~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Aaron Bentley
  • Date: 2007-03-12 19:56:41 UTC
  • mto: (1551.19.24 Aaron's mergeable stuff)
  • mto: This revision was merged to the branch mainline in revision 2353.
  • Revision ID: abentley@panoramicfeedback.com-20070312195641-ezjnseqwgjtkh0iu
merge3 auto-detects line endings for conflict markers

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
    @needs_tree_write_lock
 
309
    def _gather_kinds(self, files, kinds):
 
310
        """See MutableTree._gather_kinds."""
 
311
        for pos, f in enumerate(files):
 
312
            if kinds[pos] is None:
 
313
                kinds[pos] = self._kind(f)
 
314
 
 
315
    def _generate_inventory(self):
 
316
        """Create and set self.inventory from the dirstate object.
 
317
        
 
318
        This is relatively expensive: we have to walk the entire dirstate.
 
319
        Ideally we would not, and can deprecate this function.
 
320
        """
 
321
        #: uncomment to trap on inventory requests.
 
322
        # import pdb;pdb.set_trace()
 
323
        state = self.current_dirstate()
 
324
        state._read_dirblocks_if_needed()
 
325
        root_key, current_entry = self._get_entry(path='')
 
326
        current_id = root_key[2]
 
327
        assert current_entry[0][0] == 'd' # directory
 
328
        inv = Inventory(root_id=current_id)
 
329
        # Turn some things into local variables
 
330
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
331
        factory = entry_factory
 
332
        utf8_decode = cache_utf8._utf8_decode
 
333
        inv_byid = inv._byid
 
334
        # we could do this straight out of the dirstate; it might be fast
 
335
        # and should be profiled - RBC 20070216
 
336
        parent_ies = {'' : inv.root}
 
337
        for block in state._dirblocks[1:]: # skip the root
 
338
            dirname = block[0]
 
339
            try:
 
340
                parent_ie = parent_ies[dirname]
 
341
            except KeyError:
 
342
                # all the paths in this block are not versioned in this tree
 
343
                continue
 
344
            for key, entry in block[1]:
 
345
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
346
                if minikind in ('a', 'r'): # absent, relocated
 
347
                    # a parent tree only entry
 
348
                    continue
 
349
                name = key[1]
 
350
                name_unicode = utf8_decode(name)[0]
 
351
                file_id = key[2]
 
352
                kind = minikind_to_kind[minikind]
 
353
                inv_entry = factory[kind](file_id, name_unicode,
 
354
                                          parent_ie.file_id)
 
355
                if kind == 'file':
 
356
                    # not strictly needed: working tree
 
357
                    #entry.executable = executable
 
358
                    #entry.text_size = size
 
359
                    #entry.text_sha1 = sha1
 
360
                    pass
 
361
                elif kind == 'directory':
 
362
                    # add this entry to the parent map.
 
363
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
364
                elif kind == 'tree-reference':
 
365
                    assert self._repo_supports_tree_reference
 
366
                    inv_entry.reference_revision = link_or_sha1 or None
 
367
                else:
 
368
                    assert 'unknown kind'
 
369
                # These checks cost us around 40ms on a 55k entry tree
 
370
                assert file_id not in inv_byid, ('file_id %s already in'
 
371
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
372
                assert name_unicode not in parent_ie.children
 
373
                inv_byid[file_id] = inv_entry
 
374
                parent_ie.children[name_unicode] = inv_entry
 
375
        self._inventory = inv
 
376
 
 
377
    def _get_entry(self, file_id=None, path=None):
 
378
        """Get the dirstate row for file_id or path.
 
379
 
 
380
        If either file_id or path is supplied, it is used as the key to lookup.
 
381
        If both are supplied, the fastest lookup is used, and an error is
 
382
        raised if they do not both point at the same row.
 
383
        
 
384
        :param file_id: An optional unicode file_id to be looked up.
 
385
        :param path: An optional unicode path to be looked up.
 
386
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
387
        """
 
388
        if file_id is None and path is None:
 
389
            raise errors.BzrError('must supply file_id or path')
 
390
        state = self.current_dirstate()
 
391
        if path is not None:
 
392
            path = path.encode('utf8')
 
393
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
394
 
 
395
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
396
        # check file id is valid unconditionally.
 
397
        entry = self._get_entry(file_id=file_id, path=path)
 
398
        assert entry[0] is not None, 'what error should this raise'
 
399
        if path is None:
 
400
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
401
 
 
402
        file_abspath = self.abspath(path)
 
403
        state = self.current_dirstate()
 
404
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
405
                                          stat_value=stat_value)
 
406
        if entry[1][0][0] == 'f':
 
407
            return link_or_sha1
 
408
        return None
 
409
 
 
410
    def _get_inventory(self):
 
411
        """Get the inventory for the tree. This is only valid within a lock."""
 
