/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: Martin Pool
  • Date: 2007-03-06 05:05:46 UTC
  • mto: (2255.2.213 dirstate.dogfood)
  • mto: This revision was merged to the branch mainline in revision 2322.
  • Revision ID: mbp@sourcefrog.net-20070306050546-3lujsd390sq65um0
Add BzrDir.retire_bzrdir and partly fix subsume

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
 
47
    dirstate,
 
48
    errors,
 
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revisiontree,
 
56
    textui,
 
57
    transform,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
 
61
    )
 
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
 
64
import bzrlib.ui
 
65
""")
 
66
 
 
67
from bzrlib import symbol_versioning
 
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
 
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
 
73
from bzrlib.mutabletree import needs_tree_write_lock
 
74
from bzrlib.osutils import (
 
75
    file_kind,
 
76
    isdir,
 
77
    normpath,
 
78
    pathjoin,
 
79
    rand_chars,
 
80
    realpath,
 
81
    safe_unicode,
 
82
    splitpath,
 
83
    )
 
84
from bzrlib.trace import mutter, note
 
85
from bzrlib.transport.local import LocalTransport
 
86
from bzrlib.tree import InterTree
 
87
from bzrlib.progress import DummyProgress, ProgressPhase
 
88
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
89
from bzrlib.rio import RioReader, rio_file, Stanza
 
90
from bzrlib.symbol_versioning import (deprecated_passed,
 
91
        deprecated_method,
 
92
        deprecated_function,
 
93
        DEPRECATED_PARAMETER,
 
94
        )
 
95
from bzrlib.tree import Tree
 
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
97
 
 
98
 
 
99
class WorkingTree4(WorkingTree3):
 
100
    """This is the Format 4 working tree.
 
101
 
 
102
    This differs from WorkingTree3 by:
 
103
     - Having a consolidated internal dirstate, stored in a
 
104
       randomly-accessible sorted file on disk.
 
105
     - Not having a regular inventory attribute.  One can be synthesized 
 
106
       on demand but this is expensive and should be avoided.
 
107
 
 
108
    This is new in bzr 0.15.
 
109
    """
 
110
 
 
111
    def __init__(self, basedir,
 
112
                 branch,
 
113
                 _control_files=None,
 
114
                 _format=None,
 
115
                 _bzrdir=None):
 
116
        """Construct a WorkingTree for basedir.
 
117
 
 
118
        If the branch is not supplied, it is opened automatically.
 
119
        If the branch is supplied, it must be the branch for this basedir.
 
120
        (branch.base is not cross checked, because for remote branches that
 
121
        would be meaningless).
 
122
        """
 
123
        self._format = _format
 
124
        self.bzrdir = _bzrdir
 
125
        from bzrlib.trace import note, mutter
 
126
        assert isinstance(basedir, basestring), \
 
127
            "base directory %r is not a string" % basedir
 
128
        basedir = safe_unicode(basedir)
 
129
        mutter("opening working tree %r", basedir)
 
130
        self._branch = branch
 
131
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
132
            "branch %r is not a Branch" % self.branch
 
133
        self.basedir = realpath(basedir)
 
134
        # if branch is at our basedir and is a format 6 or less
 
135
        # assume all other formats have their own control files.
 
136
        assert isinstance(_control_files, LockableFiles), \
 
137
            "_control_files must be a LockableFiles, not %r" % _control_files
 
138
        self._control_files = _control_files
 
139
        self._dirty = None
 
140
        #-------------
 
141
        # during a read or write lock these objects are set, and are
 
142
        # None the rest of the time.
 
143
        self._dirstate = None
 
144
        self._inventory = None
 
145
        #-------------
 
146
 
 
147
    @needs_tree_write_lock
 
148
    def _add(self, files, ids, kinds):
 
149
        """See MutableTree._add."""
 
150
        state = self.current_dirstate()
 
151
        for f, file_id, kind in zip(files, ids, kinds):
 
152
            f = f.strip('/')
 
153
            assert '//' not in f
 
154
            assert '..' not in f
 
155
            if self.path2id(f):
 
156
                # special case tree root handling.
 
157
                if f == '' and self.path2id(f) == ROOT_ID:
 
158
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
159
                continue
 
160
            if file_id is None:
 
161
                file_id = generate_ids.gen_file_id(f)
 
162
            # deliberately add the file with no cached stat or sha1
 
163
            # - on the first access it will be gathered, and we can
 
164
            # always change this once tests are all passing.
 
165
            state.add(f, file_id, kind, None, '')
 
166
        self._make_dirty(reset_inventory=True)
 
167
 
 
168
    def _make_dirty(self, reset_inventory):
 
169
        """Make the tree state dirty.
 
170
 
 
171
        :param reset_inventory: True if the cached inventory should be removed
 
172
            (presuming there is one).
 
173
        """
 
174
        self._dirty = True
 
175
        if reset_inventory and self._inventory is not None:
 
176
            self._inventory = None
 
177
 
 
178
    @needs_tree_write_lock
 
179
    def add_reference(self, sub_tree):
 
180
        # use standard implementation, which calls back to self._add
 
181
        # 
 
182
        # So we don't store the reference_revision in the working dirstate,
 
183
        # it's just recorded at the moment of commit. 
 
184
        self._add_reference(sub_tree)
 
185
 
 
186
    def break_lock(self):
 
187
        """Break a lock if one is present from another instance.
 
188
 
 
189
        Uses the ui factory to ask for confirmation if the lock may be from
 
190
        an active process.
 
191
 
 
192
        This will probe the repository for its lock as well.
 
193
        """
 
194
        # if the dirstate is locked by an active process, reject the break lock
 
195
        # call.
 
196
        try:
 
197
            if self._dirstate is None:
 
198
                clear = True
 
199
            else:
 
200
                clear = False
 
201
            state = self._current_dirstate()
 
202
            if state._lock_token is not None:
 
203
                # we already have it locked. sheese, cant break our own lock.
 
204
                raise errors.LockActive(self.basedir)
 
205
            else:
 
206
                try:
 
207
                    # try for a write lock - need permission to get one anyhow
 
208
                    # to break locks.
 
209
                    state.lock_write()
 
210
                except errors.LockContention:
 
211
                    # oslocks fail when a process is still live: fail.
 
212
                    # TODO: get the locked lockdir info and give to the user to
 
213
                    # assist in debugging.
 
214
                    raise errors.LockActive(self.basedir)
 
215
                else:
 
216
                    state.unlock()
 
217
        finally:
 
218
            if clear:
 
219
                self._dirstate = None
 
220
        self._control_files.break_lock()
 
221
        self.branch.break_lock()
 
222
 
 
223
    def _comparison_data(self, entry, path):
 
224
        kind, executable, stat_value = \
 
225
            WorkingTree3._comparison_data(self, entry, path)
 
226
        # it looks like a plain directory, but it's really a reference -- see
 
227
        # also kind()
 
228
        if kind == 'directory' and self._directory_is_tree_reference(path):
 
229
            kind = 'tree-reference'
 
230
        return kind, executable, stat_value
 
231
 
 
232
    @needs_write_lock
 
233
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
234
        # mark the tree as dirty post commit - commit
 
235
        # can change the current versioned list by doing deletes.
 
236
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
237
        self._make_dirty(reset_inventory=True)
 
238
        return result
 
239
 
 
240
    def current_dirstate(self):
 
241
        """Return the current dirstate object.
 
242
 
 
243
        This is not part of the tree interface and only exposed for ease of
 
244
        testing.
 
245
 
 
246
        :raises errors.NotWriteLocked: when not in a lock.
 
247
        """
 
248
        self._must_be_locked()
 
249
        return self._current_dirstate()
 
250
 
 
251
    def _current_dirstate(self):
 
252
        """Internal function that does not check lock status.
 
253
 
 
254
        This is needed for break_lock which also needs the dirstate.
 
255
        """
 
256
        if self._dirstate is not None:
 
257
            return self._dirstate
 
258
        local_path = self.bzrdir.get_workingtree_transport(None
 
259
            ).local_abspath('dirstate')
 
260
        self._dirstate = dirstate.DirState.on_file(local_path)
 
261
        return self._dirstate
 
262
 
 
263
    def _directory_is_tree_reference(self, relpath):
 
264
        # as a special case, if a directory contains control files then 
 
265
        # it's a tree reference, except that the root of the tree is not
 
266
        return relpath and osutils.isdir(self.abspath(relpath) + "/.bzr")
 
267
        # TODO: We could ask all the control formats whether they
 
268
        # recognize this directory, but at the moment there's no cheap api
 
269
        # to do that.  Since we probably can only nest bzr checkouts and
 
270
        # they always use this name it's ok for now.  -- mbp 20060306
 
271
        #
 
272
        # FIXME: There is an unhandled case here of a subdirectory
 
273
        # containing .bzr but not a branch; that will probably blow up
 
274
        # when you try to commit it.  It might happen if there is a
 
275
        # checkout in a subdirectory.  This can be avoided by not adding
 
276
        # it.  mbp 20070306
 
277
 
 
278
    def filter_unversioned_files(self, paths):
 
279
        """Filter out paths that are versioned.
 
280
 
 
281
        :return: set of paths.
 
282
        """
 
283
        # TODO: make a generic multi-bisect routine roughly that should list
 
284
        # the paths, then process one half at a time recursively, and feed the
 
285
        # results of each bisect in further still
 
286
        paths = sorted(paths)
 
287
        result = set()
 
288
        state = self.current_dirstate()
 
289
        # TODO we want a paths_to_dirblocks helper I think
 
290
        for path in paths:
 
291
            dirname, basename = os.path.split(path.encode('utf8'))
 
292
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
293
                dirname, basename, 0)
 
294
            if not path_is_versioned:
 
295
                result.add(path)
 
296
        return result
 
297
 
 
298
    def flush(self):
 
299
        """Write all cached data to disk."""
 
300
        if self._control_files._lock_mode != 'w':
 
301
            raise errors.NotWriteLocked(self)
 
302
        self.current_dirstate().save()
 
303
        self._inventory = None
 
304
        self._dirty = False
 
305
 
 
306
    def _generate_inventory(self):
 
307
        """Create and set self.inventory from the dirstate object.
 
308
        
 
309
        This is relatively expensive: we have to walk the entire dirstate.
 
310
        Ideally we would not, and can deprecate this function.
 
311
        """
 
312
        #: uncomment to trap on inventory requests.
 
313
        # import pdb;pdb.set_trace()
 
314
        state = self.current_dirstate()
 
315
        state._read_dirblocks_if_needed()
 
316
        root_key, current_entry = self._get_entry(path='')
 
317
        current_id = root_key[2]
 
318
        assert current_entry[0][0] == 'd' # directory
 
319
        inv = Inventory(root_id=current_id)
 
320
        # Turn some things into local variables
 
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
322
        factory = entry_factory
 
323
        utf8_decode = cache_utf8._utf8_decode
 
324
        inv_byid = inv._byid
 
325
        # we could do this straight out of the dirstate; it might be fast
 
326
        # and should be profiled - RBC 20070216
 
327
        parent_ies = {'' : inv.root}
 
328
        for block in state._dirblocks[1:]: # skip the root
 
329
            dirname = block[0]
 
330
            try:
 
331
                parent_ie = parent_ies[dirname]
 
332
            except KeyError:
 
333
                # all the paths in this block are not versioned in this tree
 
334
                continue
 
335
            for key, entry in block[1]:
 
336
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
337
                if minikind in ('a', 'r'): # absent, relocated
 
338
                    # a parent tree only entry
 
339
                    continue
 
340
                name = key[1]
 
341
                name_unicode = utf8_decode(name)[0]
 
342
                file_id = key[2]
 
343
                kind = minikind_to_kind[minikind]
 
344
                inv_entry = factory[kind](file_id, name_unicode,
 
345
                                          parent_ie.file_id)
 
346
                if kind == 'file':
 
347
                    # not strictly needed: working tree
 
348
                    #entry.executable = executable
 
349
                    #entry.text_size = size
 
350
                    #entry.text_sha1 = sha1
 
351
                    pass
 
352
                elif kind == 'directory':
 
353
                    # add this entry to the parent map.
 
