~ubuntu-branches/ubuntu/hardy/bzr/hardy-updates

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Bazaar Package Importer
  • Author(s): Etienne Goyer
  • Date: 2007-03-07 08:55:49 UTC
  • mto: This revision was merged to the branch mainline in revision 35.
  • Revision ID: james.westby@ubuntu.com-20070307085549-2y0nzag1hotluiah
Tags: upstream-0.15~rc1
ImportĀ upstreamĀ versionĀ 0.15~rc1

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 (self._repo_supports_tree_reference and
 
229
            kind == 'directory' and
 
230
            self._directory_is_tree_reference(path)):
 
231
            kind = 'tree-reference'
 
232
        return kind, executable, stat_value
 
233
 
 
234
    @needs_write_lock
 
235
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
236
        # mark the tree as dirty post commit - commit
 
237
        # can change the current versioned list by doing deletes.
 
238
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
239
        self._make_dirty(reset_inventory=True)
 
240
        return result
 
241
 
 
242
    def current_dirstate(self):
 
243
        """Return the current dirstate object.
 
244
 
 
245
        This is not part of the tree interface and only exposed for ease of
 
246
        testing.
 
247
 
 
248
        :raises errors.NotWriteLocked: when not in a lock.
 
249
        """
 
250
        self._must_be_locked()
 
251
        return self._current_dirstate()
 
252
 
 
253
    def _current_dirstate(self):
 
254
        """Internal function that does not check lock status.
 
255
 
 
256
        This is needed for break_lock which also needs the dirstate.
 
257
        """
 
258
        if self._dirstate is not None:
 
259
            return self._dirstate
 
260
        local_path = self.bzrdir.get_workingtree_transport(None
 
261
            ).local_abspath('dirstate')
 
262
        self._dirstate = dirstate.DirState.on_file(local_path)
 
263
        return self._dirstate
 
264
 
 
265
    def _directory_is_tree_reference(self, relpath):
 
266
        # as a special case, if a directory contains control files then 
 
267
        # it's a tree reference, except that the root of the tree is not
 
268
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
269
        # TODO: We could ask all the control formats whether they
 
270
        # recognize this directory, but at the moment there's no cheap api
 
271
        # to do that.  Since we probably can only nest bzr checkouts and
 
272
        # they always use this name it's ok for now.  -- mbp 20060306
 
273
        #
 
274
        # FIXME: There is an unhandled case here of a subdirectory
 
275
        # containing .bzr but not a branch; that will probably blow up
 
276
        # when you try to commit it.  It might happen if there is a
 
277
        # checkout in a subdirectory.  This can be avoided by not adding
 
278
        # it.  mbp 20070306
 
279
 
 
280
    def filter_unversioned_files(self, paths):
 
281
        """Filter out paths that are versioned.
 
282
 
 
283
        :return: set of paths.
 
284
        """
 
285
        # TODO: make a generic multi-bisect routine roughly that should list
 
286
        # the paths, then process one half at a time recursively, and feed the
 
287
        # results of each bisect in further still
 
288
        paths = sorted(paths)
 
289
        result = set()
 
290
        state = self.current_dirstate()
 
291
        # TODO we want a paths_to_dirblocks helper I think
 
292
        for path in paths:
 
293
            dirname, basename = os.path.split(path.encode('utf8'))
 
294
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
295
                dirname, basename, 0)
 
296
            if not path_is_versioned:
 
297
                result.add(path)
 
298
        return result
 
299
 
 
300
    def flush(self):
 
301
        """Write all cached data to disk."""
 
302
        if self._control_files._lock_mode != 'w':
 
303
            raise errors.NotWriteLocked(self)
 
304
        self.current_dirstate().save()
 
305
        self._inventory = None
 
306
        self._dirty = False
 
307
 
 
308
    def _generate_inventory(self):
 
309
        """Create and set self.inventory from the dirstate object.
 
310
        
 
311
        This is relatively expensive: we have to walk the entire dirstate.
 
312
        Ideally we would not, and can deprecate this function.
 
313
        """
 
314
        #: uncomment to trap on inventory requests.
 
315
        # import pdb;pdb.set_trace()
 
316
        state = self.current_dirstate()
 
317
        state._read_dirblocks_if_needed()
 
318
        root_key, current_entry = self._get_entry(path='')
 
319
        current_id = root_key[2]
 
320
        assert current_entry[0][0] == 'd' # directory
 
321
        inv = Inventory(root_id=current_id)
 
322
        # Turn some things into local variables
 
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
324
        factory = entry_factory
 
325
        utf8_decode = cache_utf8._utf8_decode
 
326
        inv_byid = inv._byid
 
327
        # we could do this straight out of the dirstate; it might be fast
 
328
        # and should be profiled - RBC 20070216
 
329
        parent_ies = {'' : inv.root}
 
330
        for block in state._dirblocks[1:]: # skip the root
 
331
            dirname = block[0]
 
332
            try:
 
333
                parent_ie = parent_ies[dirname]
 
334
            except KeyError:
 
335
                # all the paths in this block are not versioned in this tree
 
336
                continue
 
337
            for key, entry in block[1]:
 
338
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
339
                if minikind in ('a', 'r'): # absent, relocated
 
340
                    # a parent tree only entry
 
341
                    continue
 
342
                name = key[1]
 
343
                name_unicode = utf8_decode(name)[0]
 
344
                file_id = key[2]
 
345
                kind = minikind_to_kind[minikind]
 
346
                inv_entry = factory[kind](file_id, name_unicode,
 
347
                                          parent_ie.file_id)
 
348
                if kind == 'file':
 
349
                    # not strictly needed: working tree
 
350
                    #entry.executable = executable
 
351
                    #entry.text_size = size
 
352
                    #entry.text_sha1 = sha1
 
353
                    pass
 
354
                elif kind == 'directory':
 
355
                    # add this entry to the parent map.
 
356
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
357
                elif kind == 'tree-reference':
 
358
                    assert self._repo_supports_tree_reference
 
359
                    inv_entry.reference_revision = link_or_sha1 or None
 
360
                else:
 
361
                    assert 'unknown kind'
 
362
                # These checks cost us around 40ms on a 55k entry tree
 
363
                assert file_id not in inv_byid, ('file_id %s already in'
 
364
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
365
                assert name_unicode not in parent_ie.children
 
366
                inv_byid[file_id] = inv_entry
 
367
                parent_ie.children[name_unicode] = inv_entry
 
368
        self._inventory = inv
 
369
 
 
370
    def _get_entry(self, file_id=None, path=None):
 
371
        """Get the dirstate row for file_id or path.
 
372
 
 
373
        If either file_id or path is supplied, it is used as the key to lookup.
 
374
        If both are supplied, the fastest lookup is used, and an error is
 
375
        raised if they do not both point at the same row.
 
376
        
 
377
        :param file_id: An optional unicode file_id to be looked up.
 
378
        :param path: An optional unicode path to be looked up.
 
379
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
380
        """
 
381
        if file_id is None and path is None:
 
382
            raise errors.BzrError('must supply file_id or path')
 
383
        state = self.current_dirstate()
 
384
        if path is not None:
 
385
            path = path.encode('utf8')
 
386
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
387
 
 
388
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
389
        # check file id is valid unconditionally.
 
390
        entry = self._get_entry(file_id=file_id, path=path)
 
391
        assert entry[0] is not None, 'what error should this raise'
 
392
        if path is None:
 
393
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
394
 
 
395
        file_abspath = self.abspath(path)
 
396
        state = self.current_dirstate()
 
397
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
398
                                          stat_value=stat_value)
 
399
        if entry[1][0][0] == 'f':
 
400
            return link_or_sha1
 
401
        return None
 
402
 
 
403
    def _get_inventory(self):
 
404
        """Get the inventory for the tree. This is only valid within a lock."""
 
405
        if self._inventory is not None:
 
406
            return self._inventory
 
407
        self._must_be_locked()
 
408
        self._generate_inventory()
 
409
        return self._inventory
 
410
 
 
411
    inventory = property(_get_inventory,
 
412
                         doc="Inventory of this Tree")
 
413
 
 
414
    @needs_read_lock
 
415
    def get_parent_ids(self):
 
416
        """See Tree.get_parent_ids.
 
417
        
 
418
        This implementation requests the ids list from the dirstate file.
 
419
        """
 
420
        return self.current_dirstate().get_parent_ids()
 
421
 
 
422
    def get_reference_revision(self, file_id, path=None):
 
423
        # referenced tree's revision is whatever's currently there
 
424
        return self.get_nested_tree(file_id, path).last_revision()
 
425
 
 
426
    def get_nested_tree(self, file_id, path=None):
 
427
        if path is None:
 
428
            path = self.id2path(file_id)
 
429
        # else: check file_id is at path?
 
430
        return WorkingTree.open(self.abspath(path))
 
431
 
 
432
    @needs_read_lock
 
433
    def get_root_id(self):
 
434
        """Return the id of this trees root"""
 
435
        return self._get_entry(path='')[0][2]
 
436
 
 
437
    def has_id(self, file_id):
 
438
        state = self.current_dirstate()
 
439
        file_id = osutils.safe_file_id(file_id)
 
440
        row, parents = self._get_entry(file_id=file_id)
 
441
        if row is None:
 
442
            return False
 
443
        return osutils.lexists(pathjoin(
 
444
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
445
 
 
446
    @needs_read_lock
 
447
    def id2path(self, file_id):
 
448
        file_id = osutils.safe_file_id(file_id)
 
449
        state = self.current_dirstate()
 
450
        entry = self._get_entry(file_id=file_id)
 
451
        if entry == (None, None):
 
452
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
453
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
454
        return path_utf8.decode('utf8')
 
455
 
 
456
    @needs_read_lock
 
457
    def __iter__(self):
 
458
        """Iterate through file_ids for this tree.
 
459
 
 
460
        file_ids are in a WorkingTree if they are in the working inventory
 
461
        and the working file exists.
 
462
        """
 
463
        result = []
 
464
        for key, tree_details in self.current_dirstate()._iter_entries():
 
465
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
466
                # not relevant to the working tree
 
467
                continue
 
468
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
469
            if osutils.lexists(path):
 
470
                result.append(key[2])
 
471
        return iter(result)
 
472
 
 
473
    def iter_references(self):
 
474
        for key, tree_details in self.current_dirstate()._iter_entries():
 
475
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
476
                # not relevant to the working tree
 
477
                continue
 
478
            if not key[1]:
 
479
                # the root is not a reference.
 
480
                continue
 
481
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
482
            try:
 
483
                if self._kind(path) == 'tree-reference':
 
484
                    yield path, key[2]
 
485
            except errors.NoSuchFile:
 
486
                # path is missing on disk.
 