412
        if self._inventory is not None:
 
413
            return self._inventory
 
414
        self._must_be_locked()
 
415
        self._generate_inventory()
 
416
        return self._inventory
 
417
 
 
418
    inventory = property(_get_inventory,
 
419
                         doc="Inventory of this Tree")
 
420
 
 
421
    @needs_read_lock
 
422
    def get_parent_ids(self):
 
423
        """See Tree.get_parent_ids.
 
424
        
 
425
        This implementation requests the ids list from the dirstate file.
 
426
        """
 
427
        return self.current_dirstate().get_parent_ids()
 
428
 
 
429
    def get_reference_revision(self, file_id, path=None):
 
430
        # referenced tree's revision is whatever's currently there
 
431
        return self.get_nested_tree(file_id, path).last_revision()
 
432
 
 
433
    def get_nested_tree(self, file_id, path=None):
 
434
        if path is None:
 
435
            path = self.id2path(file_id)
 
436
        # else: check file_id is at path?
 
437
        return WorkingTree.open(self.abspath(path))
 
438
 
 
439
    @needs_read_lock
 
440
    def get_root_id(self):
 
441
        """Return the id of this trees root"""
 
442
        return self._get_entry(path='')[0][2]
 
443
 
 
444
    def has_id(self, file_id):
 
445
        state = self.current_dirstate()
 
446
        file_id = osutils.safe_file_id(file_id)
 
447
        row, parents = self._get_entry(file_id=file_id)
 
448
        if row is None:
 
449
            return False
 
450
        return osutils.lexists(pathjoin(
 
451
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
452
 
 
453
    @needs_read_lock
 
454
    def id2path(self, file_id):
 
455
        file_id = osutils.safe_file_id(file_id)
 
456
        state = self.current_dirstate()
 
457
        entry = self._get_entry(file_id=file_id)
 
458
        if entry == (None, None):
 
459
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
460
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
461
        return path_utf8.decode('utf8')
 
462
 
 
463
    @needs_read_lock
 
464
    def __iter__(self):
 
465
        """Iterate through file_ids for this tree.
 
466
 
 
467
        file_ids are in a WorkingTree if they are in the working inventory
 
468
        and the working file exists.
 
469
        """
 
470
        result = []
 
471
        for key, tree_details in self.current_dirstate()._iter_entries():
 
472
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
473
                # not relevant to the working tree
 
474
                continue
 
475
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
476
            if osutils.lexists(path):
 
477
                result.append(key[2])
 
478
        return iter(result)
 
479
 
 
480
    def iter_references(self):
 
481
        for key, tree_details in self.current_dirstate()._iter_entries():
 
482
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
483
                # not relevant to the working tree
 
484
                continue
 
485
            if not key[1]:
 
486
                # the root is not a reference.
 
487
                continue
 
488
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
489
            try:
 
490
                if self._kind(path) == 'tree-reference':
 
491
                    yield path, key[2]
 
492
            except errors.NoSuchFile:
 
493
                # path is missing on disk.
 
494
                continue
 
495
 
 
496
    @needs_read_lock
 
497
    def kind(self, file_id):
 
498
        """Return the kind of a file.
 
499
 
 
500
        This is always the actual kind that's on disk, regardless of what it
 
501
        was added as.
 
502
        """
 
503
        relpath = self.id2path(file_id)
 
504
        assert relpath != None, \
 
505
            "path for id {%s} is None!" % file_id
 
506
        return self._kind(relpath)
 
507
 
 
508
    def _kind(self, relpath):
 
509
        abspath = self.abspath(relpath)
 
510
        kind = file_kind(abspath)
 
511
        if (self._repo_supports_tree_reference and
 
512
            kind == 'directory' and
 
513
            self._directory_is_tree_reference(relpath)):
 
514
            kind = 'tree-reference'
 
515
        return kind
 
516
 
 
517
    @needs_read_lock
 
518
    def _last_revision(self):
 
519
        """See Mutable.last_revision."""
 
520
        parent_ids = self.current_dirstate().get_parent_ids()
 
521
        if parent_ids:
 
522
            return parent_ids[0]
 
523
        else:
 
524
            return None
 
525
 
 
526
    def lock_read(self):
 
527
        """See Branch.lock_read, and WorkingTree.unlock."""
 
