 
2010
                                    yield (None, (None, current_path_info[0]),
 
2011
                                        True,
 
2012
                                        (False, False),
 
2013
                                        (None, None),
 
2014
                                        (None, current_path_info[1]),
 
2015
                                        (None, current_path_info[2]),
 
2016
                                        (None, new_executable))
 
2017
                            # dont descend into this unversioned path if it is
 
2018
                            # a dir
 
2019
                            if current_path_info[2] == 'directory':
 
2020
                                del current_dir_info[1][path_index]
 
2021
                                path_index -= 1
 
2022
                        path_index += 1
 
2023
                        if path_index < len(current_dir_info[1]):
 
2024
                            current_path_info = current_dir_info[1][path_index]
 
2025
                        else:
 
2026
                            current_path_info = None
 
2027
                        path_handled = False
 
2028
                    else:
 
2029
                        advance_path = True # reset the advance flagg.
 
2030
                if current_block is not None:
 
2031
                    block_index += 1
 
2032
                    if (block_index < len(state._dirblocks) and
 
2033
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2034
                        current_block = state._dirblocks[block_index]
 
2035
                    else:
 
2036
                        current_block = None
 
2037
                if current_dir_info is not None:
 
2038
                    try:
 
2039
                        current_dir_info = dir_iterator.next()
 
2040
                    except StopIteration:
 
2041
                        current_dir_info = None
 
2042
 
2116
2043
 
2117
2044
    @staticmethod
2118
2045
    def is_compatible(source, target):
2119
2046
        # the target must be a dirstate working tree
2120
 
        if not isinstance(target, DirStateWorkingTree):
 
2047
        if not isinstance(target, WorkingTree4):
2121
2048
            return False
2122
 
        # the source must be a revtree or dirstate rev tree.
 
2049
        # the source must be a revtreee or dirstate rev tree.
2123
2050
        if not isinstance(source,
2124
2051
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2125
2052
            return False
2126
2053
        # the source revid must be in the target dirstate
2127
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2054
        if not (source._revision_id == NULL_REVISION or
2128
2055
            source._revision_id in target.get_parent_ids()):
2129
 
            # TODO: what about ghosts? it may well need to
 
2056
            # TODO: what about ghosts? it may well need to 
2130
2057
            # check for them explicitly.
2131
2058
            return False
2132
2059
        return True
2133
2060
 
2134
2061
InterTree.register_optimiser(InterDirStateTree)
2135
 
 
2136
 
 
2137
 
class Converter3to4(object):
2138
 
    """Perform an in-place upgrade of format 3 to format 4 trees."""
2139
 
 
2140
 
    def __init__(self):
2141
 
        self.target_format = WorkingTreeFormat4()
2142
 
 
2143
 
    def convert(self, tree):
2144
 
        # lock the control files not the tree, so that we dont get tree
2145
 
        # on-unlock behaviours, and so that noone else diddles with the
2146
 
        # tree during upgrade.
2147
 
        tree._control_files.lock_write()
2148
 
        try:
2149
 
            tree.read_working_inventory()
2150
 
            self.create_dirstate_data(tree)
2151
 
            self.update_format(tree)
2152
 
            self.remove_xml_files(tree)
2153
 
        finally:
2154
 
            tree._control_files.unlock()
2155
 
 
2156
 
    def create_dirstate_data(self, tree):
2157
 
        """Create the dirstate based data for tree."""
2158
 
        local_path = tree.bzrdir.get_workingtree_transport(None
2159
 
            ).local_abspath('dirstate')
2160
 
        state = dirstate.DirState.from_tree(tree, local_path)
2161
 
        state.save()
2162
 
        state.unlock()
2163
 
 
2164
 
    def remove_xml_files(self, tree):
2165
 
        """Remove the oldformat 3 data."""
2166
 
        transport = tree.bzrdir.get_workingtree_transport(None)
2167
 
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
2168
 
            'pending-merges', 'stat-cache']:
2169
 
            try:
2170
 
                transport.delete(path)
2171
 
            except errors.NoSuchFile:
2172
 
                # some files are optional - just deal.
2173
 
                pass
2174
 
 
2175
 
    def update_format(self, tree):
2176
 
        """Change the format marker."""
2177
 
        tree._transport.put_bytes('format',
2178
 
            self.target_format.get_format_string(),
2179
 
            mode=tree.bzrdir._get_file_mode())
2180
 
 
2181
 
 
2182
 
class Converter4to5(object):
2183
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2184
 
 
2185
 
    def __init__(self):
2186
 
        self.target_format = WorkingTreeFormat5()
2187
 
 
2188
 
    def convert(self, tree):
2189
 
        # lock the control files not the tree, so that we don't get tree
2190
 
        # on-unlock behaviours, and so that no-one else diddles with the
2191
 
        # tree during upgrade.
2192
 
        tree._control_files.lock_write()
2193
 
        try:
2194
 
            self.update_format(tree)
2195
 
        finally:
2196
 
            tree._control_files.unlock()
2197
 
 
2198
 
    def update_format(self, tree):
2199
 
        """Change the format marker."""
2200
 
        tree._transport.put_bytes('format',
2201
 
            self.target_format.get_format_string(),
2202
 
            mode=tree.bzrdir._get_file_mode())
2203
 
 
2204
 
 
2205
 
class Converter4or5to6(object):
2206
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2207
 
 
2208
 
    def __init__(self):
2209
 
        self.target_format = WorkingTreeFormat6()
2210
 
 
2211
 
    def convert(self, tree):
2212
 
        # lock the control files not the tree, so that we don't get tree
2213
 
        # on-unlock behaviours, and so that no-one else diddles with the
2214
 
        # tree during upgrade.
2215
 
        tree._control_files.lock_write()
2216
 
        try:
2217
 
            self.init_custom_control_files(tree)
2218
 
            self.update_format(tree)
2219
 
        finally:
2220
 
            tree._control_files.unlock()
2221
 
 
2222
 
    def init_custom_control_files(self, tree):
2223
 
        """Initialize custom control files."""
2224
 
        tree._transport.put_bytes('views', '',
2225
 
            mode=tree.bzrdir._get_file_mode())
2226
 
 
2227
 
    def update_format(self, tree):
2228
 
        """Change the format marker."""
2229
 
        tree._transport.put_bytes('format',
2230
 
            self.target_format.get_format_string(),
2231
 
            mode=tree.bzrdir._get_file_mode())