487
                continue
 
488
 
 
489
    @needs_read_lock
 
490
    def kind(self, file_id):
 
491
        """Return the kind of a file.
 
492
 
 
493
        This is always the actual kind that's on disk, regardless of what it
 
494
        was added as.
 
495
        """
 
496
        relpath = self.id2path(file_id)
 
497
        assert relpath != None, \
 
498
            "path for id {%s} is None!" % file_id
 
499
        return self._kind(relpath)
 
500
 
 
501
    def _kind(self, relpath):
 
502
        abspath = self.abspath(relpath)
 
503
        kind = file_kind(abspath)
 
504
        if (self._repo_supports_tree_reference and
 
505
            kind == 'directory' and
 
506
            self._directory_is_tree_reference(relpath)):
 
507
            kind = 'tree-reference'
 
508
        return kind
 
509
 
 
510
    @needs_read_lock
 
511
    def _last_revision(self):
 
512
        """See Mutable.last_revision."""
 
513
        parent_ids = self.current_dirstate().get_parent_ids()
 
514
        if parent_ids:
 
515
            return parent_ids[0]
 
516
        else:
 
517
            return None
 
518
 
 
519
    def lock_read(self):
 
520
        """See Branch.lock_read, and WorkingTree.unlock."""
 
521
        self.branch.lock_read()
 
522
        try:
 
523
            self._control_files.lock_read()
 
524
            try:
 
525
                state = self.current_dirstate()
 
526
                if not state._lock_token:
 
527
                    state.lock_read()
 
528
                # set our support for tree references from the repository in
 
529
                # use.
 
530
                self._repo_supports_tree_reference = getattr(
 
531
                    self.branch.repository._format, "support_tree_reference",
 
532
                    False)
 
533
            except:
 
534
                self._control_files.unlock()
 
535
                raise
 
536
        except:
 
537
            self.branch.unlock()
 
538
            raise
 
539
 
 
540
    def _lock_self_write(self):
 
541
        """This should be called after the branch is locked."""
 
542
        try:
 
543
            self._control_files.lock_write()
 
544
            try:
 
545
                state = self.current_dirstate()
 
546
                if not state._lock_token:
 
547
                    state.lock_write()
 
548
                # set our support for tree references from the repository in
 
549
                # use.
 
550
                self._repo_supports_tree_reference = getattr(
 
551
                    self.branch.repository._format, "support_tree_reference",
 
552
                    False)
 
553
            except:
 
554
                self._control_files.unlock()
 
555
                raise
 
556
        except:
 
557
            self.branch.unlock()
 
558
            raise
 
559
 
 
560
    def lock_tree_write(self):
 
561
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
562
        self.branch.lock_read()
 
563
        self._lock_self_write()
 
564
 
 
565
    def lock_write(self):
 
566
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
567
        self.branch.lock_write()
 
568
        self._lock_self_write()
 
569
 
 
570
    @needs_tree_write_lock
 
571
    def move(self, from_paths, to_dir, after=False):
 
572
        """See WorkingTree.move()."""
 
573
        result = []
 
574
        if not from_paths:
 
575
            return result
 
576
 
 
577
        state = self.current_dirstate()
 
578
 
 
579
        assert not isinstance(from_paths, basestring)
 
580
        to_dir_utf8 = to_dir.encode('utf8')
 
581
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
582
        id_index = state._get_id_index()
 
583
        # check destination directory
 
584
        # get the details for it
 
585
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
586
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
587
        if not entry_present:
 
588
            raise errors.BzrMoveFailedError('', to_dir,
 
589
                errors.NotVersionedError(to_dir))
 
590
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
591
        # get a handle on the block itself.
 
592
        to_block_index = state._ensure_block(
 
593
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
594
        to_block = state._dirblocks[to_block_index]
 
595
        to_abs = self.abspath(to_dir)
 
596
        if not isdir(to_abs):
 
597
            raise errors.BzrMoveFailedError('',to_dir,
 
598
                errors.NotADirectory(to_abs))
 
599
 
 
600
        if to_entry[1][0][0] != 'd':
 
601
            raise errors.BzrMoveFailedError('',to_dir,
 
602
                errors.NotADirectory(to_abs))
 
603
 
 
604
        if self._inventory is not None:
 
605
            update_inventory = True
 
606
            inv = self.inventory
 
607
            to_dir_ie = inv[to_dir_id]
 
608
            to_dir_id = to_entry[0][2]
 
609
        else:
 
610
            update_inventory = False
 
611
 
 
612
        rollbacks = []
 
613
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
614
                     fingerprint, packed_stat, size,
 
615
                     to_block, to_key, to_path_utf8):
 
616
            state._make_absent(old_entry)
 
617
            from_key = old_entry[0]
 
618
            rollbacks.append(
 
619
                lambda:state.update_minimal(from_key,
 
620
                    minikind,
 
621
                    executable=executable,
 
622
                    fingerprint=fingerprint,
 
623
                    packed_stat=packed_stat,
 
624
                    size=size,
 
625
                    path_utf8=from_path_utf8))
 
626
            state.update_minimal(to_key,
 
627
                    minikind,
 
628
                    executable=executable,
 
629
                    fingerprint=fingerprint,
 
630
                    packed_stat=packed_stat,
 
631
                    size=size,
 
632
                    path_utf8=to_path_utf8)
 
633
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
634
            new_entry = to_block[1][added_entry_index]
 
635
            rollbacks.append(lambda:state._make_absent(new_entry))
 
636
 
 
637
        # create rename entries and tuples
 
638
        for from_rel in from_paths:
 
639
            # from_rel is 'pathinroot/foo/bar'
 
640
            from_rel_utf8 = from_rel.encode('utf8')
 
641
            from_dirname, from_tail = osutils.split(from_rel)
 
642
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
643
            from_entry = self._get_entry(path=from_rel)
 
644
            if from_entry == (None, None):
 
645
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
646
                    errors.NotVersionedError(path=str(from_rel)))
 
647
 
 
648
            from_id = from_entry[0][2]
 
649
            to_rel = pathjoin(to_dir, from_tail)
 
650
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
651
            item_to_entry = self._get_entry(path=to_rel)
 
652
            if item_to_entry != (None, None):
 
653
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
654
                    "Target is already versioned.")
 
655
 
 
656
            if from_rel == to_rel:
 
657
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
658
                    "Source and target are identical.")
 
659
 
 
660
            from_missing = not self.has_filename(from_rel)
 
661
            to_missing = not self.has_filename(to_rel)
 
662
            if after:
 
663
                move_file = False
 
664
            else:
 
665
                move_file = True
 
666
            if to_missing:
 
667
                if not move_file:
 
668
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
669
                        errors.NoSuchFile(path=to_rel,
 
670
                        extra="New file has not been created yet"))
 
671
                elif from_missing:
 
672
                    # neither path exists
 
673
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
674
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
675
            else:
 
676
                if from_missing: # implicitly just update our path mapping
 
677
                    move_file = False
 
678
                elif not after:
 
679
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
680
                        extra="(Use --after to update the Bazaar id)")
 
681
 
 
682
            rollbacks = []
 
683
            def rollback_rename():
 
684
                """A single rename has failed, roll it back."""
 
685
                # roll back everything, even if we encounter trouble doing one
 
686
                # of them.
 
687
                #
 
688
                # TODO: at least log the other exceptions rather than just
 
689
                # losing them mbp 20070307
 
690
                exc_info = None
 
691
                for rollback in reversed(rollbacks):
 
692
                    try:
 
693
                        rollback()
 
694
                    except Exception, e:
 
695
                        exc_info = sys.exc_info()
 
696
                if exc_info:
 
697
                    raise exc_info[0], exc_info[1], exc_info[2]
 
698
 
 
699
            # perform the disk move first - its the most likely failure point.
 
700
            if move_file:
 
701
                from_rel_abs = self.abspath(from_rel)
 
702
                to_rel_abs = self.abspath(to_rel)
 
703
                try:
 
704
                    osutils.rename(from_rel_abs, to_rel_abs)
 
705
                except OSError, e:
 
706
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
707
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
708
            try:
 
709
                # perform the rename in the inventory next if needed: its easy
 
710
                # to rollback
 
711
                if update_inventory:
 
712
                    # rename the entry
 
713
                    from_entry = inv[from_id]
 
714
                    current_parent = from_entry.parent_id
 
715
                    inv.rename(from_id, to_dir_id, from_tail)
 
716
                    rollbacks.append(
 
717
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
718
                # finally do the rename in the dirstate, which is a little
 
719
                # tricky to rollback, but least likely to need it.
 
720
                old_block_index, old_entry_index, dir_present, file_present = \
 
721
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
722
                old_block = state._dirblocks[old_block_index][1]
 
723
                old_entry = old_block[old_entry_index]
 
724
                from_key, old_entry_details = old_entry
 
725
                cur_details = old_entry_details[0]
 
726
                # remove the old row
 
727
                to_key = ((to_block[0],) + from_key[1:3])
 
728
                minikind = cur_details[0]
 
729
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
730
                         minikind=minikind,
 
731
                         executable=cur_details[3],
 
732
                         fingerprint=cur_details[1],
 
733
                         packed_stat=cur_details[4],
 
734
                         size=cur_details[2],
 
735
                         to_block=to_block,
 
736
                         to_key=to_key,
 
737
                         to_path_utf8=to_rel_utf8)
 
738
 
 
739
                if minikind == 'd':
 
740
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
741
                        """all entries in this block need updating.
 
742
 
 
743
                        TODO: This is pretty ugly, and doesn't support
 
744
                        reverting, but it works.
 
745
                        """
 
746
                        assert from_dir != '', "renaming root not supported"
 
747
                        from_key = (from_dir, '')
 
748
                        from_block_idx, present = \
 
749
                            state._find_block_index_from_key(from_key)
 
750
                        if not present:
 
751
                            # This is the old record, if it isn't present, then
 
752
                            # there is theoretically nothing to update.
 