528
        self.branch.lock_read()
 
529
        try:
 
530
            self._control_files.lock_read()
 
531
            try:
 
532
                state = self.current_dirstate()
 
533
                if not state._lock_token:
 
534
                    state.lock_read()
 
535
                # set our support for tree references from the repository in
 
536
                # use.
 
537
                self._repo_supports_tree_reference = getattr(
 
538
                    self.branch.repository._format, "support_tree_reference",
 
539
                    False)
 
540
            except:
 
541
                self._control_files.unlock()
 
542
                raise
 
543
        except:
 
544
            self.branch.unlock()
 
545
            raise
 
546
 
 
547
    def _lock_self_write(self):
 
548
        """This should be called after the branch is locked."""
 
549
        try:
 
550
            self._control_files.lock_write()
 
551
            try:
 
552
                state = self.current_dirstate()
 
553
                if not state._lock_token:
 
554
                    state.lock_write()
 
555
                # set our support for tree references from the repository in
 
556
                # use.
 
557
                self._repo_supports_tree_reference = getattr(
 
558
                    self.branch.repository._format, "support_tree_reference",
 
559
                    False)
 
560
            except:
 
561
                self._control_files.unlock()
 
562
                raise
 
563
        except:
 
564
            self.branch.unlock()
 
565
            raise
 
566
 
 
567
    def lock_tree_write(self):
 
568
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
569
        self.branch.lock_read()
 
570
        self._lock_self_write()
 
571
 
 
572
    def lock_write(self):
 
573
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
574
        self.branch.lock_write()
 
575
        self._lock_self_write()
 
576
 
 
577
    @needs_tree_write_lock
 
578
    def move(self, from_paths, to_dir, after=False):
 
579
        """See WorkingTree.move()."""
 
580
        result = []
 
581
        if not from_paths:
 
582
            return result
 
583
 
 
584
        state = self.current_dirstate()
 
585
 
 
586
        assert not isinstance(from_paths, basestring)
 
587
        to_dir_utf8 = to_dir.encode('utf8')
 
588
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
589
        id_index = state._get_id_index()
 
590
        # check destination directory
 
591
        # get the details for it
 
592
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
593
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
594
        if not entry_present:
 
595
            raise errors.BzrMoveFailedError('', to_dir,
 
596
                errors.NotVersionedError(to_dir))
 
597
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
598
        # get a handle on the block itself.
 
599
        to_block_index = state._ensure_block(
 
600
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
601
        to_block = state._dirblocks[to_block_index]
 
602
        to_abs = self.abspath(to_dir)
 
603
        if not isdir(to_abs):
 
604
            raise errors.BzrMoveFailedError('',to_dir,
 
605
                errors.NotADirectory(to_abs))
 
606
 
 
607
        if to_entry[1][0][0] != 'd':
 
608
            raise errors.BzrMoveFailedError('',to_dir,
 
609
                errors.NotADirectory(to_abs))
 
610
 
 
611
        if self._inventory is not None:
 
612
            update_inventory = True
 
613
            inv = self.inventory
 
614
            to_dir_ie = inv[to_dir_id]
 
615
            to_dir_id = to_entry[0][2]
 
616
        else:
 
617
            update_inventory = False
 
618
 
 
619
        rollbacks = []
 
620
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
621
                     fingerprint, packed_stat, size,
 
622
                     to_block, to_key, to_path_utf8):
 
623
            state._make_absent(old_entry)
 
624
            from_key = old_entry[0]
 
625
            rollbacks.append(
 
626
                lambda:state.update_minimal(from_key,
 
627
                    minikind,
 
628
                    executable=executable,
 
629
                    fingerprint=fingerprint,
 
630
                    packed_stat=packed_stat,
 
631
                    size=size,
 
632
                    path_utf8=from_path_utf8))
 
633
            state.update_minimal(to_key,
 
634
                    minikind,
 
635
                    executable=executable,
 
636
                    fingerprint=fingerprint,
 
637
                    packed_stat=packed_stat,
 
638
                    size=size,
 
639
                    path_utf8=to_path_utf8)
 
640
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
641
            new_entry = to_block[1][added_entry_index]
 
642
            rollbacks.append(lambda:state._make_absent(new_entry))
 
643
 
 
644
        # create rename entries and tuples
 
645
        for from_rel in from_paths:
 
646
            # from_rel is 'pathinroot/foo/bar'
 
647
            from_rel_utf8 = from_rel.encode('utf8')
 
648
            from_dirname, from_tail = osutils.split(from_rel)
 
649
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
650
            from_entry = self._get_entry(path=from_rel)
 
651
            if from_entry == (None, None):
 
652
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
653
                    errors.NotVersionedError(path=str(from_rel)))
 
654
 
 
655
            from_id = from_entry[0][2]
 
656
            to_rel = pathjoin(to_dir, from_tail)
 
657
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
658
            item_to_entry = self._get_entry(path=to_rel)
 
659
            if item_to_entry != (None, None):
 
660
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
661
                    "Target is already versioned.")
 