354
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
355
                elif kind == 'tree-reference':
 
356
                    inv_entry.reference_revision = link_or_sha1
 
357
                else:
 
358
                    assert 'unknown kind'
 
359
                # These checks cost us around 40ms on a 55k entry tree
 
360
                assert file_id not in inv_byid, ('file_id %s already in'
 
361
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
362
                assert name_unicode not in parent_ie.children
 
363
                inv_byid[file_id] = inv_entry
 
364
                parent_ie.children[name_unicode] = inv_entry
 
365
        self._inventory = inv
 
366
 
 
367
    def _get_entry(self, file_id=None, path=None):
 
368
        """Get the dirstate row for file_id or path.
 
369
 
 
370
        If either file_id or path is supplied, it is used as the key to lookup.
 
371
        If both are supplied, the fastest lookup is used, and an error is
 
372
        raised if they do not both point at the same row.
 
373
        
 
374
        :param file_id: An optional unicode file_id to be looked up.
 
375
        :param path: An optional unicode path to be looked up.
 
376
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
377
        """
 
378
        if file_id is None and path is None:
 
379
            raise errors.BzrError('must supply file_id or path')
 
380
        state = self.current_dirstate()
 
381
        if path is not None:
 
382
            path = path.encode('utf8')
 
383
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
384
 
 
385
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
386
        # check file id is valid unconditionally.
 
387
        entry = self._get_entry(file_id=file_id, path=path)
 
388
        assert entry[0] is not None, 'what error should this raise'
 
389
        # TODO:
 
390
        # if row stat is valid, use cached sha1, else, get a new sha1.
 
391
        if path is None:
 
392
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
393
 
 
394
        file_abspath = self.abspath(path)
 
395
        state = self.current_dirstate()
 
396
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
397
                                          stat_value=stat_value)
 
398
        if entry[1][0][0] == 'f':
 
399
            return link_or_sha1
 
400
        return None
 
401
 
 
402
    def _get_inventory(self):
 
403
        """Get the inventory for the tree. This is only valid within a lock."""
 
404
        if self._inventory is not None:
 
405
            return self._inventory
 
406
        self._must_be_locked()
 
407
        self._generate_inventory()
 
408
        return self._inventory
 
409
 
 
410
    inventory = property(_get_inventory,
 
411
                         doc="Inventory of this Tree")
 
412
 
 
413
    @needs_read_lock
 
414
    def get_parent_ids(self):
 
415
        """See Tree.get_parent_ids.
 
416
        
 
417
        This implementation requests the ids list from the dirstate file.
 
418
        """
 
419
        return self.current_dirstate().get_parent_ids()
 
420
 
 
421
    def get_reference_revision(self, entry, path=None):
 
422
        # referenced tree's revision is whatever's currently there
 
423
        return self.get_nested_tree(entry, path).last_revision()
 
424
 
 
425
    def get_nested_tree(self, entry, path=None):
 
426
        if path is None:
 
427
            path = self.id2path(entry.file_id)
 
428
        return WorkingTree.open(self.abspath(path))
 
429
 
 
430
    @needs_read_lock
 
431
    def get_root_id(self):
 
432
        """Return the id of this trees root"""
 
433
        return self._get_entry(path='')[0][2]
 
434
 
 
435
    def has_id(self, file_id):
 
436
        state = self.current_dirstate()
 
437
        file_id = osutils.safe_file_id(file_id)
 
438
        row, parents = self._get_entry(file_id=file_id)
 
439
        if row is None:
 
440
            return False
 
441
        return osutils.lexists(pathjoin(
 
442
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
443
 
 
444
    @needs_read_lock
 
445
    def id2path(self, file_id):
 
446
        file_id = osutils.safe_file_id(file_id)
 
447
        state = self.current_dirstate()
 
448
        entry = self._get_entry(file_id=file_id)
 
449
        if entry == (None, None):
 
450
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
451
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
452
        return path_utf8.decode('utf8')
 
453
 
 
454
    @needs_read_lock
 
455
    def __iter__(self):
 
456
        """Iterate through file_ids for this tree.
 
457
 
 
458
        file_ids are in a WorkingTree if they are in the working inventory
 
459
        and the working file exists.
 
460
        """
 
461
        result = []
 
462
        for key, tree_details in self.current_dirstate()._iter_entries():
 
463
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
464
                # not relevant to the working tree
 
465
                continue
 
466
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
467
            if osutils.lexists(path):
 
468
                result.append(key[2])
 
469
        return iter(result)
 
470
 
 
471
    @needs_read_lock
 
472
    def kind(self, file_id):
 
473
        """Return the kind of a file.
 
474
 
 
475
        This is always the actual kind that's on disk, regardless of what it
 
476
        was added as.
 
477
        """
 
478
        relpath = self.id2path(file_id)
 
479
        assert relpath != None, \
 
480
            "path for id {%s} is None!" % file_id
 
481
        abspath = self.abspath(relpath)
 
482
        kind = file_kind(abspath)
 
483
        if kind == 'directory' and self._directory_is_tree_reference(relpath):
 
484
            kind = 'tree-reference'
 
485
        return kind
 
486
 
 
487
    @needs_read_lock
 
488
    def _last_revision(self):
 
489
        """See Mutable.last_revision."""
 
490
        parent_ids = self.current_dirstate().get_parent_ids()
 
491
        if parent_ids:
 
492
            return parent_ids[0]
 
493
        else:
 
494
            return None
 
495
 
 
496
    def lock_read(self):
 
497
        """See Branch.lock_read, and WorkingTree.unlock."""
 
498
        self.branch.lock_read()
 
499
        try:
 
500
            self._control_files.lock_read()
 
501
            try:
 
502
                state = self.current_dirstate()
 
503
                if not state._lock_token:
 
504
                    state.lock_read()
 
505
            except:
 
506
                self._control_files.unlock()
 
507
                raise
 
508
        except:
 
509
            self.branch.unlock()
 
510
            raise
 
511
 
 
512
    def _lock_self_write(self):
 
513
        """This should be called after the branch is locked."""
 
514
        try:
 
515
            self._control_files.lock_write()
 
516
            try:
 
517
                state = self.current_dirstate()
 
518
                if not state._lock_token:
 
519
                    state.lock_write()
 
520
            except:
 
521
                self._control_files.unlock()
 
522
                raise
 
523
        except:
 
524
            self.branch.unlock()
 
525
            raise
 
526
 
 
527
    def lock_tree_write(self):
 
528
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
529
        self.branch.lock_read()
 
530
        self._lock_self_write()
 
531
 
 
532
    def lock_write(self):
 
533
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
534
        self.branch.lock_write()
 
535
        self._lock_self_write()
 
536
 
 
537
    @needs_tree_write_lock
 
538
    def move(self, from_paths, to_dir, after=False):
 
539
        """See WorkingTree.move()."""
 
540
        result = []
 
541
        if not from_paths:
 
542
            return result
 
543
 
 
544
        state = self.current_dirstate()
 
545
 
 
546
        assert not isinstance(from_paths, basestring)
 
547
        to_dir_utf8 = to_dir.encode('utf8')
 
548
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
549
        id_index = state._get_id_index()
 
550
        # check destination directory
 
551
        # get the details for it
 
552
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
553
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
554
        if not entry_present:
 
555
            raise errors.BzrMoveFailedError('', to_dir,
 
556
                errors.NotVersionedError(to_dir))
 
557
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
558
        # get a handle on the block itself.
 
559
        to_block_index = state._ensure_block(
 
560
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
561
        to_block = state._dirblocks[to_block_index]
 
562
        to_abs = self.abspath(to_dir)
 
563
        if not isdir(to_abs):
 
564
            raise errors.BzrMoveFailedError('',to_dir,
 
565
                errors.NotADirectory(to_abs))
 
566
 
 
567
        if to_entry[1][0][0] != 'd':
 
568
            raise errors.BzrMoveFailedError('',to_dir,
 
569
                errors.NotADirectory(to_abs))
 
570
 
 
571
        if self._inventory is not None:
 
572
            update_inventory = True
 
573
            inv = self.inventory
 
574
            to_dir_ie = inv[to_dir_id]
 
575
            to_dir_id = to_entry[0][2]
 
576
        else:
 
577
            update_inventory = False
 
578
 
 
579
        rollbacks = []
 
580
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
581
                     fingerprint, packed_stat, size,
 
582
                     to_block, to_key, to_path_utf8):
 
583
            state._make_absent(old_entry)
 
584
            from_key = old_entry[0]
 
585
            rollbacks.append(
 
586
                lambda:state.update_minimal(from_key,
 
587
                    minikind,
 
588
                    executable=executable,
 
589
                    fingerprint=fingerprint,
 
590
                    packed_stat=packed_stat,
 
591
                    size=size,
 
592
                    path_utf8=from_path_utf8))
 
593
            state.update_minimal(to_key,
 
594
                    minikind,
 
595
                    executable=executable,
 
596
                    fingerprint=fingerprint,
 
597
                    packed_stat=packed_stat,
 
598
                    size=size,
 
599
                    path_utf8=to_path_utf8)
 
600
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
601
            new_entry = to_block[1][added_entry_index]
 
602
            rollbacks.append(lambda:state._make_absent(new_entry))
 
603
 
 
604
        # create rename entries and tuples
 
605
        for from_rel in from_paths:
 
606
            # from_rel is 'pathinroot/foo/bar'
 
607
            from_rel_utf8 = from_rel.encode('utf8')
 
608
            from_dirname, from_tail = osutils.split(from_rel)
 
609
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
610
            from_entry = self._get_entry(path=from_rel)
 
611
            if from_entry == (None, None):
 
612
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
613
                    errors.NotVersionedError(path=str(from_rel)))
 
614
 
 
615
            from_id = from_entry[0][2]
 
616
            to_rel = pathjoin(to_dir, from_tail)
 
617
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
618
            item_to_entry = self._get_entry(path=to_rel)
 
619
            if item_to_entry != (None, None):
 
620
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
621
                    "Target is already versioned.")
 
622
 
 
623
            if from_rel == to_rel:
 
624
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
625
                    "Source and target are identical.")
 
626
 
 
627
            from_missing = not self.has_filename(from_rel)
 
628
            to_missing = not self.has_filename(to_rel)
 
629
            if after:
 
630
                move_file = False
 
631
            else:
 
632
                move_file = True
 
633
            if to_missing:
 
634
                if not move_file:
 
635
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
636
                        errors.NoSuchFile(path=to_rel,
 
637
                        extra="New file has not been created yet"))
 
638
                elif from_missing:
 
639
                    # neither path exists
 
640
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
641
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
642
            else:
 
643
                if from_missing: # implicitly just update our path mapping
 
644
                    move_file = False
 
645
                elif not after:
 
646
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
647
                        extra="(Use --after to update the Bazaar id)")
 
648
 
 
649
            rollbacks = []
 
650
            def rollback_rename():
 
651
                """A single rename has failed, roll it back."""
 