753
                            # (Unless it isn't present because of lazy loading,
 
754
                            # but we don't do that yet)
 
755
                            return
 
756
                        from_block = state._dirblocks[from_block_idx]
 
757
                        to_block_index, to_entry_index, _, _ = \
 
758
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
759
                        to_block_index = state._ensure_block(
 
760
                            to_block_index, to_entry_index, to_dir_utf8)
 
761
                        to_block = state._dirblocks[to_block_index]
 
762
                        for entry in from_block[1]:
 
763
                            assert entry[0][0] == from_dir
 
764
                            cur_details = entry[1][0]
 
765
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
766
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
767
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
768
                            minikind = cur_details[0]
 
769
                            move_one(entry, from_path_utf8=from_path_utf8,
 
770
                                     minikind=minikind,
 
771
                                     executable=cur_details[3],
 
772
                                     fingerprint=cur_details[1],
 
773
                                     packed_stat=cur_details[4],
 
774
                                     size=cur_details[2],
 
775
                                     to_block=to_block,
 
776
                                     to_key=to_key,
 
777
                                     to_path_utf8=to_rel_utf8)
 
778
                            if minikind == 'd':
 
779
                                # We need to move all the children of this
 
780
                                # entry
 
781
                                update_dirblock(from_path_utf8, to_key,
 
782
                                                to_path_utf8)
 
783
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
784
            except:
 
785
                rollback_rename()
 
786
                raise
 
787
            result.append((from_rel, to_rel))
 
788
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
789
            self._make_dirty(reset_inventory=False)
 
790
 
 
791
        return result
 
792
 
 
793
    def _must_be_locked(self):
 
794
        if not self._control_files._lock_count:
 
795
            raise errors.ObjectNotLocked(self)
 
796
 
 
797
    def _new_tree(self):
 
798
        """Initialize the state in this tree to be a new tree."""
 
799
        self._dirty = True
 
800
 
 
801
    @needs_read_lock
 
802
    def path2id(self, path):
 
803
        """Return the id for path in this tree."""
 
804
        path = path.strip('/')
 
805
        entry = self._get_entry(path=path)
 
806
        if entry == (None, None):
 
807
            return None
 
808
        return entry[0][2]
 
809
 
 
810
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
811
        """See Tree.paths2ids().
 
812
 
 
813
        This specialisation fast-paths the case where all the trees are in the
 
814
        dirstate.
 
815
        """
 
816
        if paths is None:
 
817
            return None
 
818
        parents = self.get_parent_ids()
 
819
        for tree in trees:
 
820
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
821
                parents):
 
822
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
823
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
824
        # -- make all paths utf8 --
 
825
        paths_utf8 = set()
 
826
        for path in paths:
 
827
            paths_utf8.add(path.encode('utf8'))
 
828
        paths = paths_utf8
 
829
        # -- paths is now a utf8 path set --
 
830
        # -- get the state object and prepare it.
 
831
        state = self.current_dirstate()
 
832
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
833
            and '' not in paths):
 
834
            paths2ids = self._paths2ids_using_bisect
 
835
        else:
 
836
            paths2ids = self._paths2ids_in_memory
 
837
        return paths2ids(paths, search_indexes,
 
838
                         require_versioned=require_versioned)
 
839
 
 
840
    def _paths2ids_in_memory(self, paths, search_indexes,
 
841
                             require_versioned=True):
 
842
        state = self.current_dirstate()
 
843
        state._read_dirblocks_if_needed()
 
844
        def _entries_for_path(path):
 
845
            """Return a list with all the entries that match path for all ids.
 
846
            """
 
847
            dirname, basename = os.path.split(path)
 
848
            key = (dirname, basename, '')
 
849
            block_index, present = state._find_block_index_from_key(key)
 
850
            if not present:
 
851
                # the block which should contain path is absent.
 
852
                return []
 
853
            result = []
 
854
            block = state._dirblocks[block_index][1]
 
855
            entry_index, _ = state._find_entry_index(key, block)
 
856
            # we may need to look at multiple entries at this path: walk while the paths match.
 
857
            while (entry_index < len(block) and
 
858
                block[entry_index][0][0:2] == key[0:2]):
 
859
                result.append(block[entry_index])
 
860
                entry_index += 1
 
861
            return result
 
862
        if require_versioned:
 
863
            # -- check all supplied paths are versioned in a search tree. --
 
864
            all_versioned = True
 
865
            for path in paths:
 
866
                path_entries = _entries_for_path(path)
 
867
                if not path_entries:
 
868
                    # this specified path is not present at all: error
 
869
                    all_versioned = False
 
870
                    break
 
871
                found_versioned = False
 
872
                # for each id at this path
 
873
                for entry in path_entries:
 
874
                    # for each tree.
 
875
                    for index in search_indexes:
 
876
                        if entry[1][index][0] != 'a': # absent
 
877
                            found_versioned = True
 
878
                            # all good: found a versioned cell
 
879
                            break
 
880
                if not found_versioned:
 
881
                    # none of the indexes was not 'absent' at all ids for this
 
882
                    # path.
 
883
                    all_versioned = False
 
884
                    break
 
885
            if not all_versioned:
 
886
                raise errors.PathsNotVersionedError(paths)
 
887
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
888
        search_paths = set()
 
889
        for path in paths:
 
890
            other_paths = paths.difference(set([path]))
 
891
            if not osutils.is_inside_any(other_paths, path):
 
892
                # this is a top level path, we must check it.
 
893
                search_paths.add(path)
 
894
        # sketch: 
 
895
        # for all search_indexs in each path at or under each element of
 
896
        # search_paths, if the detail is relocated: add the id, and add the
 
897
        # relocated path as one to search if its not searched already. If the
 
898
        # detail is not relocated, add the id.
 
899
        searched_paths = set()
 
900
        found_ids = set()
 
901
        def _process_entry(entry):
 
902
            """Look at search_indexes within entry.
 
903
 
 
904
            If a specific tree's details are relocated, add the relocation
 
905
            target to search_paths if not searched already. If it is absent, do
 
906
            nothing. Otherwise add the id to found_ids.
 
907
            """
 
908
            for index in search_indexes:
 
909
                if entry[1][index][0] == 'r': # relocated
 
910
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
911
                        search_paths.add(entry[1][index][1])
 
912
                elif entry[1][index][0] != 'a': # absent
 
913
                    found_ids.add(entry[0][2])
 
914
        while search_paths:
 
915
            current_root = search_paths.pop()
 
916
            searched_paths.add(current_root)
 
917
            # process the entries for this containing directory: the rest will be
 
918
            # found by their parents recursively.
 
919
            root_entries = _entries_for_path(current_root)
 
920
            if not root_entries:
 
921
                # this specified path is not present at all, skip it.
 
922
                continue
 
923
            for entry in root_entries:
 
924
                _process_entry(entry)
 
925
            initial_key = (current_root, '', '')
 
926
            block_index, _ = state._find_block_index_from_key(initial_key)
 
927
            while (block_index < len(state._dirblocks) and
 
928
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
929
                for entry in state._dirblocks[block_index][1]:
 
930
                    _process_entry(entry)
 
931
                block_index += 1
 
932
        return found_ids
 
933
 
 
934
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
935
                                require_versioned=True):
 
936
        state = self.current_dirstate()
 
937
        found_ids = set()
 
938
 
 
939
        split_paths = sorted(osutils.split(p) for p in paths)
 
940
        found = state._bisect_recursive(split_paths)
 
941
 
 
942
        if require_versioned:
 
943
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
944
            for dir_name in split_paths:
 
945
                if dir_name not in found_dir_names:
 
946
                    raise errors.PathsNotVersionedError(paths)
 
947
 
 
948
        for dir_name_id, trees_info in found.iteritems():
 
949
            for index in search_indexes:
 
950
                if trees_info[index][0] not in ('r', 'a'):
 
951
                    found_ids.add(dir_name_id[2])
 
952
        return found_ids
 
953
 
 
954
    def read_working_inventory(self):
 
955
        """Read the working inventory.
 
956
        
 
957
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
958
        """
 
959
        return self.inventory
 
960
 
 
961
    @needs_read_lock
 
962
    def revision_tree(self, revision_id):
 
963
        """See Tree.revision_tree.
 
964
 
 
965
        WorkingTree4 supplies revision_trees for any basis tree.
 
966
        """
 
967
        revision_id = osutils.safe_revision_id(revision_id)
 
968
        dirstate = self.current_dirstate()
 
969
        parent_ids = dirstate.get_parent_ids()
 
970
        if revision_id not in parent_ids:
 
971
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
972
        if revision_id in dirstate.get_ghosts():
 
973
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
974
        return DirStateRevisionTree(dirstate, revision_id,
 
975
            self.branch.repository)
 
976
 
 
977
    @needs_tree_write_lock
 
978
    def set_last_revision(self, new_revision):
 
979
        """Change the last revision in the working tree."""
 
980
        new_revision = osutils.safe_revision_id(new_revision)
 
981
        parents = self.get_parent_ids()
 
982
        if new_revision in (NULL_REVISION, None):
 
983
            assert len(parents) < 2, (
 
984
                "setting the last parent to none with a pending merge is "
 
985
                "unsupported.")
 
986
            self.set_parent_ids([])
 
987
        else:
 
988
            self.set_parent_ids([new_revision] + parents[1:],
 
989
                allow_leftmost_as_ghost=True)
 
990
 
 
991
    @needs_tree_write_lock
 
992
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
993
        """Set the parent ids to revision_ids.
 
994
        
 
995
        See also set_parent_trees. This api will try to retrieve the tree data
 
996
        for each element of revision_ids from the trees repository. If you have
 
997
        tree data already available, it is more efficient to use
 
998
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
999
        an easier API to use.
 
1000
 
 
1001
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1002
            working tree. Any of these may be ghosts.
 
1003
        """
 
1004
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1005
        trees = []
 
1006
        for revision_id in revision_ids:
 
1007
            try:
 
1008
                revtree = self.branch.repository.revision_tree(revision_id)
 
1009
                # TODO: jam 20070213 KnitVersionedFile raises
 
1010
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1011
                #       given revision_id is not present. Should Repository be
 
1012
                #       catching it and re-raising NoSuchRevision?
 
1013
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1014
                revtree = None
 
1015
            trees.append((revision_id, revtree))
 
1016
        self.current_dirstate()._validate()
 
1017
        self.set_parent_trees(trees,
 
1018
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1019
        self.current_dirstate()._validate()
 
1020
 
 
1021
    @needs_tree_write_lock
 
1022
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1023
        """Set the parents of the working tree.
 
1024
 
 
1025
        :param parents_list: A list of (revision_id, tree) tuples.
 
1026
            If tree is None, then that element is treated as an unreachable
 
1027
            parent tree - i.e. a ghost.
 
1028
        """
 
1029
        dirstate = self.current_dirstate()
 
1030
        dirstate._validate()
 
1031
        if len(parents_list) > 0:
 
1032
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1033
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1034
        real_trees = []
 
1035
        ghosts = []
 
1036
        # convert absent trees to the null tree, which we convert back to
 
1037
        # missing on access.
 
1038
        for rev_id, tree in parents_list:
 
1039
            rev_id = osutils.safe_revision_id(rev_id)
 
1040
            if tree is not None:
 
1041
                real_trees.append((rev_id, tree))
 
1042
            else:
 
1043
                real_trees.append((rev_id,
 
1044
                    self.branch.repository.revision_tree(None)))
 
1045
                ghosts.append(rev_id)
 
1046
        dirstate._validate()
 
1047
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1048
        dirstate._validate()
 
1049
        self._make_dirty(reset_inventory=False)
 
1050
        dirstate._validate()
 
1051
 
 
1052
    def _set_root_id(self, file_id):
 
1053
        """See WorkingTree.set_root_id."""
 