662
 
 
663
            if from_rel == to_rel:
 
664
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
665
                    "Source and target are identical.")
 
666
 
 
667
            from_missing = not self.has_filename(from_rel)
 
668
            to_missing = not self.has_filename(to_rel)
 
669
            if after:
 
670
                move_file = False
 
671
            else:
 
672
                move_file = True
 
673
            if to_missing:
 
674
                if not move_file:
 
675
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
676
                        errors.NoSuchFile(path=to_rel,
 
677
                        extra="New file has not been created yet"))
 
678
                elif from_missing:
 
679
                    # neither path exists
 
680
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
681
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
682
            else:
 
683
                if from_missing: # implicitly just update our path mapping
 
684
                    move_file = False
 
685
                elif not after:
 
686
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
687
                        extra="(Use --after to update the Bazaar id)")
 
688
 
 
689
            rollbacks = []
 
690
            def rollback_rename():
 
691
                """A single rename has failed, roll it back."""
 
692
                # roll back everything, even if we encounter trouble doing one
 
693
                # of them.
 
694
                #
 
695
                # TODO: at least log the other exceptions rather than just
 
696
                # losing them mbp 20070307
 
697
                exc_info = None
 
698
                for rollback in reversed(rollbacks):
 
699
                    try:
 
700
                        rollback()
 
701
                    except Exception, e:
 
702
                        exc_info = sys.exc_info()
 
703
                if exc_info:
 
704
                    raise exc_info[0], exc_info[1], exc_info[2]
 
705
 
 
706
            # perform the disk move first - its the most likely failure point.
 
707
            if move_file:
 
708
                from_rel_abs = self.abspath(from_rel)
 
709
                to_rel_abs = self.abspath(to_rel)
 
710
                try:
 
711
                    osutils.rename(from_rel_abs, to_rel_abs)
 
712
                except OSError, e:
 
713
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
714
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
715
            try:
 
716
                # perform the rename in the inventory next if needed: its easy
 
717
                # to rollback
 
718
                if update_inventory:
 
719
                    # rename the entry
 
720
                    from_entry = inv[from_id]
 
721
                    current_parent = from_entry.parent_id
 
722
                    inv.rename(from_id, to_dir_id, from_tail)
 
723
                    rollbacks.append(
 
724
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
725
                # finally do the rename in the dirstate, which is a little
 
726
                # tricky to rollback, but least likely to need it.
 
727
                old_block_index, old_entry_index, dir_present, file_present = \
 
728
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
729
                old_block = state._dirblocks[old_block_index][1]
 
730
                old_entry = old_block[old_entry_index]
 
731
                from_key, old_entry_details = old_entry
 
732
                cur_details = old_entry_details[0]
 
733
                # remove the old row
 
734
                to_key = ((to_block[0],) + from_key[1:3])
 
735
                minikind = cur_details[0]
 
736
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
737
                         minikind=minikind,
 
738
                         executable=cur_details[3],
 
739
                         fingerprint=cur_details[1],
 
740
                         packed_stat=cur_details[4],
 
741
                         size=cur_details[2],
 
742
                         to_block=to_block,
 
743
                         to_key=to_key,
 
744
                         to_path_utf8=to_rel_utf8)
 
745
 
 
746
                if minikind == 'd':
 
747
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
748
                        """all entries in this block need updating.
 
749
 
 
750
                        TODO: This is pretty ugly, and doesn't support
 
751
                        reverting, but it works.
 
752
                        """
 
753
                        assert from_dir != '', "renaming root not supported"
 
754
                        from_key = (from_dir, '')
 
755
                        from_block_idx, present = \
 
756
                            state._find_block_index_from_key(from_key)
 
757
                        if not present:
 
758
                            # This is the old record, if it isn't present, then
 
759
                            # there is theoretically nothing to update.
 