652
                exc_info = None
 
653
                for rollback in reversed(rollbacks):
 
654
                    try:
 
655
                        rollback()
 
656
                    except Exception, e:
 
657
                        import pdb;pdb.set_trace()
 
658
                        exc_info = sys.exc_info()
 
659
                if exc_info:
 
660
                    raise exc_info[0], exc_info[1], exc_info[2]
 
661
 
 
662
            # perform the disk move first - its the most likely failure point.
 
663
            if move_file:
 
664
                from_rel_abs = self.abspath(from_rel)
 
665
                to_rel_abs = self.abspath(to_rel)
 
666
                try:
 
667
                    osutils.rename(from_rel_abs, to_rel_abs)
 
668
                except OSError, e:
 
669
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
670
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
671
            try:
 
672
                # perform the rename in the inventory next if needed: its easy
 
673
                # to rollback
 
674
                if update_inventory:
 
675
                    # rename the entry
 
676
                    from_entry = inv[from_id]
 
677
                    current_parent = from_entry.parent_id
 
678
                    inv.rename(from_id, to_dir_id, from_tail)
 
679
                    rollbacks.append(
 
680
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
681
                # finally do the rename in the dirstate, which is a little
 
682
                # tricky to rollback, but least likely to need it.
 
683
                old_block_index, old_entry_index, dir_present, file_present = \
 
684
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
685
                old_block = state._dirblocks[old_block_index][1]
 
686
                old_entry = old_block[old_entry_index]
 
687
                from_key, old_entry_details = old_entry
 
688
                cur_details = old_entry_details[0]
 
689
                # remove the old row
 
690
                to_key = ((to_block[0],) + from_key[1:3])
 
691
                minikind = cur_details[0]
 
692
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
693
                         minikind=minikind,
 
694
                         executable=cur_details[3],
 
695
                         fingerprint=cur_details[1],
 
696
                         packed_stat=cur_details[4],
 
697
                         size=cur_details[2],
 
698
                         to_block=to_block,
 
699
                         to_key=to_key,
 
700
                         to_path_utf8=to_rel_utf8)
 
701
 
 
702
                if minikind == 'd':
 
703
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
704
                        """all entries in this block need updating.
 
705
 
 
706
                        TODO: This is pretty ugly, and doesn't support
 
707
                        reverting, but it works.
 
708
                        """
 
709
                        assert from_dir != '', "renaming root not supported"
 
710
                        from_key = (from_dir, '')
 
711
                        from_block_idx, present = \
 
712
                            state._find_block_index_from_key(from_key)
 
713
                        if not present:
 
714
                            # This is the old record, if it isn't present, then
 
715
                            # there is theoretically nothing to update.
 
716
                            # (Unless it isn't present because of lazy loading,
 
717
                            # but we don't do that yet)
 
718
                            return
 
719
                        from_block = state._dirblocks[from_block_idx]
 
720
                        to_block_index, to_entry_index, _, _ = \
 
721
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
722
                        to_block_index = state._ensure_block(
 
723
                            to_block_index, to_entry_index, to_dir_utf8)
 
724
                        to_block = state._dirblocks[to_block_index]
 
725
                        for entry in from_block[1]:
 
726
                            assert entry[0][0] == from_dir
 
727
                            cur_details = entry[1][0]
 
728
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
729
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
730
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
731
                            minikind = cur_details[0]
 
732
                            move_one(entry, from_path_utf8=from_path_utf8,
 
733
                                     minikind=minikind,
 
734
                                     executable=cur_details[3],
 
735
                                     fingerprint=cur_details[1],
 
736
                                     packed_stat=cur_details[4],
 
737
                                     size=cur_details[2],
 
738
                                     to_block=to_block,
 
739
                                     to_key=to_key,
 
740
                                     to_path_utf8=to_rel_utf8)
 
741
                            if minikind == 'd':
 
742
                                # We need to move all the children of this
 
743
                                # entry
 
744
                                update_dirblock(from_path_utf8, to_key,
 
745
                                                to_path_utf8)
 
746
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
747
            except:
 
748
                rollback_rename()
 
749
                raise
 
750
            result.append((from_rel, to_rel))
 
751
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
752
            self._make_dirty(reset_inventory=False)
 
753
 
 
754
        return result
 
755
 
 
756
    def _must_be_locked(self):
 
757
        if not self._control_files._lock_count:
 
758
            raise errors.ObjectNotLocked(self)
 
759
 
 
760
    def _new_tree(self):
 
761
        """Initialize the state in this tree to be a new tree."""
 
762
        self._dirty = True
 
763
 
 
764
    @needs_read_lock
 
765
    def path2id(self, path):
 
766
        """Return the id for path in this tree."""
 
767
        path = path.strip('/')
 
768
        entry = self._get_entry(path=path)
 
769
        if entry == (None, None):
 
770
            return None
 
771
        return entry[0][2]
 
772
 
 
773
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
774
        """See Tree.paths2ids().
 
775
 
 
776
        This specialisation fast-paths the case where all the trees are in the
 
777
        dirstate.
 
778
        """
 
779
        if paths is None:
 
780
            return None
 
781
        parents = self.get_parent_ids()
 
782
        for tree in trees:
 
783
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
784
                parents):
 
785
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
786
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
787
        # -- make all paths utf8 --
 
788
        paths_utf8 = set()
 
789
        for path in paths:
 
790
            paths_utf8.add(path.encode('utf8'))
 
791
        paths = paths_utf8
 
792
        # -- paths is now a utf8 path set --
 
793
        # -- get the state object and prepare it.
 
794
        state = self.current_dirstate()
 
795
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
796
            and '' not in paths):
 
797
            paths2ids = self._paths2ids_using_bisect
 
798
        else:
 
799
            paths2ids = self._paths2ids_in_memory
 
800
        return paths2ids(paths, search_indexes,
 
801
                         require_versioned=require_versioned)
 
802
 
 
803
    def _paths2ids_in_memory(self, paths, search_indexes,
 
804
                             require_versioned=True):
 
805
        state = self.current_dirstate()
 
806
        state._read_dirblocks_if_needed()
 
807
        def _entries_for_path(path):
 
808
            """Return a list with all the entries that match path for all ids.
 
809
            """
 
810
            dirname, basename = os.path.split(path)
 
811
            key = (dirname, basename, '')
 
812
            block_index, present = state._find_block_index_from_key(key)
 
813
            if not present:
 
814
                # the block which should contain path is absent.
 
815
                return []
 
816
            result = []
 
817
            block = state._dirblocks[block_index][1]
 
818
            entry_index, _ = state._find_entry_index(key, block)
 
819
            # we may need to look at multiple entries at this path: walk while the paths match.
 
820
            while (entry_index < len(block) and
 
821
                block[entry_index][0][0:2] == key[0:2]):
 
822
                result.append(block[entry_index])
 
823
                entry_index += 1
 
824
            return result
 
825
        if require_versioned:
 
826
            # -- check all supplied paths are versioned in a search tree. --
 
827
            all_versioned = True
 
828
            for path in paths:
 
829
                path_entries = _entries_for_path(path)
 
830
                if not path_entries:
 
831
                    # this specified path is not present at all: error
 
832
                    all_versioned = False
 
833
                    break
 
834
                found_versioned = False
 
835
                # for each id at this path
 
836
                for entry in path_entries:
 
837
                    # for each tree.
 
838
                    for index in search_indexes:
 
839
                        if entry[1][index][0] != 'a': # absent
 
840
                            found_versioned = True
 
841
                            # all good: found a versioned cell
 
842
                            break
 
843
                if not found_versioned:
 
844
                    # none of the indexes was not 'absent' at all ids for this
 
845
                    # path.
 
846
                    all_versioned = False
 
847
                    break
 
848
            if not all_versioned:
 
849
                raise errors.PathsNotVersionedError(paths)
 
850
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
851
        search_paths = set()
 
852
        for path in paths:
 
853
            other_paths = paths.difference(set([path]))
 
854
            if not osutils.is_inside_any(other_paths, path):
 
855
                # this is a top level path, we must check it.
 
856
                search_paths.add(path)
 
857
        # sketch: 
 
858
        # for all search_indexs in each path at or under each element of
 
859
        # search_paths, if the detail is relocated: add the id, and add the
 
860
        # relocated path as one to search if its not searched already. If the
 
861
        # detail is not relocated, add the id.
 
862
        searched_paths = set()
 
863
        found_ids = set()
 
864
        def _process_entry(entry):
 
865
            """Look at search_indexes within entry.
 
866
 
 
867
            If a specific tree's details are relocated, add the relocation
 
868
            target to search_paths if not searched already. If it is absent, do
 
869
            nothing. Otherwise add the id to found_ids.
 
870
            """
 
871
            for index in search_indexes:
 
872
                if entry[1][index][0] == 'r': # relocated
 
873
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
874
                        search_paths.add(entry[1][index][1])
 
875
                elif entry[1][index][0] != 'a': # absent
 
876
                    found_ids.add(entry[0][2])
 
877
        while search_paths:
 
878
            current_root = search_paths.pop()
 
879
            searched_paths.add(current_root)
 
880
            # process the entries for this containing directory: the rest will be
 
881
            # found by their parents recursively.
 
882
            root_entries = _entries_for_path(current_root)
 
883
            if not root_entries:
 
884
                # this specified path is not present at all, skip it.
 
885
                continue
 
886
            for entry in root_entries:
 
887
                _process_entry(entry)
 
888
            initial_key = (current_root, '', '')
 
889
            block_index, _ = state._find_block_index_from_key(initial_key)
 
890
            while (block_index < len(state._dirblocks) and
 
891
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
892
                for entry in state._dirblocks[block_index][1]:
 
893
                    _process_entry(entry)
 
894
                block_index += 1
 
895
        return found_ids
 
896
 
 
897
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
898
                                require_versioned=True):
 
899
        state = self.current_dirstate()
 
900
        found_ids = set()
 
901
 
 
902
        split_paths = sorted(osutils.split(p) for p in paths)
 
903
        found = state._bisect_recursive(split_paths)
 
904
 
 
905
        if require_versioned:
 
906
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
907
            for dir_name in split_paths:
 
908
                if dir_name not in found_dir_names:
 
909
                    raise errors.PathsNotVersionedError(paths)
 
910
 
 
911
        for dir_name_id, trees_info in found.iteritems():
 
912
            for index in search_indexes:
 
913
                if trees_info[index][0] not in ('r', 'a'):
 
914
                    found_ids.add(dir_name_id[2])
 
915
        return found_ids
 
916
 
 
917
    def read_working_inventory(self):
 
918
        """Read the working inventory.
 
919
        
 
920
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
921
        """
 
922
        return self.inventory
 
923
 
 
924
    @needs_read_lock
 
925
    def revision_tree(self, revision_id):
 
926
        """See Tree.revision_tree.
 