1054
        state = self.current_dirstate()
 
1055
        state.set_path_id('', file_id)
 
1056
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1057
            self._make_dirty(reset_inventory=True)
 
1058
 
 
1059
    @needs_read_lock
 
1060
    def supports_tree_reference(self):
 
1061
        return self._repo_supports_tree_reference
 
1062
 
 
1063
    def unlock(self):
 
1064
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1065
        if self._control_files._lock_count == 1:
 
1066
            # eventually we should do signature checking during read locks for
 
1067
            # dirstate updates.
 
1068
            if self._control_files._lock_mode == 'w':
 
1069
                if self._dirty:
 
1070
                    self.flush()
 
1071
            if self._dirstate is not None:
 
1072
                # This is a no-op if there are no modifications.
 
1073
                self._dirstate.save()
 
1074
                self._dirstate.unlock()
 
1075
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1076
            #       point. Instead, it could check if the header has been
 
1077
            #       modified when it is locked, and if not, it can hang on to
 
1078
            #       the data it has in memory.
 
1079
            self._dirstate = None
 
1080
            self._inventory = None
 
1081
        # reverse order of locking.
 
1082
        try:
 
1083
            return self._control_files.unlock()
 
1084
        finally:
 
1085
            self.branch.unlock()
 
1086
 
 
1087
    @needs_tree_write_lock
 
1088
    def unversion(self, file_ids):
 
1089
        """Remove the file ids in file_ids from the current versioned set.
 
1090
 
 
1091
        When a file_id is unversioned, all of its children are automatically
 
1092
        unversioned.
 
1093
 
 
1094
        :param file_ids: The file ids to stop versioning.
 
1095
        :raises: NoSuchId if any fileid is not currently versioned.
 
1096
        """
 
1097
        if not file_ids:
 
1098
            return
 
1099
        state = self.current_dirstate()
 
1100
        state._read_dirblocks_if_needed()
 
1101
        ids_to_unversion = set()
 
1102
        for file_id in file_ids:
 
1103
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1104
        paths_to_unversion = set()
 
1105
        # sketch:
 
1106
        # check if the root is to be unversioned, if so, assert for now.
 
1107
        # walk the state marking unversioned things as absent.
 
1108
        # if there are any un-unversioned ids at the end, raise
 
1109
        for key, details in state._dirblocks[0][1]:
 
1110
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1111
                key[2] in ids_to_unversion):
 
1112
                # I haven't written the code to unversion / yet - it should be
 
1113
                # supported.
 
1114
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1115
        block_index = 0
 
1116
        while block_index < len(state._dirblocks):
 
1117
            # process one directory at a time.
 
1118
            block = state._dirblocks[block_index]
 
1119
            # first check: is the path one to remove - it or its children
 
1120
            delete_block = False
 
1121
            for path in paths_to_unversion:
 
1122
                if (block[0].startswith(path) and
 
1123
                    (len(block[0]) == len(path) or
 
1124
                     block[0][len(path)] == '/')):
 
1125
                    # this entire block should be deleted - its the block for a
 
1126
                    # path to unversion; or the child of one
 
1127
                    delete_block = True
 
1128
                    break
 
1129
            # TODO: trim paths_to_unversion as we pass by paths
 
1130
            if delete_block:
 
1131
                # this block is to be deleted: process it.
 
1132
                # TODO: we can special case the no-parents case and
 
1133
                # just forget the whole block.
 
1134
                entry_index = 0
 
1135
                while entry_index < len(block[1]):
 
1136
                    # Mark this file id as having been removed
 
1137
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1138
                    if not state._make_absent(block[1][entry_index]):
 
1139
                        entry_index += 1
 
1140
                # go to the next block. (At the moment we dont delete empty
 
1141
                # dirblocks)
 
1142
                block_index += 1
 
1143
                continue
 
1144
            entry_index = 0
 
1145
            while entry_index < len(block[1]):
 
1146
                entry = block[1][entry_index]
 
1147
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1148
                    # ^ some parent row.
 
1149
                    entry[0][2] not in ids_to_unversion):
 
1150
                    # ^ not an id to unversion
 
1151
                    entry_index += 1
 
1152
                    continue
 
1153
                if entry[1][0][0] == 'd':
 
1154
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1155
                if not state._make_absent(entry):
 
1156
                    entry_index += 1
 
1157
                # we have unversioned this id
 
1158
                ids_to_unversion.remove(entry[0][2])
 
1159
            block_index += 1
 
1160
        if ids_to_unversion:
 
1161
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1162
        self._make_dirty(reset_inventory=False)
 
1163
        # have to change the legacy inventory too.
 
1164
        if self._inventory is not None:
 
1165
            for file_id in file_ids:
 
1166
                self._inventory.remove_recursive_id(file_id)
 
1167
 
 
1168
    @needs_tree_write_lock
 
1169
    def _write_inventory(self, inv):
 
1170
        """Write inventory as the current inventory."""
 
1171
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1172
        self.current_dirstate().set_state_from_inventory(inv)
 
1173
        self._make_dirty(reset_inventory=False)
 
1174
        if self._inventory is not None:
 
1175
            self._inventory = inv
 
1176
        self.flush()
 
1177
 
 
1178
 
 
1179
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1180
    """The first consolidated dirstate working tree format.
 
1181
 
 
1182
    This format:
 
1183
        - exists within a metadir controlling .bzr
 
1184
        - includes an explicit version marker for the workingtree control
 
1185
          files, separate from the BzrDir format
 
1186
        - modifies the hash cache format
 
1187
        - is new in bzr 0.15
 
1188
        - uses a LockDir to guard access to it.
 
1189
    """
 
1190
 
 
1191
    def get_format_string(self):
 
1192
        """See WorkingTreeFormat.get_format_string()."""
 
1193
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1194
 
 
1195
    def get_format_description(self):
 
1196
        """See WorkingTreeFormat.get_format_description()."""
 
1197
        return "Working tree format 4"
 
1198
 
 
1199
    def initialize(self, a_bzrdir, revision_id=None):
 
1200
        """See WorkingTreeFormat.initialize().
 
1201
 
 
1202
        :param revision_id: allows creating a working tree at a different
 
1203
        revision than the branch is at.
 
1204
 
 
1205
        These trees get an initial random root id.
 
1206
        """
 
1207
        revision_id = osutils.safe_revision_id(revision_id)
 
1208
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1209
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1210
        transport = a_bzrdir.get_workingtree_transport(self)
 
1211
        control_files = self._open_control_files(a_bzrdir)
 
1212
        control_files.create_lock()
 
1213
        control_files.lock_write()
 
1214
        control_files.put_utf8('format', self.get_format_string())
 
1215
        branch = a_bzrdir.open_branch()
 
1216
        if revision_id is None:
 
1217
            revision_id = branch.last_revision()
 
1218
        local_path = transport.local_abspath('dirstate')
 
1219
        # write out new dirstate (must exist when we create the tree)
 
1220
        state = dirstate.DirState.initialize(local_path)
 
1221
        state.unlock()
 
1222
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1223
                         branch,
 
1224
                         _format=self,
 
1225
                         _bzrdir=a_bzrdir,
 
1226
                         _control_files=control_files)
 
1227
        wt._new_tree()
 
1228
        wt.lock_tree_write()
 
1229
        state._validate()
 
1230
        try:
 
1231
            if revision_id in (None, NULL_REVISION):
 
1232
                wt._set_root_id(generate_ids.gen_root_id())
 
1233
                wt.flush()
 
1234
                wt.current_dirstate()._validate()
 
1235
            wt.set_last_revision(revision_id)
 
1236
            wt.flush()
 
1237
            basis = wt.basis_tree()
 
1238
            basis.lock_read()
 
1239
            # if the basis has a root id we have to use that; otherwise we use
 
1240
            # a new random one
 
1241
            basis_root_id = basis.get_root_id()
 
1242
            if basis_root_id is not None:
 
1243
                wt._set_root_id(basis_root_id)
 
1244
                wt.flush()
 
1245
            transform.build_tree(basis, wt)
 
1246
            basis.unlock()
 
1247
        finally:
 
1248
            control_files.unlock()
 
1249
            wt.unlock()
 
1250
        return wt
 
1251
 
 
1252
    def _open(self, a_bzrdir, control_files):
 
1253
        """Open the tree itself.
 
1254
 
 
1255
        :param a_bzrdir: the dir for the tree.
 
1256
        :param control_files: the control files for the tree.
 
1257
        """
 
1258
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1259
                           branch=a_bzrdir.open_branch(),
 
1260
                           _format=self,
 
1261
                           _bzrdir=a_bzrdir,
 
1262
                           _control_files=control_files)
 
1263
 
 
1264
    def __get_matchingbzrdir(self):
 
1265
        # please test against something that will let us do tree references
 
1266
        return bzrdir.format_registry.make_bzrdir(
 
1267
            'dirstate-with-subtree')
 
1268
 
 
1269
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1270
 
 
1271
 
 
1272
class DirStateRevisionTree(Tree):
 
1273
    """A revision tree pulling the inventory from a dirstate."""
 
1274
 
 
1275
    def __init__(self, dirstate, revision_id, repository):
 
1276
        self._dirstate = dirstate
 
1277
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1278
        self._repository = repository
 
1279
        self._inventory = None
 
1280
        self._locked = 0
 
1281
        self._dirstate_locked = False
 
1282
 
 
1283
    def __repr__(self):
 
1284
        return "<%s of %s in %s>" % \
 
1285
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1286
 
 
1287
    def annotate_iter(self, file_id):
 
1288
        """See Tree.annotate_iter"""
 
1289
        w = self._repository.weave_store.get_weave(file_id,
 
1290
                           self._repository.get_transaction())
 
1291
        return w.annotate_iter(self.inventory[file_id].revision)
 
1292
 
 
1293
    def _comparison_data(self, entry, path):
 
1294
        """See Tree._comparison_data."""
 
1295
        if entry is None:
 
1296
            return None, False, None
 
1297
        # trust the entry as RevisionTree does, but this may not be
 
1298
        # sensible: the entry might not have come from us?
 
1299
        return entry.kind, entry.executable, None
 
1300
 
 
1301
    def _file_size(self, entry, stat_value):
 
1302
        return entry.text_size
 
1303
 
 
1304
    def filter_unversioned_files(self, paths):
 
1305
        """Filter out paths that are not versioned.
 
1306
 
 
1307
        :return: set of paths.
 
1308
        """
 
1309
        pred = self.has_filename
 
1310
        return set((p for p in paths if not pred(p)))
 
1311
 
 
1312
    def get_root_id(self):
 
1313
        return self.path2id('')
 
1314
 
 
1315
    def _get_parent_index(self):
 
1316
        """Return the index in the dirstate referenced by this tree."""
 