760
                            # (Unless it isn't present because of lazy loading,
 
761
                            # but we don't do that yet)
 
762
                            return
 
763
                        from_block = state._dirblocks[from_block_idx]
 
764
                        to_block_index, to_entry_index, _, _ = \
 
765
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
766
                        to_block_index = state._ensure_block(
 
767
                            to_block_index, to_entry_index, to_dir_utf8)
 
768
                        to_block = state._dirblocks[to_block_index]
 
769
                        for entry in from_block[1]:
 
770
                            assert entry[0][0] == from_dir
 
771
                            cur_details = entry[1][0]
 
772
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
773
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
774
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
775
                            minikind = cur_details[0]
 
776
                            move_one(entry, from_path_utf8=from_path_utf8,
 
777
                                     minikind=minikind,
 
778
                                     executable=cur_details[3],
 
779
                                     fingerprint=cur_details[1],
 
780
                                     packed_stat=cur_details[4],
 
781
                                     size=cur_details[2],
 
782
                                     to_block=to_block,
 
783
                                     to_key=to_key,
 
784
                                     to_path_utf8=to_rel_utf8)
 
785
                            if minikind == 'd':
 
786
                                # We need to move all the children of this
 
787
                                # entry
 
788
                                update_dirblock(from_path_utf8, to_key,
 
789
                                                to_path_utf8)
 
790
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
791
            except:
 
792
                rollback_rename()
 
793
                raise
 
794
            result.append((from_rel, to_rel))
 
795
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
796
            self._make_dirty(reset_inventory=False)
 
797
 
 
798
        return result
 
799
 
 
800
    def _must_be_locked(self):
 
801
        if not self._control_files._lock_count:
 
802
            raise errors.ObjectNotLocked(self)
 
803
 
 
804
    def _new_tree(self):
 
805
        """Initialize the state in this tree to be a new tree."""
 
806
        self._dirty = True
 
807
 
 
808
    @needs_read_lock
 
809
    def path2id(self, path):
 
810
        """Return the id for path in this tree."""
 
811
        path = path.strip('/')
 
812
        entry = self._get_entry(path=path)
 
813
        if entry == (None, None):
 
814
            return None
 
815
        return entry[0][2]
 
816
 
 
817
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
818
        """See Tree.paths2ids().
 
819
 
 
820
        This specialisation fast-paths the case where all the trees are in the
 
821
        dirstate.
 
822
        """
 
823
        if paths is None:
 
824
            return None
 
825
        parents = self.get_parent_ids()
 
826
        for tree in trees:
 
827
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
828
                parents):
 
829
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
830
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
831
        # -- make all paths utf8 --
 
832
        paths_utf8 = set()
 
833
        for path in paths:
 
834
            paths_utf8.add(path.encode('utf8'))
 
835
        paths = paths_utf8
 
836
        # -- paths is now a utf8 path set --
 
837
        # -- get the state object and prepare it.
 
838
        state = self.current_dirstate()
 
839
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
840
            and '' not in paths):
 
841
            paths2ids = self._paths2ids_using_bisect
 
842
        else:
 
843
            paths2ids = self._paths2ids_in_memory
 
844
        return paths2ids(paths, search_indexes,
 
845
                         require_versioned=require_versioned)
 
846
 
 
847
    def _paths2ids_in_memory(self, paths, search_indexes,
 
848
                             require_versioned=True):
 
849
        state = self.current_dirstate()
 
850
        state._read_dirblocks_if_needed()
 
851
        def _entries_for_path(path):
 
852
            """Return a list with all the entries that match path for all ids.
 
853
            """
 
854
            dirname, basename = os.path.split(path)
 
855
            key = (dirname, basename, '')
 
856
            block_index, present = state._find_block_index_from_key(key)
 
857
            if not present:
 
858
                # the block which should contain path is absent.
 
859
                return []
 
860
            result = []
 
861
            block = state._dirblocks[block_index][1]
 
862
            entry_index, _ = state._find_entry_index(key, block)
 
863
            # we may need to look at multiple entries at this path: walk while the paths match.
 
864
            while (entry_index < len(block) and
 
865
                block[entry_index][0][0:2] == key[0:2]):
 
866
                result.append(block[entry_index])
 
867
                entry_index += 1
 
868
            return result
 
869
        if require_versioned:
 
870
            # -- check all supplied paths are versioned in a search tree. --
 
871
            all_versioned = True
 
872
            for path in paths:
 
873
                path_entries = _entries_for_path(path)
 
874
                if not path_entries:
 
875
                    # this specified path is not present at all: error
 
876
                    all_versioned = False
 
877
                    break
 
878
                found_versioned = False
 
879
                # for each id at this path
 
880
                for entry in path_entries:
 
881
                    # for each tree.
 
882
                    for index in search_indexes:
 
883
                        if entry[1][index][0] != 'a': # absent
 
884
                            found_versioned = True
 
885
                            # all good: found a versioned cell
 
886
                            break
 
887
                if not found_versioned:
 
888
                    # none of the indexes was not 'absent' at all ids for this
 
889
                    # path.
 