927
 
 
928
        WorkingTree4 supplies revision_trees for any basis tree.
 
929
        """
 
930
        revision_id = osutils.safe_revision_id(revision_id)
 
931
        dirstate = self.current_dirstate()
 
932
        parent_ids = dirstate.get_parent_ids()
 
933
        if revision_id not in parent_ids:
 
934
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
935
        if revision_id in dirstate.get_ghosts():
 
936
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
937
        return DirStateRevisionTree(dirstate, revision_id,
 
938
            self.branch.repository)
 
939
 
 
940
    @needs_tree_write_lock
 
941
    def set_last_revision(self, new_revision):
 
942
        """Change the last revision in the working tree."""
 
943
        new_revision = osutils.safe_revision_id(new_revision)
 
944
        parents = self.get_parent_ids()
 
945
        if new_revision in (NULL_REVISION, None):
 
946
            assert len(parents) < 2, (
 
947
                "setting the last parent to none with a pending merge is "
 
948
                "unsupported.")
 
949
            self.set_parent_ids([])
 
950
        else:
 
951
            self.set_parent_ids([new_revision] + parents[1:],
 
952
                allow_leftmost_as_ghost=True)
 
953
 
 
954
    @needs_tree_write_lock
 
955
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
956
        """Set the parent ids to revision_ids.
 
957
        
 
958
        See also set_parent_trees. This api will try to retrieve the tree data
 
959
        for each element of revision_ids from the trees repository. If you have
 
960
        tree data already available, it is more efficient to use
 
961
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
962
        an easier API to use.
 
963
 
 
964
        :param revision_ids: The revision_ids to set as the parent ids of this
 
965
            working tree. Any of these may be ghosts.
 
966
        """
 
967
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
968
        trees = []
 
969
        for revision_id in revision_ids:
 
970
            try:
 
971
                revtree = self.branch.repository.revision_tree(revision_id)
 
972
                # TODO: jam 20070213 KnitVersionedFile raises
 
973
                #       RevisionNotPresent rather than NoSuchRevision if a
 
974
                #       given revision_id is not present. Should Repository be
 
975
                #       catching it and re-raising NoSuchRevision?
 
976
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
977
                revtree = None
 
978
            trees.append((revision_id, revtree))
 
979
        self.current_dirstate()._validate()
 
980
        self.set_parent_trees(trees,
 
981
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
982
        self.current_dirstate()._validate()
 
983
 
 
984
    @needs_tree_write_lock
 
985
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
986
        """Set the parents of the working tree.
 
987
 
 
988
        :param parents_list: A list of (revision_id, tree) tuples.
 
989
            If tree is None, then that element is treated as an unreachable
 
990
            parent tree - i.e. a ghost.
 
991
        """
 
992
        dirstate = self.current_dirstate()
 
993
        dirstate._validate()
 
994
        if len(parents_list) > 0:
 
995
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
996
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
997
        real_trees = []
 
998
        ghosts = []
 
999
        # convert absent trees to the null tree, which we convert back to
 
1000
        # missing on access.
 
1001
        for rev_id, tree in parents_list:
 
1002
            rev_id = osutils.safe_revision_id(rev_id)
 
1003
            if tree is not None:
 
1004
                real_trees.append((rev_id, tree))
 
1005
            else:
 
1006
                real_trees.append((rev_id,
 
1007
                    self.branch.repository.revision_tree(None)))
 
1008
                ghosts.append(rev_id)
 
1009
        dirstate._validate()
 
1010
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1011
        dirstate._validate()
 
1012
        self._make_dirty(reset_inventory=False)
 
1013
        dirstate._validate()
 
1014
 
 
1015
    def _set_root_id(self, file_id):
 
1016
        """See WorkingTree.set_root_id."""
 
1017
        state = self.current_dirstate()
 
1018
        state.set_path_id('', file_id)
 
1019
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1020
            self._make_dirty(reset_inventory=True)
 
1021
 
 
1022
    def unlock(self):
 
1023
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1024
        if self._control_files._lock_count == 1:
 
1025
            # eventually we should do signature checking during read locks for
 
1026
            # dirstate updates.
 
1027
            if self._control_files._lock_mode == 'w':
 
1028
                if self._dirty:
 
1029
                    self.flush()
 
1030
            if self._dirstate is not None:
 
1031
                # This is a no-op if there are no modifications.
 
1032
                self._dirstate.save()
 
1033
                self._dirstate.unlock()
 
1034
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1035
            #       point. Instead, it could check if the header has been
 
1036
            #       modified when it is locked, and if not, it can hang on to
 
1037
            #       the data it has in memory.
 
1038
            self._dirstate = None
 
1039
            self._inventory = None
 
1040
        # reverse order of locking.
 
1041
        try:
 
1042
            return self._control_files.unlock()
 
1043
        finally:
 
1044
            self.branch.unlock()
 
1045
 
 
1046
    @needs_tree_write_lock
 
1047
    def unversion(self, file_ids):
 
1048
        """Remove the file ids in file_ids from the current versioned set.
 
1049
 
 
1050
        When a file_id is unversioned, all of its children are automatically
 
1051
        unversioned.
 
1052
 
 
1053
        :param file_ids: The file ids to stop versioning.
 
1054
        :raises: NoSuchId if any fileid is not currently versioned.
 
1055
        """
 
1056
        if not file_ids:
 
1057
            return
 
1058
        state = self.current_dirstate()
 
1059
        state._read_dirblocks_if_needed()
 
1060
        ids_to_unversion = set()
 
1061
        for file_id in file_ids:
 
1062
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1063
        paths_to_unversion = set()
 
1064
        # sketch:
 
1065
        # check if the root is to be unversioned, if so, assert for now.
 
1066
        # walk the state marking unversioned things as absent.
 
1067
        # if there are any un-unversioned ids at the end, raise
 
1068
        for key, details in state._dirblocks[0][1]:
 
1069
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1070
                key[2] in ids_to_unversion):
 
1071
                # I haven't written the code to unversion / yet - it should be
 
1072
                # supported.
 
1073
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1074
        block_index = 0
 
1075
        while block_index < len(state._dirblocks):
 
1076
            # process one directory at a time.
 
1077
            block = state._dirblocks[block_index]
 
1078
            # first check: is the path one to remove - it or its children
 
1079
            delete_block = False
 
1080
            for path in paths_to_unversion:
 
1081
                if (block[0].startswith(path) and
 
1082
                    (len(block[0]) == len(path) or
 
1083
                     block[0][len(path)] == '/')):
 
1084
                    # this entire block should be deleted - its the block for a
 
1085
                    # path to unversion; or the child of one
 
1086
                    delete_block = True
 
1087
                    break
 
1088
            # TODO: trim paths_to_unversion as we pass by paths
 
1089
            if delete_block:
 
1090
                # this block is to be deleted: process it.
 
1091
                # TODO: we can special case the no-parents case and
 
1092
                # just forget the whole block.
 
1093
                entry_index = 0
 
1094
                while entry_index < len(block[1]):
 
1095
                    # Mark this file id as having been removed
 
1096
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1097
                    if not state._make_absent(block[1][entry_index]):
 
1098
                        entry_index += 1
 
1099
                # go to the next block. (At the moment we dont delete empty
 
1100
                # dirblocks)
 
1101
                block_index += 1
 
1102
                continue
 
1103
            entry_index = 0
 
1104
            while entry_index < len(block[1]):
 
1105
                entry = block[1][entry_index]
 
1106
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1107
                    # ^ some parent row.
 
1108
                    entry[0][2] not in ids_to_unversion):
 
1109
                    # ^ not an id to unversion
 
1110
                    entry_index += 1
 
1111
                    continue
 
1112
                if entry[1][0][0] == 'd':
 
1113
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1114
                if not state._make_absent(entry):
 
1115
                    entry_index += 1
 
1116
                # we have unversioned this id
 
1117
                ids_to_unversion.remove(entry[0][2])
 
1118
            block_index += 1
 
1119
        if ids_to_unversion:
 
1120
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1121
        self._make_dirty(reset_inventory=False)
 
1122
        # have to change the legacy inventory too.
 
1123
        if self._inventory is not None:
 
1124
            for file_id in file_ids:
 
1125
                self._inventory.remove_recursive_id(file_id)
 
1126
 
 
1127
    @needs_tree_write_lock
 
1128
    def _write_inventory(self, inv):
 
1129
        """Write inventory as the current inventory."""
 
1130
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1131
        self.current_dirstate().set_state_from_inventory(inv)
 
1132
        self._make_dirty(reset_inventory=False)
 
1133
        if self._inventory is not None:
 
1134
            self._inventory = inv
 
1135
        self.flush()
 
1136
 
 
1137
 
 
1138
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1139
    """The first consolidated dirstate working tree format.
 
1140
 
 
1141
    This format:
 
1142
        - exists within a metadir controlling .bzr
 
1143
        - includes an explicit version marker for the workingtree control
 
1144
          files, separate from the BzrDir format
 
1145
        - modifies the hash cache format
 
1146
        - is new in bzr TODO FIXME SETBEFOREMERGE
 
1147
        - uses a LockDir to guard access to it.
 
1148
    """
 
1149
 
 
1150
    supports_tree_reference = True
 
1151
 
 
1152
    def get_format_string(self):
 
1153
        """See WorkingTreeFormat.get_format_string()."""
 
1154
        return "Bazaar Working Tree format 4\n"
 
1155
 
 
1156
    def get_format_description(self):
 
1157
        """See WorkingTreeFormat.get_format_description()."""
 
1158
        return "Working tree format 4"
 
1159
 
 
1160
    def initialize(self, a_bzrdir, revision_id=None):
 
1161
        """See WorkingTreeFormat.initialize().
 
1162
 
 
1163
        :param revision_id: allows creating a working tree at a different
 
1164
        revision than the branch is at.
 
1165
 
 
1166
        These trees get an initial random root id.
 
1167
        """
 
1168
        revision_id = osutils.safe_revision_id(revision_id)
 
1169
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1170
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1171
        transport = a_bzrdir.get_workingtree_transport(self)
 
1172
        control_files = self._open_control_files(a_bzrdir)
 
1173
        control_files.create_lock()
 
1174
        control_files.lock_write()
 
1175
        control_files.put_utf8('format', self.get_format_string())
 
1176
        branch = a_bzrdir.open_branch()
 
1177
        if revision_id is None:
 
1178
            revision_id = branch.last_revision()
 
1179
        local_path = transport.local_abspath('dirstate')
 
1180
        # write out new dirstate (must exist when we create the tree)
 
1181
        state = dirstate.DirState.initialize(local_path)
 
1182
        state.unlock()
 
1183
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1184
                         branch,
 
1185
                         _format=self,
 
1186
                         _bzrdir=a_bzrdir,
 
1187
                         _control_files=control_files)
 
1188
        wt._new_tree()
 
1189
        wt.lock_tree_write()
 
1190
        state._validate()
 
1191
        try:
 
1192
            if revision_id in (None, NULL_REVISION):
 
1193
                wt._set_root_id(generate_ids.gen_root_id())
 
1194
                wt.flush()
 
1195
                wt.current_dirstate()._validate()
 
1196
            wt.set_last_revision(revision_id)
 
1197
            wt.flush()
 
1198
            basis = wt.basis_tree()
 
1199
            basis.lock_read()
 
1200
            # if the basis has a root id we have to use that; otherwise we use
 
1201
            # a new random one
 
1202
            basis_root_id = basis.get_root_id()
 
1203
            if basis_root_id is not None:
 
1204
                wt._set_root_id(basis_root_id)
 
1205
                wt.flush()
 
1206
            transform.build_tree(basis, wt)
 
1207
            basis.unlock()
 
1208
        finally:
 
1209
            control_files.unlock()
 
1210
            wt.unlock()
 
1211
        return wt
 
1212
 
 
1213
    def _open(self, a_bzrdir, control_files):
 
1214
        """Open the tree itself.
 