1317
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1318
 
 
1319
    def _get_entry(self, file_id=None, path=None):
 
1320
        """Get the dirstate row for file_id or path.
 
1321
 
 
1322
        If either file_id or path is supplied, it is used as the key to lookup.
 
1323
        If both are supplied, the fastest lookup is used, and an error is
 
1324
        raised if they do not both point at the same row.
 
1325
        
 
1326
        :param file_id: An optional unicode file_id to be looked up.
 
1327
        :param path: An optional unicode path to be looked up.
 
1328
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1329
        """
 
1330
        if file_id is None and path is None:
 
1331
            raise errors.BzrError('must supply file_id or path')
 
1332
        file_id = osutils.safe_file_id(file_id)
 
1333
        if path is not None:
 
1334
            path = path.encode('utf8')
 
1335
        parent_index = self._get_parent_index()
 
1336
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1337
 
 
1338
    def _generate_inventory(self):
 
1339
        """Create and set self.inventory from the dirstate object.
 
1340
 
 
1341
        (So this is only called the first time the inventory is requested for
 
1342
        this tree; it then remains in memory until it's out of date.)
 
1343
 
 
1344
        This is relatively expensive: we have to walk the entire dirstate.
 
1345
        """
 
1346
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1347
            'dirstate revision tree'
 
1348
        # separate call for profiling - makes it clear where the costs are.
 
1349
        self._dirstate._read_dirblocks_if_needed()
 
1350
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1351
            'parent %s has disappeared from %s' % (
 
1352
            self._revision_id, self._dirstate.get_parent_ids())
 
1353
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1354
        # This is identical now to the WorkingTree _generate_inventory except
 
1355
        # for the tree index use.
 
1356
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1357
        current_id = root_key[2]
 
1358
        assert current_entry[parent_index][0] == 'd'
 
1359
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1360
        inv.root.revision = current_entry[parent_index][4]
 
1361
        # Turn some things into local variables
 
1362
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1363
        factory = entry_factory
 
1364
        utf8_decode = cache_utf8._utf8_decode
 
1365
        inv_byid = inv._byid
 
1366
        # we could do this straight out of the dirstate; it might be fast
 
1367
        # and should be profiled - RBC 20070216
 
1368
        parent_ies = {'' : inv.root}
 
1369
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1370
            dirname = block[0]
 
1371
            try:
 
1372
                parent_ie = parent_ies[dirname]
 
1373
            except KeyError:
 
1374
                # all the paths in this block are not versioned in this tree
 
1375
                continue
 
1376
            for key, entry in block[1]:
 
1377
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1378
                if minikind in ('a', 'r'): # absent, relocated
 
1379
                    # not this tree
 
1380
                    continue
 
1381
                name = key[1]
 
1382
                name_unicode = utf8_decode(name)[0]
 
1383
                file_id = key[2]
 
1384
                kind = minikind_to_kind[minikind]
 
1385
                inv_entry = factory[kind](file_id, name_unicode,
 
1386
                                          parent_ie.file_id)
 
1387
                inv_entry.revision = revid
 
1388
                if kind == 'file':
 
1389
                    inv_entry.executable = executable
 
1390
                    inv_entry.text_size = size
 
1391
                    inv_entry.text_sha1 = fingerprint
 
1392
                elif kind == 'directory':
 
1393
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1394
                elif kind == 'symlink':
 
1395
                    inv_entry.executable = False
 
1396
                    inv_entry.text_size = size
 
1397
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1398
                elif kind == 'tree-reference':
 
1399
                    inv_entry.reference_revision = fingerprint or None
 
1400
                else:
 
1401
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1402
                            % entry)
 
1403
                # These checks cost us around 40ms on a 55k entry tree
 
1404
                assert file_id not in inv_byid
 
1405
                assert name_unicode not in parent_ie.children
 
1406
                inv_byid[file_id] = inv_entry
 
1407
                parent_ie.children[name_unicode] = inv_entry
 
1408
        self._inventory = inv
 
1409
 
 
1410
    def get_file_mtime(self, file_id, path=None):
 
1411
        """Return the modification time for this record.
 
1412
 
 
1413
        We return the timestamp of the last-changed revision.
 
1414
        """
 
1415
        # Make sure the file exists
 
1416
        entry = self._get_entry(file_id, path=path)
 
1417
        if entry == (None, None): # do we raise?
 
1418
            return None
 
1419
        parent_index = self._get_parent_index()
 
1420
        last_changed_revision = entry[1][parent_index][4]
 
1421
        return self._repository.get_revision(last_changed_revision).timestamp
 
1422
 
 
1423
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1424
        entry = self._get_entry(file_id=file_id, path=path)
 
1425
        parent_index = self._get_parent_index()
 
1426
        parent_details = entry[1][parent_index]
 
1427
        if parent_details[0] == 'f':
 
1428
            return parent_details[1]
 
1429
        return None
 
1430
 
 
1431
    def get_file(self, file_id):
 
1432
        return StringIO(self.get_file_text(file_id))
 
1433
 
 
1434
    def get_file_lines(self, file_id):
 
1435
        ie = self.inventory[file_id]
 
1436
        return self._repository.weave_store.get_weave(file_id,
 
1437
                self._repository.get_transaction()).get_lines(ie.revision)
 
1438
 
 
1439
    def get_file_size(self, file_id):
 
1440
        return self.inventory[file_id].text_size
 
1441
 
 
1442
    def get_file_text(self, file_id):
 
1443
        return ''.join(self.get_file_lines(file_id))
 
1444
 
 
1445
    def get_reference_revision(self, file_id, path=None):
 
1446
        return self.inventory[file_id].reference_revision
 
1447
 
 
1448
    def get_symlink_target(self, file_id):
 
1449
        entry = self._get_entry(file_id=file_id)
 
1450
        parent_index = self._get_parent_index()
 
1451
        if entry[1][parent_index][0] != 'l':
 
1452
            return None
 
1453
        else:
 
1454
            # At present, none of the tree implementations supports non-ascii
 
1455
            # symlink targets. So we will just assume that the dirstate path is
 
1456
            # correct.
 
1457
            return entry[1][parent_index][1]
 
1458
 
 
1459
    def get_revision_id(self):
 
1460
        """Return the revision id for this tree."""
 
1461
        return self._revision_id
 
1462
 
 
1463
    def _get_inventory(self):
 
1464
        if self._inventory is not None:
 
1465
            return self._inventory
 
1466
        self._must_be_locked()
 
1467
        self._generate_inventory()
 
1468
        return self._inventory
 
1469
 
 
1470
    inventory = property(_get_inventory,
 
1471
                         doc="Inventory of this Tree")
 
1472
 
 
1473
    def get_parent_ids(self):
 
1474
        """The parents of a tree in the dirstate are not cached."""
 
1475
        return self._repository.get_revision(self._revision_id).parent_ids
 
1476
 
 
1477
    def has_filename(self, filename):
 
1478
        return bool(self.path2id(filename))
 
1479
 
 
1480
    def kind(self, file_id):
 
1481
        return self.inventory[file_id].kind
 
1482
 
 
1483
    def is_executable(self, file_id, path=None):
 
1484
        ie = self.inventory[file_id]
 
1485
        if ie.kind != "file":
 
1486
            return None
 
1487
        return ie.executable
 
1488
 
 
1489
    def list_files(self, include_root=False):
 
1490
        # We use a standard implementation, because DirStateRevisionTree is
 
1491
        # dealing with one of the parents of the current state
 
1492
        inv = self._get_inventory()
 
1493
        entries = inv.iter_entries()
 
1494
        if self.inventory.root is not None and not include_root:
 
1495
            entries.next()
 
1496
        for path, entry in entries:
 
1497
            yield path, 'V', entry.kind, entry.file_id, entry
 
1498
 
 
1499
    def lock_read(self):
 
1500
        """Lock the tree for a set of operations."""
 
1501
        if not self._locked:
 
1502
            self._repository.lock_read()
 
1503
            if self._dirstate._lock_token is None:
 
1504
                self._dirstate.lock_read()
 
1505
                self._dirstate_locked = True
 
1506
        self._locked += 1
 
1507
 
 
1508
    def _must_be_locked(self):
 
1509
        if not self._locked:
 
1510
            raise errors.ObjectNotLocked(self)
 
1511
 
 
1512
    @needs_read_lock
 
1513
    def path2id(self, path):
 
1514
        """Return the id for path in this tree."""
 
1515
        # lookup by path: faster than splitting and walking the ivnentory.
 
1516
        entry = self._get_entry(path=path)
 
1517
        if entry == (None, None):
 
1518
            return None
 
1519
        return entry[0][2]
 
1520
 
 
1521
    def unlock(self):
 
1522
        """Unlock, freeing any cache memory used during the lock."""
 
1523
        # outside of a lock, the inventory is suspect: release it.
 
1524
        self._locked -=1
 
1525
        if not self._locked:
 
1526
            self._inventory = None
 
1527
            self._locked = 0
 
1528
            if self._dirstate_locked:
 
1529
                self._dirstate.unlock()
 
1530
                self._dirstate_locked = False
 
1531
            self._repository.unlock()
 
1532
 
 
1533
    def walkdirs(self, prefix=""):
 
1534
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1535
        # implementation based on an inventory.  
 
1536
        # This should be cleaned up to use the much faster Dirstate code
 
1537
        # So for now, we just build up the parent inventory, and extract
 
1538
        # it the same way RevisionTree does.
 
1539
        _directory = 'directory'
 
1540
        inv = self._get_inventory()
 
1541
        top_id = inv.path2id(prefix)
 
1542
        if top_id is None:
 
1543
            pending = []
 
1544
        else:
 
1545
            pending = [(prefix, top_id)]
 
1546
        while pending:
 
1547
            dirblock = []
 
1548
            relpath, file_id = pending.pop()
 
1549
            # 0 - relpath, 1- file-id
 
1550
            if relpath:
 
1551
                relroot = relpath + '/'
 
1552
            else:
 
1553
                relroot = ""
 
1554
            # FIXME: stash the node in pending
 
1555
            entry = inv[file_id]
 
1556
            for name, child in entry.sorted_children():
 
1557
                toppath = relroot + name
 
1558
                dirblock.append((toppath, name, child.kind, None,
 
1559
                    child.file_id, child.kind
 
1560
                    ))
 
1561
            yield (relpath, entry.file_id), dirblock
 
1562
            # push the user specified dirs from dirblock
 
1563
            for dir in reversed(dirblock):
 
1564
                if dir[2] == _directory:
 
1565
                    pending.append((dir[0], dir[4]))
 
1566
 
 
1567
 
 
1568
class InterDirStateTree(InterTree):
 
1569
    """Fast path optimiser for changes_from with dirstate trees.
 