890
                    all_versioned = False
 
891
                    break
 
892
            if not all_versioned:
 
893
                raise errors.PathsNotVersionedError(paths)
 
894
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
895
        search_paths = set()
 
896
        for path in paths:
 
897
            other_paths = paths.difference(set([path]))
 
898
            if not osutils.is_inside_any(other_paths, path):
 
899
                # this is a top level path, we must check it.
 
900
                search_paths.add(path)
 
901
        # sketch: 
 
902
        # for all search_indexs in each path at or under each element of
 
903
        # search_paths, if the detail is relocated: add the id, and add the
 
904
        # relocated path as one to search if its not searched already. If the
 
905
        # detail is not relocated, add the id.
 
906
        searched_paths = set()
 
907
        found_ids = set()
 
908
        def _process_entry(entry):
 
909
            """Look at search_indexes within entry.
 
910
 
 
911
            If a specific tree's details are relocated, add the relocation
 
912
            target to search_paths if not searched already. If it is absent, do
 
913
            nothing. Otherwise add the id to found_ids.
 
914
            """
 
915
            for index in search_indexes:
 
916
                if entry[1][index][0] == 'r': # relocated
 
917
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
918
                        search_paths.add(entry[1][index][1])
 
919
                elif entry[1][index][0] != 'a': # absent
 
920
                    found_ids.add(entry[0][2])
 
921
        while search_paths:
 
922
            current_root = search_paths.pop()
 
923
            searched_paths.add(current_root)
 
924
            # process the entries for this containing directory: the rest will be
 
925
            # found by their parents recursively.
 
926
            root_entries = _entries_for_path(current_root)
 
927
            if not root_entries:
 
928
                # this specified path is not present at all, skip it.
 
929
                continue
 
930
            for entry in root_entries:
 
931
                _process_entry(entry)
 
932
            initial_key = (current_root, '', '')
 
933
            block_index, _ = state._find_block_index_from_key(initial_key)
 
934
            while (block_index < len(state._dirblocks) and
 
935
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
936
                for entry in state._dirblocks[block_index][1]:
 
937
                    _process_entry(entry)
 
938
                block_index += 1
 
939
        return found_ids
 
940
 
 
941
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
942
                                require_versioned=True):
 
943
        state = self.current_dirstate()
 
944
        found_ids = set()
 
945
 
 
946
        split_paths = sorted(osutils.split(p) for p in paths)
 
947
        found = state._bisect_recursive(split_paths)
 
948
 
 
949
        if require_versioned:
 
950
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
951
            for dir_name in split_paths:
 
952
                if dir_name not in found_dir_names:
 
953
                    raise errors.PathsNotVersionedError(paths)
 
954
 
 
955
        for dir_name_id, trees_info in found.iteritems():
 
956
            for index in search_indexes:
 
957
                if trees_info[index][0] not in ('r', 'a'):
 
958
                    found_ids.add(dir_name_id[2])
 
959
        return found_ids
 
960
 
 
961
    def read_working_inventory(self):
 
962
        """Read the working inventory.
 
963
        
 
964
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
965
        """
 
966
        return self.inventory
 
967
 
 
968
    @needs_read_lock
 
969
    def revision_tree(self, revision_id):
 
970
        """See Tree.revision_tree.
 
971
 
 
972
        WorkingTree4 supplies revision_trees for any basis tree.
 
973
        """
 
974
        revision_id = osutils.safe_revision_id(revision_id)
 
975
        dirstate = self.current_dirstate()
 
976
        parent_ids = dirstate.get_parent_ids()
 
977
        if revision_id not in parent_ids:
 
978
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
979
        if revision_id in dirstate.get_ghosts():
 
980
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
981
        return DirStateRevisionTree(dirstate, revision_id,
 
982
            self.branch.repository)
 
983
 
 
984
    @needs_tree_write_lock
 
985
    def set_last_revision(self, new_revision):
 
986
        """Change the last revision in the working tree."""
 
987
        new_revision = osutils.safe_revision_id(new_revision)
 
988
        parents = self.get_parent_ids()
 
989
        if new_revision in (NULL_REVISION, None):
 
990
            assert len(parents) < 2, (
 
991
                "setting the last parent to none with a pending merge is "
 
992
                "unsupported.")
 
993
            self.set_parent_ids([])
 
994
        else:
 
995
            self.set_parent_ids([new_revision] + parents[1:],
 
996
                allow_leftmost_as_ghost=True)
 
997
 
 
998
    @needs_tree_write_lock
 
999
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1000
        """Set the parent ids to revision_ids.
 
1001
        
 
1002
        See also set_parent_trees. This api will try to retrieve the tree data
 
1003
        for each element of revision_ids from the trees repository. If you have
 
1004
        tree data already available, it is more efficient to use
 
1005
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1006
        an easier API to use.
 