1215
 
 
1216
        :param a_bzrdir: the dir for the tree.
 
1217
        :param control_files: the control files for the tree.
 
1218
        """
 
1219
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1220
                           branch=a_bzrdir.open_branch(),
 
1221
                           _format=self,
 
1222
                           _bzrdir=a_bzrdir,
 
1223
                           _control_files=control_files)
 
1224
 
 
1225
    def __get_matchingbzrdir(self):
 
1226
        # please test against something that will let us do tree references
 
1227
        return bzrdir.format_registry.make_bzrdir(
 
1228
            'dirstate-with-subtree')
 
1229
 
 
1230
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1231
 
 
1232
 
 
1233
class DirStateRevisionTree(Tree):
 
1234
    """A revision tree pulling the inventory from a dirstate."""
 
1235
 
 
1236
    def __init__(self, dirstate, revision_id, repository):
 
1237
        self._dirstate = dirstate
 
1238
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1239
        self._repository = repository
 
1240
        self._inventory = None
 
1241
        self._locked = 0
 
1242
        self._dirstate_locked = False
 
1243
 
 
1244
    def __repr__(self):
 
1245
        return "<%s of %s in %s>" % \
 
1246
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1247
 
 
1248
    def annotate_iter(self, file_id):
 
1249
        """See Tree.annotate_iter"""
 
1250
        w = self._repository.weave_store.get_weave(file_id,
 
1251
                           self._repository.get_transaction())
 
1252
        return w.annotate_iter(self.inventory[file_id].revision)
 
1253
 
 
1254
    def _comparison_data(self, entry, path):
 
1255
        """See Tree._comparison_data."""
 
1256
        if entry is None:
 
1257
            return None, False, None
 
1258
        # trust the entry as RevisionTree does, but this may not be
 
1259
        # sensible: the entry might not have come from us?
 
1260
        return entry.kind, entry.executable, None
 
1261
 
 
1262
    def _file_size(self, entry, stat_value):
 
1263
        return entry.text_size
 
1264
 
 
1265
    def filter_unversioned_files(self, paths):
 
1266
        """Filter out paths that are not versioned.
 
1267
 
 
1268
        :return: set of paths.
 
1269
        """
 
1270
        pred = self.has_filename
 
1271
        return set((p for p in paths if not pred(p)))
 
1272
 
 
1273
    def get_root_id(self):
 
1274
        return self.path2id('')
 
1275
 
 
1276
    def _get_parent_index(self):
 
1277
        """Return the index in the dirstate referenced by this tree."""
 
1278
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1279
 
 
1280
    def _get_entry(self, file_id=None, path=None):
 
1281
        """Get the dirstate row for file_id or path.
 
1282
 
 
1283
        If either file_id or path is supplied, it is used as the key to lookup.
 
1284
        If both are supplied, the fastest lookup is used, and an error is
 
1285
        raised if they do not both point at the same row.
 
1286
        
 
1287
        :param file_id: An optional unicode file_id to be looked up.
 
1288
        :param path: An optional unicode path to be looked up.
 
1289
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1290
        """
 
1291
        if file_id is None and path is None:
 
1292
            raise errors.BzrError('must supply file_id or path')
 
1293
        file_id = osutils.safe_file_id(file_id)
 
1294
        if path is not None:
 
1295
            path = path.encode('utf8')
 
1296
        parent_index = self._get_parent_index()
 
1297
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1298
 
 
1299
    def _generate_inventory(self):
 
1300
        """Create and set self.inventory from the dirstate object.
 
1301
 
 
1302
        (So this is only called the first time the inventory is requested for
 
1303
        this tree; it then remains in memory until it's out of date.)
 
1304
 
 
1305
        This is relatively expensive: we have to walk the entire dirstate.
 
1306
        """
 
1307
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1308
            'dirstate revision tree'
 
1309
        # separate call for profiling - makes it clear where the costs are.
 
1310
        self._dirstate._read_dirblocks_if_needed()
 
1311
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1312
            'parent %s has disappeared from %s' % (
 
1313
            self._revision_id, self._dirstate.get_parent_ids())
 
1314
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1315
        # This is identical now to the WorkingTree _generate_inventory except
 
1316
        # for the tree index use.
 
1317
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1318
        current_id = root_key[2]
 
1319
        assert current_entry[parent_index][0] == 'd'
 
1320
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1321
        inv.root.revision = current_entry[parent_index][4]
 
1322
        # Turn some things into local variables
 
1323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1324
        factory = entry_factory
 
1325
        utf8_decode = cache_utf8._utf8_decode
 
1326
        inv_byid = inv._byid
 
1327
        # we could do this straight out of the dirstate; it might be fast
 
1328
        # and should be profiled - RBC 20070216
 
1329
        parent_ies = {'' : inv.root}
 
1330
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1331
            dirname = block[0]
 
1332
            try:
 
1333
                parent_ie = parent_ies[dirname]
 
1334
            except KeyError:
 
1335
                # all the paths in this block are not versioned in this tree
 
1336
                continue
 
1337
            for key, entry in block[1]:
 
1338
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1339
                if minikind in ('a', 'r'): # absent, relocated
 
1340
                    # not this tree
 
1341
                    continue
 
1342
                name = key[1]
 
1343
                name_unicode = utf8_decode(name)[0]
 
1344
                file_id = key[2]
 
1345
                kind = minikind_to_kind[minikind]
 
1346
                inv_entry = factory[kind](file_id, name_unicode,
 
1347
                                          parent_ie.file_id)
 
1348
                inv_entry.revision = revid
 
1349
                if kind == 'file':
 
1350
                    inv_entry.executable = executable
 
1351
                    inv_entry.text_size = size
 
1352
                    inv_entry.text_sha1 = fingerprint
 
1353
                elif kind == 'directory':
 
1354
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1355
                elif kind == 'symlink':
 
1356
                    inv_entry.executable = False
 
1357
                    inv_entry.text_size = size
 
1358
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1359
                elif kind == 'tree-reference':
 
1360
                    inv_entry.reference_revision = fingerprint
 
1361
                else:
 
1362
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1363
                            % entry)
 
1364
                # These checks cost us around 40ms on a 55k entry tree
 
1365
                assert file_id not in inv_byid
 
1366
                assert name_unicode not in parent_ie.children
 
1367
                inv_byid[file_id] = inv_entry
 
1368
                parent_ie.children[name_unicode] = inv_entry
 
1369
        self._inventory = inv
 
1370
 
 
1371
    def get_file_mtime(self, file_id, path=None):
 
1372
        """Return the modification time for this record.
 
1373
 
 
1374
        We return the timestamp of the last-changed revision.
 
1375
        """
 
1376
        # Make sure the file exists
 
1377
        entry = self._get_entry(file_id, path=path)
 
1378
        if entry == (None, None): # do we raise?
 
1379
            return None
 
1380
        parent_index = self._get_parent_index()
 
1381
        last_changed_revision = entry[1][parent_index][4]
 
1382
        return self._repository.get_revision(last_changed_revision).timestamp
 
1383
 
 
1384
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1385
        entry = self._get_entry(file_id=file_id, path=path)
 
1386
        parent_index = self._get_parent_index()
 
1387
        parent_details = entry[1][parent_index]
 
1388
        if parent_details[0] == 'f':
 
1389
            return parent_details[1]
 
1390
        return None
 
1391
 
 
1392
    def get_file(self, file_id):
 
1393
        return StringIO(self.get_file_text(file_id))
 
1394
 
 
1395
    def get_file_lines(self, file_id):
 
1396
        ie = self.inventory[file_id]
 
1397
        return self._repository.weave_store.get_weave(file_id,
 
1398
                self._repository.get_transaction()).get_lines(ie.revision)
 
1399
 
 
1400
    def get_file_size(self, file_id):
 
1401
        return self.inventory[file_id].text_size
 
1402
 
 
1403
    def get_file_text(self, file_id):
 
1404
        return ''.join(self.get_file_lines(file_id))
 
1405
 
 
1406
    def get_symlink_target(self, file_id):
 
1407
        entry = self._get_entry(file_id=file_id)
 
1408
        parent_index = self._get_parent_index()
 
1409
        if entry[1][parent_index][0] != 'l':
 
1410
            return None
 
1411
        else:
 
1412
            # At present, none of the tree implementations supports non-ascii
 
1413
            # symlink targets. So we will just assume that the dirstate path is
 
1414
            # correct.
 
1415
            return entry[1][parent_index][1]
 
1416
 
 
1417
    def get_revision_id(self):
 
1418
        """Return the revision id for this tree."""
 
1419
        return self._revision_id
 
1420
 
 
1421
    def _get_inventory(self):
 
1422
        if self._inventory is not None:
 
1423
            return self._inventory
 
1424
        self._must_be_locked()
 
1425
        self._generate_inventory()
 
1426
        return self._inventory
 
1427
 
 
1428
    inventory = property(_get_inventory,
 
1429
                         doc="Inventory of this Tree")
 
1430
 
 
1431
    def get_parent_ids(self):
 
1432
        """The parents of a tree in the dirstate are not cached."""
 
1433
        return self._repository.get_revision(self._revision_id).parent_ids
 
1434
 
 
1435
    def has_filename(self, filename):
 
1436
        return bool(self.path2id(filename))
 
1437
 
 
1438
    def kind(self, file_id):
 
1439
        return self.inventory[file_id].kind
 
1440
 
 
1441
    def is_executable(self, file_id, path=None):
 
1442
        ie = self.inventory[file_id]
 
1443
        if ie.kind != "file":
 
1444
            return None
 
1445
        return ie.executable
 
1446
 
 
1447
    def list_files(self, include_root=False):
 
1448
        # We use a standard implementation, because DirStateRevisionTree is
 
1449
        # dealing with one of the parents of the current state
 
1450
        inv = self._get_inventory()
 
1451
        entries = inv.iter_entries()
 
1452
        if self.inventory.root is not None and not include_root:
 
1453
            entries.next()
 
1454
        for path, entry in entries:
 
1455
            yield path, 'V', entry.kind, entry.file_id, entry
 
1456
 
 
1457
    def lock_read(self):
 
1458
        """Lock the tree for a set of operations."""
 
1459
        if not self._locked:
 
1460
            self._repository.lock_read()
 
1461
            if self._dirstate._lock_token is None:
 
1462
                self._dirstate.lock_read()
 
1463
                self._dirstate_locked = True
 
1464
        self._locked += 1
 
1465
 
 
1466
    def _must_be_locked(self):
 
1467
        if not self._locked:
 
1468
            raise errors.ObjectNotLocked(self)
 
1469
 
 
1470
    @needs_read_lock
 
1471
    def path2id(self, path):
 
1472
        """Return the id for path in this tree."""
 
1473
        # lookup by path: faster than splitting and walking the ivnentory.
 