1570
    
 
1571
    This is used only when both trees are in the dirstate working file, and 
 
1572
    the source is any parent within the dirstate, and the destination is 
 
1573
    the current working tree of the same dirstate.
 
1574
    """
 
1575
    # this could be generalized to allow comparisons between any trees in the
 
1576
    # dirstate, and possibly between trees stored in different dirstates.
 
1577
 
 
1578
    def __init__(self, source, target):
 
1579
        super(InterDirStateTree, self).__init__(source, target)
 
1580
        if not InterDirStateTree.is_compatible(source, target):
 
1581
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1582
 
 
1583
    @staticmethod
 
1584
    def make_source_parent_tree(source, target):
 
1585
        """Change the source tree into a parent of the target."""
 
1586
        revid = source.commit('record tree')
 
1587
        target.branch.repository.fetch(source.branch.repository, revid)
 
1588
        target.set_parent_ids([revid])
 
1589
        return target.basis_tree(), target
 
1590
 
 
1591
    _matching_from_tree_format = WorkingTreeFormat4()
 
1592
    _matching_to_tree_format = WorkingTreeFormat4()
 
1593
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1594
 
 
1595
    def _iter_changes(self, include_unchanged=False,
 
1596
                      specific_files=None, pb=None, extra_trees=[],
 
1597
                      require_versioned=True, want_unversioned=False):
 
1598
        """Return the changes from source to target.
 
1599
 
 
1600
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1601
            for details.
 
1602
        :param specific_files: An optional list of file paths to restrict the
 
1603
            comparison to. When mapping filenames to ids, all matches in all
 
1604
            trees (including optional extra_trees) are used, and all children of
 
1605
            matched directories are included.
 
1606
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1607
            unchanged entries in the result.
 
1608
        :param extra_trees: An optional list of additional trees to use when
 
1609
            mapping the contents of specific_files (paths) to file_ids.
 
1610
        :param require_versioned: If True, all files in specific_files must be
 
1611
            versioned in one of source, target, extra_trees or
 
1612
            PathsNotVersionedError is raised.
 
1613
        :param want_unversioned: Should unversioned files be returned in the
 
1614
            output. An unversioned file is defined as one with (False, False)
 
1615
            for the versioned pair.
 
1616
        """
 
1617
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1618
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1619
        # NB: show_status depends on being able to pass in non-versioned files
 
1620
        # and report them as unknown
 
1621
        # TODO: handle extra trees in the dirstate.
 
1622
        # TODO: handle comparisons as an empty tree as a different special
 
1623
        # case? mbp 20070226
 
1624
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1625
            # we can't fast-path these cases (yet)
 
1626
            for f in super(InterDirStateTree, self)._iter_changes(
 
1627
                include_unchanged, specific_files, pb, extra_trees,
 
1628
                require_versioned, want_unversioned=want_unversioned):
 
1629
                yield f
 
1630
            return
 
1631
        parent_ids = self.target.get_parent_ids()
 
1632
        assert (self.source._revision_id in parent_ids), \
 
1633
                "revision {%s} is not stored in {%s}, but %s " \
 
1634
                "can only be used for trees stored in the dirstate" \
 
1635
                % (self.source._revision_id, self.target, self._iter_changes)
 
1636
        target_index = 0
 
1637
        if self.source._revision_id == NULL_REVISION:
 
1638
            source_index = None
 
1639
            indices = (target_index,)
 
1640
        else:
 
1641
            assert (self.source._revision_id in parent_ids), \
 
1642
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1643
                self.source._revision_id, parent_ids)
 
1644
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1645
            indices = (source_index,target_index)
 
1646
        # -- make all specific_files utf8 --
 
1647
        if specific_files:
 
1648
            specific_files_utf8 = set()
 
1649
            for path in specific_files:
 
1650
                specific_files_utf8.add(path.encode('utf8'))
 
1651
            specific_files = specific_files_utf8
 
1652
        else:
 
1653
            specific_files = set([''])
 
1654
        # -- specific_files is now a utf8 path set --
 
1655
        # -- get the state object and prepare it.
 
1656
        state = self.target.current_dirstate()
 
1657
        state._read_dirblocks_if_needed()
 
1658
        def _entries_for_path(path):
 
1659
            """Return a list with all the entries that match path for all ids.
 
1660
            """
 
1661
            dirname, basename = os.path.split(path)
 
1662
            key = (dirname, basename, '')
 
1663
            block_index, present = state._find_block_index_from_key(key)
 
1664
            if not present:
 
1665
                # the block which should contain path is absent.
 
1666
                return []
 
1667
            result = []
 
1668
            block = state._dirblocks[block_index][1]
 
1669
            entry_index, _ = state._find_entry_index(key, block)
 
1670
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1671
            while (entry_index < len(block) and
 
1672
                block[entry_index][0][0:2] == key[0:2]):
 
1673
                result.append(block[entry_index])
 
1674
                entry_index += 1
 
1675
            return result
 
1676
        if require_versioned:
 
1677
            # -- check all supplied paths are versioned in a search tree. --
 
1678
            all_versioned = True
 
1679
            for path in specific_files:
 
1680
                path_entries = _entries_for_path(path)
 
1681
                if not path_entries:
 
1682
                    # this specified path is not present at all: error
 
1683
                    all_versioned = False
 
1684
                    break
 
1685
                found_versioned = False
 
1686
                # for each id at this path
 
1687
                for entry in path_entries:
 
1688
                    # for each tree.
 
1689
                    for index in indices:
 
1690
                        if entry[1][index][0] != 'a': # absent
 
1691
                            found_versioned = True
 
1692
                            # all good: found a versioned cell
 
1693
                            break
 
1694
                if not found_versioned:
 
1695
                    # none of the indexes was not 'absent' at all ids for this
 
1696
                    # path.
 
1697
                    all_versioned = False
 
1698
                    break
 
1699
            if not all_versioned:
 
1700
                raise errors.PathsNotVersionedError(specific_files)
 
1701
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1702
        search_specific_files = set()
 
1703
        for path in specific_files:
 
1704
            other_specific_files = specific_files.difference(set([path]))
 
1705
            if not osutils.is_inside_any(other_specific_files, path):
 
1706
                # this is a top level path, we must check it.
 
1707
                search_specific_files.add(path)
 
1708
        # sketch: 
 
1709
        # compare source_index and target_index at or under each element of search_specific_files.
 
1710
        # follow the following comparison table. Note that we only want to do diff operations when
 
1711
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1712
        # for the target.
 
1713
        # cases:
 
1714
        # 
 
1715
        # Source | Target | disk | action
 
1716
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1717
        #        |        |      | diff check on source-target
 
1718
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1719
        #        |        |      | ???
 
1720
        #   r    |  a     |      | add source to search
 
1721
        #   r    |  a     |  a   | 
 
1722
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1723
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1724
        #   a    | fdlt   |      | add new id
 
1725
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1726
        #   a    |  a     |      | not present in either tree, skip
 
1727
        #   a    |  a     |  a   | not present in any tree, skip
 
1728
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1729
        #        |        |      | may not be selected by the users list of paths.
 
1730
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1731
        #        |        |      | may not be selected by the users list of paths.
 
1732
        #  fdlt  | fdlt   |      | content in both: diff them
 
1733
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1734
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1735
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1736
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1737
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1738
        #        |        |      | this id at the other path.
 
1739
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1740
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1741
        #        |        |      | this id at the other path.
 
1742
 
 
1743
        # for all search_indexs in each path at or under each element of
 
1744
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1745
        # relocated path as one to search if its not searched already. If the
 
1746
        # detail is not relocated, add the id.
 
1747
        searched_specific_files = set()
 
1748
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1749
        # Using a list so that we can access the values and change them in
 
1750
        # nested scope. Each one is [path, file_id, entry]
 
1751
        last_source_parent = [None, None, None]
 
1752
        last_target_parent = [None, None, None]
 
1753
 
 
1754
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1755
 
 
1756
        def _process_entry(entry, path_info):
 
1757
            """Compare an entry and real disk to generate delta information.
 
1758
 
 
1759
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1760
                the path of entry. If None, then the path is considered absent.
 
1761
                (Perhaps we should pass in a concrete entry for this ?)
 
1762
                Basename is returned as a utf8 string because we expect this
 
1763
                tuple will be ignored, and don't want to take the time to
 
1764
                decode.
 
1765
            """
 
1766
            if source_index is None:
 
1767
                source_details = NULL_PARENT_DETAILS
 
1768
            else:
 
1769
                source_details = entry[1][source_index]
 
1770
            target_details = entry[1][target_index]
 
1771
            target_minikind = target_details[0]
 
1772
            if path_info is not None and target_minikind in 'fdlt':
 
1773
                assert target_index == 0
 
1774
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1775
                                                  stat_value=path_info[3])
 
1776
                # The entry may have been modified by update_entry
 
1777
                target_details = entry[1][target_index]
 
1778
                target_minikind = target_details[0]
 
1779
            else:
 
1780
                link_or_sha1 = None
 
1781
            source_minikind = source_details[0]
 
1782
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1783
                # claimed content in both: diff
 
1784
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1785
                #        |        |      | diff check on source-target
 
1786
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1787
                #        |        |      | ???
 
1788
                if source_minikind in 'r':
 
1789
                    # add the source to the search path to find any children it
 
1790
                    # has.  TODO ? : only add if it is a container ?
 
1791
                    if not osutils.is_inside_any(searched_specific_files,
 
1792
                                                 source_details[1]):
 
1793
                        search_specific_files.add(source_details[1])
 
1794
                    # generate the old path; this is needed for stating later
 
1795
                    # as well.
 
1796
                    old_path = source_details[1]
 
1797
                    old_dirname, old_basename = os.path.split(old_path)
 
1798
                    path = pathjoin(entry[0][0], entry[0][1])
 
1799
                    old_entry = state._get_entry(source_index,
 
1800
                                                 path_utf8=old_path)
 
1801
                    # update the source details variable to be the real
 
1802
                    # location.
 
1803
                    source_details = old_entry[1][source_index]
 
1804
                    source_minikind = source_details[0]
 
1805
                else:
 
1806
                    old_dirname = entry[0][0]
 
1807
                    old_basename = entry[0][1]
 
1808
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1809
                if path_info is None:
 
1810
                    # the file is missing on disk, show as removed.
 
1811
                    content_change = True
 
1812
                    target_kind = None
 
1813
                    target_exec = False
 
1814
                else:
 
1815
                    # source and target are both versioned and disk file is present.
 