1007
 
 
1008
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1009
            working tree. Any of these may be ghosts.
 
1010
        """
 
1011
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1012
        trees = []
 
1013
        for revision_id in revision_ids:
 
1014
            try:
 
1015
                revtree = self.branch.repository.revision_tree(revision_id)
 
1016
                # TODO: jam 20070213 KnitVersionedFile raises
 
1017
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1018
                #       given revision_id is not present. Should Repository be
 
1019
                #       catching it and re-raising NoSuchRevision?
 
1020
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1021
                revtree = None
 
1022
            trees.append((revision_id, revtree))
 
1023
        self.current_dirstate()._validate()
 
1024
        self.set_parent_trees(trees,
 
1025
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1026
        self.current_dirstate()._validate()
 
1027
 
 
1028
    @needs_tree_write_lock
 
1029
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1030
        """Set the parents of the working tree.
 
1031
 
 
1032
        :param parents_list: A list of (revision_id, tree) tuples.
 
1033
            If tree is None, then that element is treated as an unreachable
 
1034
            parent tree - i.e. a ghost.
 
1035
        """
 
1036
        dirstate = self.current_dirstate()
 
1037
        dirstate._validate()
 
1038
        if len(parents_list) > 0:
 
1039
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1040
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1041
        real_trees = []
 
1042
        ghosts = []
 
1043
        # convert absent trees to the null tree, which we convert back to
 
1044
        # missing on access.
 
1045
        for rev_id, tree in parents_list:
 
1046
            rev_id = osutils.safe_revision_id(rev_id)
 
1047
            if tree is not None:
 
1048
                real_trees.append((rev_id, tree))
 
1049
            else:
 
1050
                real_trees.append((rev_id,
 
1051
                    self.branch.repository.revision_tree(None)))
 
1052
                ghosts.append(rev_id)
 
1053
        dirstate._validate()
 
1054
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1055
        dirstate._validate()
 
1056
        self._make_dirty(reset_inventory=False)
 
1057
        dirstate._validate()
 
1058
 
 
1059
    def _set_root_id(self, file_id):
 
1060
        """See WorkingTree.set_root_id."""
 
1061
        state = self.current_dirstate()
 
1062
        state.set_path_id('', file_id)
 
1063
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1064
            self._make_dirty(reset_inventory=True)
 
1065
 
 
1066
    @needs_read_lock
 
1067
    def supports_tree_reference(self):
 
1068
        return self._repo_supports_tree_reference
 
1069
 
 
1070
    def unlock(self):
 
1071
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1072
        if self._control_files._lock_count == 1:
 
1073
            # eventually we should do signature checking during read locks for
 
1074
            # dirstate updates.
 
1075
            if self._control_files._lock_mode == 'w':
 
1076
                if self._dirty:
 
1077
                    self.flush()
 
1078
            if self._dirstate is not None:
 
1079
                # This is a no-op if there are no modifications.
 
1080
                self._dirstate.save()
 
1081
                self._dirstate.unlock()
 
1082
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1083
            #       point. Instead, it could check if the header has been
 
1084
            #       modified when it is locked, and if not, it can hang on to
 
1085
            #       the data it has in memory.
 
1086
            self._dirstate = None
 
1087
            self._inventory = None
 
1088
        # reverse order of locking.
 
1089
        try:
 
1090
            return self._control_files.unlock()
 
1091
        finally:
 
1092
            self.branch.unlock()
 
1093
 
 
1094
    @needs_tree_write_lock
 
1095
    def unversion(self, file_ids):
 
1096
        """Remove the file ids in file_ids from the current versioned set.
 
1097
 
 
1098
        When a file_id is unversioned, all of its children are automatically
 
1099
        unversioned.
 
1100
 
 
1101
        :param file_ids: The file ids to stop versioning.
 
1102
        :raises: NoSuchId if any fileid is not currently versioned.
 
1103
        """
 
1104
        if not file_ids:
 
1105
            return
 
1106
        state = self.current_dirstate()
 
1107
        state._read_dirblocks_if_needed()
 
1108
        ids_to_unversion = set()
 
1109
        for file_id in file_ids:
 
1110
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1111
        paths_to_unversion = set()
 
1112
        # sketch:
 
1113
        # check if the root is to be unversioned, if so, assert for now.
 
1114
        # walk the state marking unversioned things as absent.
 
1115
        # if there are any un-unversioned ids at the end, raise
 
1116
        for key, details in state._dirblocks[0][1]:
 
1117
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1118
                key[2] in ids_to_unversion):
 
1119
                # I haven't written the code to unversion / yet - it should be
 
1120
                # supported.
 