1474
        entry = self._get_entry(path=path)
 
1475
        if entry == (None, None):
 
1476
            return None
 
1477
        return entry[0][2]
 
1478
 
 
1479
    def unlock(self):
 
1480
        """Unlock, freeing any cache memory used during the lock."""
 
1481
        # outside of a lock, the inventory is suspect: release it.
 
1482
        self._locked -=1
 
1483
        if not self._locked:
 
1484
            self._inventory = None
 
1485
            self._locked = 0
 
1486
            if self._dirstate_locked:
 
1487
                self._dirstate.unlock()
 
1488
                self._dirstate_locked = False
 
1489
            self._repository.unlock()
 
1490
 
 
1491
    def walkdirs(self, prefix=""):
 
1492
        # TODO: jam 20070215 This is the cheap way by cheating and using the
 
1493
        #       RevisionTree implementation.
 
1494
        #       This should be cleaned up to use the much faster Dirstate code
 
1495
        #       This is a little tricky, though, because the dirstate is
 
1496
        #       indexed by current path, not by parent path.
 
1497
        #       So for now, we just build up the parent inventory, and extract
 
1498
        #       it the same way RevisionTree does.
 
1499
        _directory = 'directory'
 
1500
        inv = self._get_inventory()
 
1501
        top_id = inv.path2id(prefix)
 
1502
        if top_id is None:
 
1503
            pending = []
 
1504
        else:
 
1505
            pending = [(prefix, top_id)]
 
1506
        while pending:
 
1507
            dirblock = []
 
1508
            relpath, file_id = pending.pop()
 
1509
            # 0 - relpath, 1- file-id
 
1510
            if relpath:
 
1511
                relroot = relpath + '/'
 
1512
            else:
 
1513
                relroot = ""
 
1514
            # FIXME: stash the node in pending
 
1515
            entry = inv[file_id]
 
1516
            for name, child in entry.sorted_children():
 
1517
                toppath = relroot + name
 
1518
                dirblock.append((toppath, name, child.kind, None,
 
1519
                    child.file_id, child.kind
 
1520
                    ))
 
1521
            yield (relpath, entry.file_id), dirblock
 
1522
            # push the user specified dirs from dirblock
 
1523
            for dir in reversed(dirblock):
 
1524
                if dir[2] == _directory:
 
1525
                    pending.append((dir[0], dir[4]))
 
1526
 
 
1527
 
 
1528
class InterDirStateTree(InterTree):
 
1529
    """Fast path optimiser for changes_from with dirstate trees.
 
1530
    
 
1531
    This is used only when both trees are in the dirstate working file, and 
 
1532
    the source is any parent within the dirstate, and the destination is 
 
1533
    the current working tree of the same dirstate.
 
1534
    """
 
1535
    # this could be generalized to allow comparisons between any trees in the
 
1536
    # dirstate, and possibly between trees stored in different dirstates.
 
1537
 
 
1538
    def __init__(self, source, target):
 
1539
        super(InterDirStateTree, self).__init__(source, target)
 
1540
        if not InterDirStateTree.is_compatible(source, target):
 
1541
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1542
 
 
1543
    @staticmethod
 
1544
    def make_source_parent_tree(source, target):
 
1545
        """Change the source tree into a parent of the target."""
 
1546
        revid = source.commit('record tree')
 
1547
        target.branch.repository.fetch(source.branch.repository, revid)
 
1548
        target.set_parent_ids([revid])
 
1549
        return target.basis_tree(), target
 
1550
 
 
1551
    _matching_from_tree_format = WorkingTreeFormat4()
 
1552
    _matching_to_tree_format = WorkingTreeFormat4()
 
1553
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1554
 
 
1555
    def _iter_changes(self, include_unchanged=False,
 
1556
                      specific_files=None, pb=None, extra_trees=[],
 
1557
                      require_versioned=True, want_unversioned=False):
 
1558
        """Return the changes from source to target.
 
1559
 
 
1560
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1561
            for details.
 
1562
        :param specific_files: An optional list of file paths to restrict the
 
1563
            comparison to. When mapping filenames to ids, all matches in all
 
1564
            trees (including optional extra_trees) are used, and all children of
 
1565
            matched directories are included.
 
1566
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1567
            unchanged entries in the result.
 
1568
        :param extra_trees: An optional list of additional trees to use when
 
1569
            mapping the contents of specific_files (paths) to file_ids.
 
1570
        :param require_versioned: If True, all files in specific_files must be
 
1571
            versioned in one of source, target, extra_trees or
 
1572
            PathsNotVersionedError is raised.
 
1573
        :param want_unversioned: Should unversioned files be returned in the
 
1574
            output. An unversioned file is defined as one with (False, False)
 
1575
            for the versioned pair.
 
1576
        """
 
1577
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1578
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1579
        # NB: show_status depends on being able to pass in non-versioned files
 
1580
        # and report them as unknown
 
1581
        # TODO: handle extra trees in the dirstate.
 
1582
        # TODO: handle comparisons as an empty tree as a different special
 
1583
        # case? mbp 20070226
 
1584
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1585
            # we can't fast-path these cases (yet)
 
1586
            for f in super(InterDirStateTree, self)._iter_changes(
 
1587
                include_unchanged, specific_files, pb, extra_trees,
 
1588
                require_versioned, want_unversioned=want_unversioned):
 
1589
                yield f
 
1590
            return
 
1591
        parent_ids = self.target.get_parent_ids()
 
1592
        assert (self.source._revision_id in parent_ids), \
 
1593
                "revision {%s} is not stored in {%s}, but %s " \
 
1594
                "can only be used for trees stored in the dirstate" \
 
1595
                % (self.source._revision_id, self.target, self._iter_changes)
 
1596
        target_index = 0
 
1597
        if self.source._revision_id == NULL_REVISION:
 
1598
            source_index = None
 
1599
            indices = (target_index,)
 
1600
        else:
 
1601
            assert (self.source._revision_id in parent_ids), \
 
1602
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1603
                self.source._revision_id, parent_ids)
 
1604
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1605
            indices = (source_index,target_index)
 
1606
        # -- make all specific_files utf8 --
 
1607
        if specific_files:
 
1608
            specific_files_utf8 = set()
 
1609
            for path in specific_files:
 
1610
                specific_files_utf8.add(path.encode('utf8'))
 
1611
            specific_files = specific_files_utf8
 
1612
        else:
 
1613
            specific_files = set([''])
 
1614
        # -- specific_files is now a utf8 path set --
 
1615
        # -- get the state object and prepare it.
 
1616
        state = self.target.current_dirstate()
 
1617
        state._read_dirblocks_if_needed()
 
1618
        def _entries_for_path(path):
 
1619
            """Return a list with all the entries that match path for all ids.
 
1620
            """
 
1621
            dirname, basename = os.path.split(path)
 
1622
            key = (dirname, basename, '')
 
1623
            block_index, present = state._find_block_index_from_key(key)
 
1624
            if not present:
 
1625
                # the block which should contain path is absent.
 
1626
                return []
 
1627
            result = []
 
1628
            block = state._dirblocks[block_index][1]
 
1629
            entry_index, _ = state._find_entry_index(key, block)
 
1630
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1631
            while (entry_index < len(block) and
 
1632
                block[entry_index][0][0:2] == key[0:2]):
 
1633
                result.append(block[entry_index])
 
1634
                entry_index += 1
 
1635
            return result
 
1636
        if require_versioned:
 
1637
            # -- check all supplied paths are versioned in a search tree. --
 
1638
            all_versioned = True
 
1639
            for path in specific_files:
 
1640
                path_entries = _entries_for_path(path)
 
1641
                if not path_entries:
 
1642
                    # this specified path is not present at all: error
 
1643
                    all_versioned = False
 
1644
                    break
 
1645
                found_versioned = False
 
1646
                # for each id at this path
 
1647
                for entry in path_entries:
 
1648
                    # for each tree.
 
1649
                    for index in indices:
 
1650
                        if entry[1][index][0] != 'a': # absent
 
1651
                            found_versioned = True
 
1652
                            # all good: found a versioned cell
 
1653
                            break
 
1654
                if not found_versioned:
 
1655
                    # none of the indexes was not 'absent' at all ids for this
 
1656
                    # path.
 
1657
                    all_versioned = False
 
1658
                    break
 
1659
            if not all_versioned:
 
1660
                raise errors.PathsNotVersionedError(specific_files)
 
1661
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1662
        search_specific_files = set()
 
1663
        for path in specific_files:
 
1664
            other_specific_files = specific_files.difference(set([path]))
 
1665
            if not osutils.is_inside_any(other_specific_files, path):
 
1666
                # this is a top level path, we must check it.
 
1667
                search_specific_files.add(path)
 
1668
        # sketch: 
 
1669
        # compare source_index and target_index at or under each element of search_specific_files.
 
1670
        # follow the following comparison table. Note that we only want to do diff operations when
 
1671
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1672
        # for the target.
 
1673
        # cases:
 
1674
        # 
 
1675
        # Source | Target | disk | action
 
1676
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1677
        #        |        |      | diff check on source-target
 
1678
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1679
        #        |        |      | ???
 
1680
        #   r    |  a     |      | add source to search
 
1681
        #   r    |  a     |  a   | 
 
1682
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1683
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1684
        #   a    | fdlt   |      | add new id
 
1685
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1686
        #   a    |  a     |      | not present in either tree, skip
 
1687
        #   a    |  a     |  a   | not present in any tree, skip
 
1688
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1689
        #        |        |      | may not be selected by the users list of paths.
 
1690
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1691
        #        |        |      | may not be selected by the users list of paths.
 
1692
        #  fdlt  | fdlt   |      | content in both: diff them
 
1693
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1694
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1695
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1696
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1697
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1698
        #        |        |      | this id at the other path.
 
1699
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1700
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1701
        #        |        |      | this id at the other path.
 
1702
 
 
1703
        # for all search_indexs in each path at or under each element of
 
1704
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1705
        # relocated path as one to search if its not searched already. If the
 
1706
        # detail is not relocated, add the id.
 
1707
        searched_specific_files = set()
 
1708
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1709
        # Using a list so that we can access the values and change them in
 
1710
        # nested scope. Each one is [path, file_id, entry]
 
1711
        last_source_parent = [None, None, None]
 
1712
        last_target_parent = [None, None, None]
 
1713
 
 
1714
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1715
 
 
1716
        def _process_entry(entry, path_info):
 
1717
            """Compare an entry and real disk to generate delta information.
 
1718
 
 
1719
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1720
                the path of entry. If None, then the path is considered absent.
 
1721
                (Perhaps we should pass in a concrete entry for this ?)
 
1722
                Basename is returned as a utf8 string because we expect this
 
1723
                tuple will be ignored, and don't want to take the time to
 
1724
                decode.
 
1725
            """
 
1726
            # TODO: when a parent has been renamed, dont emit path renames for children,
 
1727
            ## if path_info[1] == 'sub':
 
1728
            ##     import pdb;pdb.set_trace()
 
1729
            if source_index is None:
 
1730
                source_details = NULL_PARENT_DETAILS
 
1731
            else:
 
1732
                source_details = entry[1][source_index]
 
1733
            target_details = entry[1][target_index]
 
1734
            target_minikind = target_details[0]
 
1735
            if path_info is not None and target_minikind in 'fdlt':
 
1736
                assert target_index == 0
 
1737
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1738
                                                  stat_value=path_info[3])
 
1739
                # The entry may have been modified by update_entry
 
1740
                target_details = entry[1][target_index]
 
1741
                target_minikind = target_details[0]
 
1742
            else:
 
1743
                link_or_sha1 = None
 
1744
            source_minikind = source_details[0]
 
1745
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1746
                # claimed content in both: diff
 
1747
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1748
                #        |        |      | diff check on source-target
 
1749
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1750
                #        |        |      | ???
 