1816
                    target_kind = path_info[2]
 
1817
                    if target_kind == 'directory':
 
1818
                        if source_minikind != 'd':
 
1819
                            content_change = True
 
1820
                        else:
 
1821
                            # directories have no fingerprint
 
1822
                            content_change = False
 
1823
                        target_exec = False
 
1824
                    elif target_kind == 'file':
 
1825
                        if source_minikind != 'f':
 
1826
                            content_change = True
 
1827
                        else:
 
1828
                            # We could check the size, but we already have the
 
1829
                            # sha1 hash.
 
1830
                            content_change = (link_or_sha1 != source_details[1])
 
1831
                        # Target details is updated at update_entry time
 
1832
                        if use_filesystem_for_exec:
 
1833
                            # We don't need S_ISREG here, because we are sure
 
1834
                            # we are dealing with a file.
 
1835
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1836
                        else:
 
1837
                            target_exec = target_details[3]
 
1838
                    elif target_kind == 'symlink':
 
1839
                        if source_minikind != 'l':
 
1840
                            content_change = True
 
1841
                        else:
 
1842
                            content_change = (link_or_sha1 != source_details[1])
 
1843
                        target_exec = False
 
1844
                    elif target_kind == 'tree-reference':
 
1845
                        if source_minikind != 't':
 
1846
                            content_change = True
 
1847
                        else:
 
1848
                            content_change = False
 
1849
                        target_exec = False
 
1850
                    else:
 
1851
                        raise Exception, "unknown kind %s" % path_info[2]
 
1852
                # parent id is the entry for the path in the target tree
 
1853
                if old_dirname == last_source_parent[0]:
 
1854
                    source_parent_id = last_source_parent[1]
 
1855
                else:
 
1856
                    source_parent_entry = state._get_entry(source_index,
 
1857
                                                           path_utf8=old_dirname)
 
1858
                    source_parent_id = source_parent_entry[0][2]
 
1859
                    if source_parent_id == entry[0][2]:
 
1860
                        # This is the root, so the parent is None
 
1861
                        source_parent_id = None
 
1862
                    else:
 
1863
                        last_source_parent[0] = old_dirname
 
1864
                        last_source_parent[1] = source_parent_id
 
1865
                        last_source_parent[2] = source_parent_entry
 
1866
 
 
1867
                new_dirname = entry[0][0]
 
1868
                if new_dirname == last_target_parent[0]:
 
1869
                    target_parent_id = last_target_parent[1]
 
1870
                else:
 
1871
                    # TODO: We don't always need to do the lookup, because the
 
1872
                    #       parent entry will be the same as the source entry.
 
1873
                    target_parent_entry = state._get_entry(target_index,
 
1874
                                                           path_utf8=new_dirname)
 
1875
                    target_parent_id = target_parent_entry[0][2]
 
1876
                    if target_parent_id == entry[0][2]:
 
1877
                        # This is the root, so the parent is None
 
1878
                        target_parent_id = None
 
1879
                    else:
 
1880
                        last_target_parent[0] = new_dirname
 
1881
                        last_target_parent[1] = target_parent_id
 
1882
                        last_target_parent[2] = target_parent_entry
 
1883
 
 
1884
                source_exec = source_details[3]
 
1885
                return ((entry[0][2], (old_path, path), content_change,
 
1886
                        (True, True),
 
1887
                        (source_parent_id, target_parent_id),
 
1888
                        (old_basename, entry[0][1]),
 
1889
                        (_minikind_to_kind[source_minikind], target_kind),
 
1890
                        (source_exec, target_exec)),)
 
1891
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1892
                # looks like a new file
 
1893
                if path_info is not None:
 
1894
                    path = pathjoin(entry[0][0], entry[0][1])
 
1895
                    # parent id is the entry for the path in the target tree
 
1896
                    # TODO: these are the same for an entire directory: cache em.
 
1897
                    parent_id = state._get_entry(target_index,
 
1898
                                                 path_utf8=entry[0][0])[0][2]
 
1899
                    if parent_id == entry[0][2]:
 
1900
                        parent_id = None
 
1901
                    if use_filesystem_for_exec:
 
1902
                        # We need S_ISREG here, because we aren't sure if this
 
1903
                        # is a file or not.
 
1904
                        target_exec = bool(
 
1905
                            stat.S_ISREG(path_info[3].st_mode)
 
1906
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1907
                    else:
 
1908
                        target_exec = target_details[3]
 
1909
                    return ((entry[0][2], (None, path), True,
 
1910
                            (False, True),
 
1911
                            (None, parent_id),
 
1912
                            (None, entry[0][1]),
 
1913
                            (None, path_info[2]),
 
1914
                            (None, target_exec)),)
 
1915
                else:
 
1916
                    # but its not on disk: we deliberately treat this as just
 
1917
                    # never-present. (Why ?! - RBC 20070224)
 
1918
                    pass
 
1919
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1920
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1921
                # if its still on disk, *and* theres no other entry at this
 
1922
                # path [we dont know this in this routine at the moment -
 
1923
                # perhaps we should change this - then it would be an unknown.
 
1924
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1925
                # parent id is the entry for the path in the target tree
 
1926
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1927
                if parent_id == entry[0][2]:
 
1928
                    parent_id = None
 
1929
                return ((entry[0][2], (old_path, None), True,
 
1930
                        (True, False),
 
1931
                        (parent_id, None),
 
1932
                        (entry[0][1], None),
 
1933
                        (_minikind_to_kind[source_minikind], None),
 
1934
                        (source_details[3], None)),)
 
1935
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1936
                # a rename; could be a true rename, or a rename inherited from
 
1937
                # a renamed parent. TODO: handle this efficiently. Its not
 
1938
                # common case to rename dirs though, so a correct but slow
 
1939
                # implementation will do.
 
1940
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1941
                    search_specific_files.add(target_details[1])
 
1942
            elif source_minikind in 'r' and target_minikind in 'r':
 
1943
                # neither of the selected trees contain this file,
 
1944
                # so skip over it. This is not currently directly tested, but
 
1945
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1946
                pass
 
1947
            else:
 
1948
                raise AssertionError("don't know how to compare "
 
1949
                    "source_minikind=%r, target_minikind=%r"
 
1950
                    % (source_minikind, target_minikind))
 
1951
                ## import pdb;pdb.set_trace()
 
1952
            return ()
 
1953
        while search_specific_files:
 
1954
            # TODO: the pending list should be lexically sorted?  the
 
1955
            # interface doesn't require it.
 
1956
            current_root = search_specific_files.pop()
 
1957
            searched_specific_files.add(current_root)
 
1958
            # process the entries for this containing directory: the rest will be
 
1959
            # found by their parents recursively.
 
1960
            root_entries = _entries_for_path(current_root)
 
1961
            root_abspath = self.target.abspath(current_root)
 
1962
            try:
 
1963
                root_stat = os.lstat(root_abspath)
 
1964
            except OSError, e:
 
1965
                if e.errno == errno.ENOENT:
 
1966
                    # the path does not exist: let _process_entry know that.
 
1967
                    root_dir_info = None
 
1968
                else:
 
1969
                    # some other random error: hand it up.
 
1970
                    raise
 
1971
            else:
 
1972
                root_dir_info = ('', current_root,
 
1973
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1974
                    root_abspath)
 
1975
                if root_dir_info[2] == 'directory':
 
1976
                    if self.target._directory_is_tree_reference(
 
1977
                        current_root.decode('utf8')):
 
1978
                        root_dir_info = root_dir_info[:2] + \
 
1979
                            ('tree-reference',) + root_dir_info[3:]
 
1980
 
 
1981
            if not root_entries and not root_dir_info:
 
1982
                # this specified path is not present at all, skip it.
 
1983
                continue
 
1984
            path_handled = False
 
1985
            for entry in root_entries:
 
1986
                for result in _process_entry(entry, root_dir_info):
 
1987
                    # this check should probably be outside the loop: one
 
1988
                    # 'iterate two trees' api, and then _iter_changes filters
 
1989
                    # unchanged pairs. - RBC 20070226
 
1990
                    path_handled = True
 
1991
                    if (include_unchanged
 
1992
                        or result[2]                    # content change
 
1993
                        or result[3][0] != result[3][1] # versioned status
 
1994
                        or result[4][0] != result[4][1] # parent id
 
1995
                        or result[5][0] != result[5][1] # name
 
1996
                        or result[6][0] != result[6][1] # kind
 
1997
                        or result[7][0] != result[7][1] # executable
 
1998
                        ):
 
1999
                        result = (result[0],
 
2000
                            ((utf8_decode(result[1][0])[0]),
 
2001
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
2002
                        yield result
 
2003
            if want_unversioned and not path_handled:
 
2004
                new_executable = bool(
 
2005
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2006
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2007
                yield (None, (None, current_root), True, (False, False),
 
2008
                    (None, None),
 
2009
                    (None, splitpath(current_root)[-1]),
 
2010
                    (None, root_dir_info[2]), (None, new_executable))
 
2011
            dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2012
            initial_key = (current_root, '', '')
 
2013
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2014
            if block_index == 0:
 
2015
                # we have processed the total root already, but because the
 
2016
                # initial key matched it we should skip it here.
 
2017
                block_index +=1
 
2018
            try:
 
2019
                current_dir_info = dir_iterator.next()
 
2020
            except OSError, e:
 
2021
                if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
2022
                    # there may be directories in the inventory even though
 
2023
                    # this path is not a file on disk: so mark it as end of
 
2024
                    # iterator
 
2025
                    current_dir_info = None
 
2026
                else:
 
2027
                    raise
 
2028
            else:
 
2029
                if current_dir_info[0][0] == '':
 
2030
                    # remove .bzr from iteration
 
2031
                    bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2032
                    assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2033
                    del current_dir_info[1][bzr_index]
 
2034
            # walk until both the directory listing and the versioned metadata
 
2035
            # are exhausted. 
 
2036
            if (block_index < len(state._dirblocks) and
 
2037
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2038
                current_block = state._dirblocks[block_index]
 
2039
            else:
 
2040
                current_block = None
 
2041
            while (current_dir_info is not None or
 
2042
                   current_block is not None):
 
2043
                if (current_dir_info and current_block
 
2044
                    and current_dir_info[0][0] != current_block[0]):
 
2045
                    if current_dir_info[0][0] < current_block[0] :
 
2046
                        # filesystem data refers to paths not covered by the dirblock.
 
2047
                        # this has two possibilities:
 
2048
                        # A) it is versioned but empty, so there is no block for it
 
2049
                        # B) it is not versioned.
 
2050
                        # in either case it was processed by the containing directories walk:
 
2051
                        # if it is root/foo, when we walked root we emitted it,
 
2052
                        # or if we ere given root/foo to walk specifically, we
 
2053
                        # emitted it when checking the walk-root entries
 
2054
                        # advance the iterator and loop - we dont need to emit it.
 