1751
                if source_minikind in 'r':
 
1752
                    # add the source to the search path to find any children it
 
1753
                    # has.  TODO ? : only add if it is a container ?
 
1754
                    if not osutils.is_inside_any(searched_specific_files,
 
1755
                                                 source_details[1]):
 
1756
                        search_specific_files.add(source_details[1])
 
1757
                    # generate the old path; this is needed for stating later
 
1758
                    # as well.
 
1759
                    old_path = source_details[1]
 
1760
                    old_dirname, old_basename = os.path.split(old_path)
 
1761
                    path = pathjoin(entry[0][0], entry[0][1])
 
1762
                    old_entry = state._get_entry(source_index,
 
1763
                                                 path_utf8=old_path)
 
1764
                    # update the source details variable to be the real
 
1765
                    # location.
 
1766
                    source_details = old_entry[1][source_index]
 
1767
                    source_minikind = source_details[0]
 
1768
                else:
 
1769
                    old_dirname = entry[0][0]
 
1770
                    old_basename = entry[0][1]
 
1771
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1772
                if path_info is None:
 
1773
                    # the file is missing on disk, show as removed.
 
1774
                    content_change = True
 
1775
                    target_kind = None
 
1776
                    target_exec = False
 
1777
                else:
 
1778
                    # source and target are both versioned and disk file is present.
 
1779
                    target_kind = path_info[2]
 
1780
                    if target_kind == 'directory':
 
1781
                        if source_minikind != 'd':
 
1782
                            content_change = True
 
1783
                        else:
 
1784
                            # directories have no fingerprint
 
1785
                            content_change = False
 
1786
                        target_exec = False
 
1787
                    elif target_kind == 'file':
 
1788
                        if source_minikind != 'f':
 
1789
                            content_change = True
 
1790
                        else:
 
1791
                            # We could check the size, but we already have the
 
1792
                            # sha1 hash.
 
1793
                            content_change = (link_or_sha1 != source_details[1])
 
1794
                        # Target details is updated at update_entry time
 
1795
                        if use_filesystem_for_exec:
 
1796
                            # We don't need S_ISREG here, because we are sure
 
1797
                            # we are dealing with a file.
 
1798
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1799
                        else:
 
1800
                            target_exec = target_details[3]
 
1801
                    elif target_kind == 'symlink':
 
1802
                        if source_minikind != 'l':
 
1803
                            content_change = True
 
1804
                        else:
 
1805
                            content_change = (link_or_sha1 != source_details[1])
 
1806
                        target_exec = False
 
1807
                    elif target_kind == 'tree-reference':
 
1808
                        if source_minikind != 't':
 
1809
                            content_change = True
 
1810
                        else:
 
1811
                            content_change = False
 
1812
                    else:
 
1813
                        raise Exception, "unknown kind %s" % path_info[2]
 
1814
                # parent id is the entry for the path in the target tree
 
1815
                if old_dirname == last_source_parent[0]:
 
1816
                    source_parent_id = last_source_parent[1]
 
1817
                else:
 
1818
                    source_parent_entry = state._get_entry(source_index,
 
1819
                                                           path_utf8=old_dirname)
 
1820
                    source_parent_id = source_parent_entry[0][2]
 
1821
                    if source_parent_id == entry[0][2]:
 
1822
                        # This is the root, so the parent is None
 
1823
                        source_parent_id = None
 
1824
                    else:
 
1825
                        last_source_parent[0] = old_dirname
 
1826
                        last_source_parent[1] = source_parent_id
 
1827
                        last_source_parent[2] = source_parent_entry
 
1828
 
 
1829
                new_dirname = entry[0][0]
 
1830
                if new_dirname == last_target_parent[0]:
 
1831
                    target_parent_id = last_target_parent[1]
 
1832
                else:
 
1833
                    # TODO: We don't always need to do the lookup, because the
 
1834
                    #       parent entry will be the same as the source entry.
 
1835
                    target_parent_entry = state._get_entry(target_index,
 
1836
                                                           path_utf8=new_dirname)
 
1837
                    target_parent_id = target_parent_entry[0][2]
 
1838
                    if target_parent_id == entry[0][2]:
 
1839
                        # This is the root, so the parent is None
 
1840
                        target_parent_id = None
 
1841
                    else:
 
1842
                        last_target_parent[0] = new_dirname
 
1843
                        last_target_parent[1] = target_parent_id
 
1844
                        last_target_parent[2] = target_parent_entry
 
1845
 
 
1846
                source_exec = source_details[3]
 
1847
                return ((entry[0][2], (old_path, path), content_change,
 
1848
                        (True, True),
 
1849
                        (source_parent_id, target_parent_id),
 
1850
                        (old_basename, entry[0][1]),
 
1851
                        (_minikind_to_kind[source_minikind], target_kind),
 
1852
                        (source_exec, target_exec)),)
 
1853
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1854
                # looks like a new file
 
1855
                if path_info is not None:
 
1856
                    path = pathjoin(entry[0][0], entry[0][1])
 
1857
                    # parent id is the entry for the path in the target tree
 
1858
                    # TODO: these are the same for an entire directory: cache em.
 
1859
                    parent_id = state._get_entry(target_index,
 
1860
                                                 path_utf8=entry[0][0])[0][2]
 
1861
                    if parent_id == entry[0][2]:
 
1862
                        parent_id = None
 
1863
                    if use_filesystem_for_exec:
 
1864
                        # We need S_ISREG here, because we aren't sure if this
 
1865
                        # is a file or not.
 
1866
                        target_exec = bool(
 
1867
                            stat.S_ISREG(path_info[3].st_mode)
 
1868
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1869
                    else:
 
1870
                        target_exec = target_details[3]
 
1871
                    return ((entry[0][2], (None, path), True,
 
1872
                            (False, True),
 
1873
                            (None, parent_id),
 
1874
                            (None, entry[0][1]),
 
1875
                            (None, path_info[2]),
 
1876
                            (None, target_exec)),)
 
1877
                else:
 
1878
                    # but its not on disk: we deliberately treat this as just
 
1879
                    # never-present. (Why ?! - RBC 20070224)
 
1880
                    pass
 
1881
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1882
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1883
                # if its still on disk, *and* theres no other entry at this
 
1884
                # path [we dont know this in this routine at the moment -
 
1885
                # perhaps we should change this - then it would be an unknown.
 
1886
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1887
                # parent id is the entry for the path in the target tree
 
1888
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1889
                if parent_id == entry[0][2]:
 
1890
                    parent_id = None
 
1891
                return ((entry[0][2], (old_path, None), True,
 
1892
                        (True, False),
 
1893
                        (parent_id, None),
 
1894
                        (entry[0][1], None),
 
1895
                        (_minikind_to_kind[source_minikind], None),
 
1896
                        (source_details[3], None)),)
 
1897
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1898
                # a rename; could be a true rename, or a rename inherited from
 
1899
                # a renamed parent. TODO: handle this efficiently. Its not
 
1900
                # common case to rename dirs though, so a correct but slow
 
1901
                # implementation will do.
 
1902
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1903
                    search_specific_files.add(target_details[1])
 
1904
            elif source_minikind in 'r' and target_minikind in 'r':
 
1905
                # neither of the selected trees contain this file,
 
1906
                # so skip over it. This is not currently directly tested, but
 
1907
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1908
                pass
 
1909
            else:
 
1910
                raise AssertionError("don't know how to compare "
 
1911
                    "source_minikind=%r, target_minikind=%r"
 
1912
                    % (source_minikind, target_minikind))
 
1913
                ## import pdb;pdb.set_trace()
 
1914
            return ()
 
1915
        while search_specific_files:
 
1916
            # TODO: the pending list should be lexically sorted?
 
1917
            current_root = search_specific_files.pop()
 
1918
            searched_specific_files.add(current_root)
 
1919
            # process the entries for this containing directory: the rest will be
 
1920
            # found by their parents recursively.
 
1921
            root_entries = _entries_for_path(current_root)
 
1922
            root_abspath = self.target.abspath(current_root)
 
1923
            try:
 
1924
                root_stat = os.lstat(root_abspath)
 
1925
            except OSError, e:
 
1926
                if e.errno == errno.ENOENT:
 
1927
                    # the path does not exist: let _process_entry know that.
 
1928
                    root_dir_info = None
 
1929
                else:
 
1930
                    # some other random error: hand it up.
 
1931
                    raise
 
1932
            else:
 
1933
                root_dir_info = ('', current_root,
 
1934
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1935
                    root_abspath)
 
1936
            if not root_entries and not root_dir_info:
 
1937
                # this specified path is not present at all, skip it.
 
1938
                continue
 
1939
            path_handled = False
 
1940
            for entry in root_entries:
 
1941
                for result in _process_entry(entry, root_dir_info):
 
1942
                    # this check should probably be outside the loop: one
 
1943
                    # 'iterate two trees' api, and then _iter_changes filters
 
1944
                    # unchanged pairs. - RBC 20070226
 
1945
                    path_handled = True
 
1946
                    if (include_unchanged
 
1947
                        or result[2]                    # content change
 
1948
                        or result[3][0] != result[3][1] # versioned status
 
1949
                        or result[4][0] != result[4][1] # parent id
 
1950
                        or result[5][0] != result[5][1] # name
 
1951
                        or result[6][0] != result[6][1] # kind
 
1952
                        or result[7][0] != result[7][1] # executable
 
1953
                        ):
 
1954
                        result = (result[0],
 
1955
                            ((utf8_decode(result[1][0])[0]),
 
1956
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
1957
                        yield result
 
1958
            if want_unversioned and not path_handled:
 
1959
                new_executable = bool(
 
1960
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
1961
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
1962
                yield (None, (None, current_root), True, (False, False),
 
1963
                    (None, None),
 
1964
                    (None, splitpath(current_root)[-1]),
 
1965
                    (None, root_dir_info[2]), (None, new_executable))
 
1966
            dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
1967
            initial_key = (current_root, '', '')
 
1968
            block_index, _ = state._find_block_index_from_key(initial_key)
 
1969
            if block_index == 0:
 
1970
                # we have processed the total root already, but because the
 
1971
                # initial key matched it we should skip it here.
 
1972
                block_index +=1
 
1973
            try:
 
1974
                current_dir_info = dir_iterator.next()
 
1975
            except OSError, e:
 
1976
                if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
1977
                    # there may be directories in the inventory even though
 
1978
                    # this path is not a file on disk: so mark it as end of
 
1979
                    # iterator
 
1980
                    current_dir_info = None
 
1981
                else:
 
1982
                    raise
 
1983
            else:
 
1984
                if current_dir_info[0][0] == '':
 
1985
                    # remove .bzr from iteration
 
1986
                    bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
1987
                    assert current_dir_info[1][bzr_index][0] == '.bzr'
 
1988
                    del current_dir_info[1][bzr_index]
 
1989
            # walk until both the directory listing and the versioned metadata
 
1990
            # are exhausted. TODO: reevaluate this, perhaps we should stop when
 
1991
            # the versioned data runs out.
 