2055
                        try:
 
2056
                            current_dir_info = dir_iterator.next()
 
2057
                        except StopIteration:
 
2058
                            current_dir_info = None
 
2059
                    else:
 
2060
                        # We have a dirblock entry for this location, but there
 
2061
                        # is no filesystem path for this. This is most likely
 
2062
                        # because a directory was removed from the disk.
 
2063
                        # We don't have to report the missing directory,
 
2064
                        # because that should have already been handled, but we
 
2065
                        # need to handle all of the files that are contained
 
2066
                        # within.
 
2067
                        for current_entry in current_block[1]:
 
2068
                            # entry referring to file not present on disk.
 
2069
                            # advance the entry only, after processing.
 
2070
                            for result in _process_entry(current_entry, None):
 
2071
                                # this check should probably be outside the loop: one
 
2072
                                # 'iterate two trees' api, and then _iter_changes filters
 
2073
                                # unchanged pairs. - RBC 20070226
 
2074
                                if (include_unchanged
 
2075
                                    or result[2]                    # content change
 
2076
                                    or result[3][0] != result[3][1] # versioned status
 
2077
                                    or result[4][0] != result[4][1] # parent id
 
2078
                                    or result[5][0] != result[5][1] # name
 
2079
                                    or result[6][0] != result[6][1] # kind
 
2080
                                    or result[7][0] != result[7][1] # executable
 
2081
                                    ):
 
2082
                                    result = (result[0],
 
2083
                                        ((utf8_decode(result[1][0])[0]),
 
2084
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2085
                                    yield result
 
2086
                        block_index +=1
 
2087
                        if (block_index < len(state._dirblocks) and
 
2088
                            osutils.is_inside(current_root,
 
2089
                                              state._dirblocks[block_index][0])):
 
2090
                            current_block = state._dirblocks[block_index]
 
2091
                        else:
 
2092
                            current_block = None
 
2093
                    continue
 
2094
                entry_index = 0
 
2095
                if current_block and entry_index < len(current_block[1]):
 
2096
                    current_entry = current_block[1][entry_index]
 
2097
                else:
 
2098
                    current_entry = None
 
2099
                advance_entry = True
 
2100
                path_index = 0
 
2101
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2102
                    current_path_info = current_dir_info[1][path_index]
 
2103
                    if current_path_info[2] == 'directory':
 
2104
                        if self.target._directory_is_tree_reference(
 
2105
                            current_path_info[0].decode('utf8')):
 
2106
                            current_path_info = current_path_info[:2] + \
 
2107
                                ('tree-reference',) + current_path_info[3:]
 
2108
                else:
 
2109
                    current_path_info = None
 
2110
                advance_path = True
 
2111
                path_handled = False
 
2112
                while (current_entry is not None or
 
2113
                    current_path_info is not None):
 
2114
                    if current_entry is None:
 
2115
                        # the check for path_handled when the path is adnvaced
 
2116
                        # will yield this path if needed.
 
2117
                        pass
 
2118
                    elif current_path_info is None:
 
2119
                        # no path is fine: the per entry code will handle it.
 
2120
                        for result in _process_entry(current_entry, current_path_info):
 
2121
                            # this check should probably be outside the loop: one
 
2122
                            # 'iterate two trees' api, and then _iter_changes filters
 
2123
                            # unchanged pairs. - RBC 20070226
 
2124
                            if (include_unchanged
 
2125
                                or result[2]                    # content change
 
2126
                                or result[3][0] != result[3][1] # versioned status
 
2127
                                or result[4][0] != result[4][1] # parent id
 
2128
                                or result[5][0] != result[5][1] # name
 
2129
                                or result[6][0] != result[6][1] # kind
 
2130
                                or result[7][0] != result[7][1] # executable
 
2131
                                ):
 
2132
                                result = (result[0],
 
2133
                                    ((utf8_decode(result[1][0])[0]),
 
2134
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2135
                                yield result
 
2136
                    elif current_entry[0][1] != current_path_info[1]:
 
2137
                        if current_path_info[1] < current_entry[0][1]:
 
2138
                            # extra file on disk: pass for now, but only
 
2139
                            # increment the path, not the entry
 
2140
                            advance_entry = False
 
2141
                        else:
 
2142
                            # entry referring to file not present on disk.
 
2143
                            # advance the entry only, after processing.
 
2144
                            for result in _process_entry(current_entry, None):
 
2145
                                # this check should probably be outside the loop: one
 
2146
                                # 'iterate two trees' api, and then _iter_changes filters
 
2147
                                # unchanged pairs. - RBC 20070226
 
2148
                                path_handled = True
 
2149
                                if (include_unchanged
 
2150
                                    or result[2]                    # content change
 
2151
                                    or result[3][0] != result[3][1] # versioned status
 
2152
                                    or result[4][0] != result[4][1] # parent id
 
2153
                                    or result[5][0] != result[5][1] # name
 
2154
                                    or result[6][0] != result[6][1] # kind
 
2155
                                    or result[7][0] != result[7][1] # executable
 
2156
                                    ):
 
2157
                                    result = (result[0],
 
2158
                                        ((utf8_decode(result[1][0])[0]),
 
2159
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2160
                                    yield result
 
2161
                            advance_path = False
 
2162
                    else:
 
2163
                        for result in _process_entry(current_entry, current_path_info):
 
2164
                            # this check should probably be outside the loop: one
 
2165
                            # 'iterate two trees' api, and then _iter_changes filters
 
2166
                            # unchanged pairs. - RBC 20070226
 
2167
                            path_handled = True
 
2168
                            if (include_unchanged
 
2169
                                or result[2]                    # content change
 
2170
                                or result[3][0] != result[3][1] # versioned status
 
2171
                                or result[4][0] != result[4][1] # parent id
 
2172
                                or result[5][0] != result[5][1] # name
 
2173
                                or result[6][0] != result[6][1] # kind
 
2174
                                or result[7][0] != result[7][1] # executable
 
2175
                                ):
 
2176
                                result = (result[0],
 
2177
                                    ((utf8_decode(result[1][0])[0]),
 
2178
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2179
                                yield result
 
2180
                    if advance_entry and current_entry is not None:
 
2181
                        entry_index += 1
 
2182
                        if entry_index < len(current_block[1]):
 
2183
                            current_entry = current_block[1][entry_index]
 
2184
                        else:
 
2185
                            current_entry = None
 
2186
                    else:
 
2187
                        advance_entry = True # reset the advance flaga
 
2188
                    if advance_path and current_path_info is not None:
 
2189
                        if not path_handled:
 
2190
                            # unversioned in all regards
 
2191
                            if want_unversioned:
 
2192
                                new_executable = bool(
 
2193
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2194
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2195
                                if want_unversioned:
 
2196
                                    yield (None, (None, current_path_info[0]),
 
2197
                                        True,
 
2198
                                        (False, False),
 
2199
                                        (None, None),
 
2200
                                        (None, current_path_info[1]),
 
2201
                                        (None, current_path_info[2]),
 
2202
                                        (None, new_executable))
 
2203
                            # dont descend into this unversioned path if it is
 
2204
                            # a dir
 
2205
                            if current_path_info[2] in (
 
2206
                                'directory', 'tree-referene'):
 
2207
                                del current_dir_info[1][path_index]
 
2208
                                path_index -= 1
 
2209
                        path_index += 1
 
2210
                        if path_index < len(current_dir_info[1]):
 
2211
                            current_path_info = current_dir_info[1][path_index]
 
2212
                            if current_path_info[2] == 'directory':
 
2213
                                if self.target._directory_is_tree_reference(
 
2214
                                    current_path_info[0].decode('utf8')):
 
2215
                                    current_path_info = current_path_info[:2] + \
 
2216
                                        ('tree-reference',) + current_path_info[3:]
 
2217
                        else:
 
2218
                            current_path_info = None
 
2219
                        path_handled = False
 
2220
                    else:
 
2221
                        advance_path = True # reset the advance flagg.
 
2222
                if current_block is not None:
 
2223
                    block_index += 1
 
2224
                    if (block_index < len(state._dirblocks) and
 
2225
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2226
                        current_block = state._dirblocks[block_index]
 
2227
                    else:
 
2228
                        current_block = None
 
2229
                if current_dir_info is not None:
 
2230
                    try:
 
2231
                        current_dir_info = dir_iterator.next()
 
2232
                    except StopIteration:
 
2233
                        current_dir_info = None
 
2234
 
 
2235
 
 
2236
    @staticmethod
 
2237
    def is_compatible(source, target):
 
2238
        # the target must be a dirstate working tree
 
2239
        if not isinstance(target, WorkingTree4):
 
2240
            return False
 
2241
        # the source must be a revtreee or dirstate rev tree.
 
2242
        if not isinstance(source,
 
2243
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2244
            return False
 
2245
        # the source revid must be in the target dirstate
 
2246
        if not (source._revision_id == NULL_REVISION or
 
2247
            source._revision_id in target.get_parent_ids()):
 
2248
            # TODO: what about ghosts? it may well need to 
 
2249
            # check for them explicitly.
 
2250
            return False
 
2251
        return True
 
2252
 
 
2253
InterTree.register_optimiser(InterDirStateTree)
 
2254
 
 
2255
 
 
2256
class Converter3to4(object):
 
2257
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2258
 
 
2259
    def __init__(self):
 
2260
        self.target_format = WorkingTreeFormat4()
 
2261
 
 
2262
    def convert(self, tree):
 
2263
        # lock the control files not the tree, so that we dont get tree
 
2264
        # on-unlock behaviours, and so that noone else diddles with the 
 
2265
        # tree during upgrade.
 
2266
        tree._control_files.lock_write()
 
2267
        try:
 
2268
            self.create_dirstate_data(tree)
 
2269
            self.update_format(tree)
 
2270
            self.remove_xml_files(tree)
 
2271
        finally:
 
2272
            tree._control_files.unlock()
 
2273
 
 
2274
    def create_dirstate_data(self, tree):
 
2275
        """Create the dirstate based data for tree."""
 
2276
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2277
            ).local_abspath('dirstate')
 
2278
        state = dirstate.DirState.from_tree(tree, local_path)
 
2279
        state.save()
 
2280
        state.unlock()
 
2281
 
 
2282
    def remove_xml_files(self, tree):
 
2283
        """Remove the oldformat 3 data."""
 
2284
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2285
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2286
            'pending-merges', 'stat-cache']:
 
2287
            try:
 
2288
                transport.delete(path)
 
2289
            except errors.NoSuchFile:
 
2290
                # some files are optional - just deal.
 
2291
                pass
 
2292
 
 
2293
    def update_format(self, tree):
 
2294
        """Change the format marker."""
 
2295
        tree._control_files.put_utf8('format',
 
2296
            self.target_format.get_format_string())