1992
            if (block_index < len(state._dirblocks) and
 
1993
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
1994
                current_block = state._dirblocks[block_index]
 
1995
            else:
 
1996
                current_block = None
 
1997
            while (current_dir_info is not None or
 
1998
                   current_block is not None):
 
1999
                if (current_dir_info and current_block
 
2000
                    and current_dir_info[0][0] != current_block[0]):
 
2001
                    if current_dir_info[0][0] < current_block[0] :
 
2002
                        # import pdb; pdb.set_trace()
 
2003
                        # print 'unversioned dir'
 
2004
                        # filesystem data refers to paths not covered by the dirblock.
 
2005
                        # this has two possibilities:
 
2006
                        # A) it is versioned but empty, so there is no block for it
 
2007
                        # B) it is not versioned.
 
2008
                        # in either case it was processed by the containing directories walk:
 
2009
                        # if it is root/foo, when we walked root we emitted it,
 
2010
                        # or if we ere given root/foo to walk specifically, we
 
2011
                        # emitted it when checking the walk-root entries
 
2012
                        # advance the iterator and loop - we dont need to emit it.
 
2013
                        try:
 
2014
                            current_dir_info = dir_iterator.next()
 
2015
                        except StopIteration:
 
2016
                            current_dir_info = None
 
2017
                    else:
 
2018
                        # We have a dirblock entry for this location, but there
 
2019
                        # is no filesystem path for this. This is most likely
 
2020
                        # because a directory was removed from the disk.
 
2021
                        # We don't have to report the missing directory,
 
2022
                        # because that should have already been handled, but we
 
2023
                        # need to handle all of the files that are contained
 
2024
                        # within.
 
2025
                        for current_entry in current_block[1]:
 
2026
                            # entry referring to file not present on disk.
 
2027
                            # advance the entry only, after processing.
 
2028
                            for result in _process_entry(current_entry, None):
 
2029
                                # this check should probably be outside the loop: one
 
2030
                                # 'iterate two trees' api, and then _iter_changes filters
 
2031
                                # unchanged pairs. - RBC 20070226
 
2032
                                if (include_unchanged
 
2033
                                    or result[2]                    # content change
 
2034
                                    or result[3][0] != result[3][1] # versioned status
 
2035
                                    or result[4][0] != result[4][1] # parent id
 
2036
                                    or result[5][0] != result[5][1] # name
 
2037
                                    or result[6][0] != result[6][1] # kind
 
2038
                                    or result[7][0] != result[7][1] # executable
 
2039
                                    ):
 
2040
                                    result = (result[0],
 
2041
                                        ((utf8_decode(result[1][0])[0]),
 
2042
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2043
                                    yield result
 
2044
                        block_index +=1
 
2045
                        if (block_index < len(state._dirblocks) and
 
2046
                            osutils.is_inside(current_root,
 
2047
                                              state._dirblocks[block_index][0])):
 
2048
                            current_block = state._dirblocks[block_index]
 
2049
                        else:
 
2050
                            current_block = None
 
2051
                    continue
 
2052
                entry_index = 0
 
2053
                if current_block and entry_index < len(current_block[1]):
 
2054
                    current_entry = current_block[1][entry_index]
 
2055
                else:
 
2056
                    current_entry = None
 
2057
                advance_entry = True
 
2058
                path_index = 0
 
2059
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2060
                    current_path_info = current_dir_info[1][path_index]
 
2061
                else:
 
2062
                    current_path_info = None
 
2063
                advance_path = True
 
2064
                path_handled = False
 
2065
                while (current_entry is not None or
 
2066
                    current_path_info is not None):
 
2067
                    if current_entry is None:
 
2068
                        # the check for path_handled when the path is adnvaced
 
2069
                        # will yield this path if needed.
 
2070
                        pass
 
2071
                    elif current_path_info is None:
 
2072
                        # no path is fine: the per entry code will handle it.
 
2073
                        for result in _process_entry(current_entry, current_path_info):
 
2074
                            # this check should probably be outside the loop: one
 
2075
                            # 'iterate two trees' api, and then _iter_changes filters
 
2076
                            # unchanged pairs. - RBC 20070226
 
2077
                            if (include_unchanged
 
2078
                                or result[2]                    # content change
 
2079
                                or result[3][0] != result[3][1] # versioned status
 
2080
                                or result[4][0] != result[4][1] # parent id
 
2081
                                or result[5][0] != result[5][1] # name
 
2082
                                or result[6][0] != result[6][1] # kind
 
2083
                                or result[7][0] != result[7][1] # executable
 
2084
                                ):
 
2085
                                result = (result[0],
 
2086
                                    ((utf8_decode(result[1][0])[0]),
 
2087
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2088
                                yield result
 
2089
                    elif current_entry[0][1] != current_path_info[1]:
 
2090
                        if current_path_info[1] < current_entry[0][1]:
 
2091
                            # extra file on disk: pass for now, but only
 
2092
                            # increment the path, not the entry
 
2093
                            # import pdb; pdb.set_trace()
 
2094
                            # print 'unversioned file'
 
2095
                            advance_entry = False
 
2096
                        else:
 
2097
                            # entry referring to file not present on disk.
 
2098
                            # advance the entry only, after processing.
 
2099
                            for result in _process_entry(current_entry, None):
 
2100
                                # this check should probably be outside the loop: one
 
2101
                                # 'iterate two trees' api, and then _iter_changes filters
 
2102
                                # unchanged pairs. - RBC 20070226
 
2103
                                path_handled = True
 
2104
                                if (include_unchanged
 
2105
                                    or result[2]                    # content change
 
2106
                                    or result[3][0] != result[3][1] # versioned status
 
2107
                                    or result[4][0] != result[4][1] # parent id
 
2108
                                    or result[5][0] != result[5][1] # name
 
2109
                                    or result[6][0] != result[6][1] # kind
 
2110
                                    or result[7][0] != result[7][1] # executable
 
2111
                                    ):
 
2112
                                    result = (result[0],
 
2113
                                        ((utf8_decode(result[1][0])[0]),
 
2114
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2115
                                    yield result
 
2116
                            advance_path = False
 
2117
                    else:
 
2118
                        for result in _process_entry(current_entry, current_path_info):
 
2119
                            # this check should probably be outside the loop: one
 
2120
                            # 'iterate two trees' api, and then _iter_changes filters
 
2121
                            # unchanged pairs. - RBC 20070226
 
2122
                            path_handled = True
 
2123
                            if (include_unchanged
 
2124
                                or result[2]                    # content change
 
2125
                                or result[3][0] != result[3][1] # versioned status
 
2126
                                or result[4][0] != result[4][1] # parent id
 
2127
                                or result[5][0] != result[5][1] # name
 
2128
                                or result[6][0] != result[6][1] # kind
 
2129
                                or result[7][0] != result[7][1] # executable
 
2130
                                ):
 
2131
                                result = (result[0],
 
2132
                                    ((utf8_decode(result[1][0])[0]),
 
2133
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2134
                                yield result
 
2135
                    if advance_entry and current_entry is not None:
 
2136
                        entry_index += 1
 
2137
                        if entry_index < len(current_block[1]):
 
2138
                            current_entry = current_block[1][entry_index]
 
2139
                        else:
 
2140
                            current_entry = None
 
2141
                    else:
 
2142
                        advance_entry = True # reset the advance flaga
 
2143
                    if advance_path and current_path_info is not None:
 
2144
                        if not path_handled:
 
2145
                            # unversioned in all regards
 
2146
                            if want_unversioned:
 
2147
                                new_executable = bool(
 
2148
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2149
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2150
                                if want_unversioned:
 
2151
                                    yield (None, (None, current_path_info[0]),
 
2152
                                        True,
 
2153
                                        (False, False),
 
2154
                                        (None, None),
 
2155
                                        (None, current_path_info[1]),
 
2156
                                        (None, current_path_info[2]),
 
2157
                                        (None, new_executable))
 
2158
                            # dont descend into this unversioned path if it is
 
2159
                            # a dir
 
2160
                            if current_path_info[2] == 'directory':
 
2161
                                del current_dir_info[1][path_index]
 
2162
                                path_index -= 1
 
2163
                        path_index += 1
 
2164
                        if path_index < len(current_dir_info[1]):
 
2165
                            current_path_info = current_dir_info[1][path_index]
 
2166
                        else:
 
2167
                            current_path_info = None
 
2168
                        path_handled = False
 
2169
                    else:
 
2170
                        advance_path = True # reset the advance flagg.
 
2171
                if current_block is not None:
 
2172
                    block_index += 1
 
2173
                    if (block_index < len(state._dirblocks) and
 
2174
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2175
                        current_block = state._dirblocks[block_index]
 
2176
                    else:
 
2177
                        current_block = None
 
2178
                if current_dir_info is not None:
 
2179
                    try:
 
2180
                        current_dir_info = dir_iterator.next()
 
2181
                    except StopIteration:
 
2182
                        current_dir_info = None
 
2183
 
 
2184
 
 
2185
    @staticmethod
 
2186
    def is_compatible(source, target):
 
2187
        # the target must be a dirstate working tree
 
2188
        if not isinstance(target, WorkingTree4):
 
2189
            return False
 
2190
        # the source must be a revtreee or dirstate rev tree.
 
2191
        if not isinstance(source,
 
2192
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2193
            return False
 
2194
        # the source revid must be in the target dirstate
 
2195
        if not (source._revision_id == NULL_REVISION or
 
2196
            source._revision_id in target.get_parent_ids()):
 
2197
            # TODO: what about ghosts? it may well need to 
 
2198
            # check for them explicitly.
 
2199
            return False
 
2200
        return True
 
2201
 
 
2202
InterTree.register_optimiser(InterDirStateTree)
 
2203
 
 
2204
 
 
2205
class Converter3to4(object):
 
2206
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2207
 
 
2208
    def __init__(self):
 
2209
        self.target_format = WorkingTreeFormat4()
 
2210
 
 
2211
    def convert(self, tree):
 
2212
        # lock the control files not the tree, so that we dont get tree
 
2213
        # on-unlock behaviours, and so that noone else diddles with the 
 
2214
        # tree during upgrade.
 
2215
        tree._control_files.lock_write()
 
2216
        try:
 
2217
            self.create_dirstate_data(tree)
 
2218
            self.update_format(tree)
 
2219
            self.remove_xml_files(tree)
 
2220
        finally:
 
2221
            tree._control_files.unlock()
 
2222
 
 
2223
    def create_dirstate_data(self, tree):
 
2224
        """Create the dirstate based data for tree."""
 
2225
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2226
            ).local_abspath('dirstate')
 
2227
        state = dirstate.DirState.from_tree(tree, local_path)
 
2228
        state.save()
 
2229
        state.unlock()
 
2230
 
 
2231
    def remove_xml_files(self, tree):
 
2232
        """Remove the oldformat 3 data."""
 
2233
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2234
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2235
            'pending-merges', 'stat-cache']:
 
2236
            try:
 
2237
                transport.delete(path)
 
2238
            except errors.NoSuchFile:
 
2239
                # some files are optional - just deal.
 
2240
                pass
 
2241
 
 
2242
    def update_format(self, tree):
 
2243
        """Change the format marker."""
 
2244
        tree._control_files.put_utf8('format',
 
2245
            self.target_format.get_format_string())