1121
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1122
        block_index = 0
 
1123
        while block_index < len(state._dirblocks):
 
1124
            # process one directory at a time.
 
1125
            block = state._dirblocks[block_index]
 
1126
            # first check: is the path one to remove - it or its children
 
1127
            delete_block = False
 
1128
            for path in paths_to_unversion:
 
1129
                if (block[0].startswith(path) and
 
1130
                    (len(block[0]) == len(path) or
 
1131
                     block[0][len(path)] == '/')):
 
1132
                    # this entire block should be deleted - its the block for a
 
1133
                    # path to unversion; or the child of one
 
1134
                    delete_block = True
 
1135
                    break
 
1136
            # TODO: trim paths_to_unversion as we pass by paths
 
1137
            if delete_block:
 
1138
                # this block is to be deleted: process it.
 
1139
                # TODO: we can special case the no-parents case and
 
1140
                # just forget the whole block.
 
1141
                entry_index = 0
 
1142
                while entry_index < len(block[1]):
 
1143
                    # Mark this file id as having been removed
 
1144
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1145
                    if not state._make_absent(block[1][entry_index]):
 
1146
                        entry_index += 1
 
1147
                # go to the next block. (At the moment we dont delete empty
 
1148
                # dirblocks)
 
1149
                block_index += 1
 
1150
                continue
 
1151
            entry_index = 0
 
1152
            while entry_index < len(block[1]):
 
1153
                entry = block[1][entry_index]
 
1154
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1155
                    # ^ some parent row.
 
1156
                    entry[0][2] not in ids_to_unversion):
 
1157
                    # ^ not an id to unversion
 
1158
                    entry_index += 1
 
1159
                    continue
 
1160
                if entry[1][0][0] == 'd':
 
1161
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1162
                if not state._make_absent(entry):
 
1163
                    entry_index += 1
 
1164
                # we have unversioned this id
 
1165
                ids_to_unversion.remove(entry[0][2])
 
1166
            block_index += 1
 
1167
        if ids_to_unversion:
 
1168
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1169
        self._make_dirty(reset_inventory=False)
 
1170
        # have to change the legacy inventory too.
 
1171
        if self._inventory is not None:
 
1172
            for file_id in file_ids:
 
1173
                self._inventory.remove_recursive_id(file_id)
 
1174
 
 
1175
    @needs_tree_write_lock
 
1176
    def _write_inventory(self, inv):
 
1177
        """Write inventory as the current inventory."""
 
1178
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1179
        self.current_dirstate().set_state_from_inventory(inv)
 
1180
        self._make_dirty(reset_inventory=False)
 
1181
        if self._inventory is not None:
 
1182
            self._inventory = inv
 
1183
        self.flush()
 
1184
 
 
1185
 
 
1186
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1187
    """The first consolidated dirstate working tree format.
 
1188
 
 
1189
    This format:
 
1190
        - exists within a metadir controlling .bzr
 
1191
        - includes an explicit version marker for the workingtree control
 
1192
          files, separate from the BzrDir format
 
1193
        - modifies the hash cache format
 
1194
        - is new in bzr 0.15
 
1195
        - uses a LockDir to guard access to it.
 
1196
    """
 
1197
 
 
1198
    def get_format_string(self):
 
1199
        """See WorkingTreeFormat.get_format_string()."""
 
1200
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1201
 
 
1202
    def get_format_description(self):
 
1203
        """See WorkingTreeFormat.get_format_description()."""
 
1204
        return "Working tree format 4"
 
1205
 
 
1206
    def initialize(self, a_bzrdir, revision_id=None):
 
1207
        """See WorkingTreeFormat.initialize().
 
1208
 
 
1209
        :param revision_id: allows creating a working tree at a different
 
1210
        revision than the branch is at.
 
1211
 
 
1212
        These trees get an initial random root id.
 
1213
        """
 
1214
        revision_id = osutils.safe_revision_id(revision_id)
 
1215
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1216
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1217
        transport = a_bzrdir.get_workingtree_transport(self)
 
1218
        control_files = self._open_control_files(a_bzrdir)
 
1219
        control_files.create_lock()
 
1220
        control_files.lock_write()
 
1221
        control_files.put_utf8('format', self.get_format_string())
 
1222
        branch = a_bzrdir.open_branch()
 
1223
        if revision_id is None:
 
1224
            revision_id = branch.last_revision()
 
1225
        local_path = transport.local_abspath('dirstate')
 
1226
        # write out new dirstate (must exist when we create the tree)
 
1227
        state = dirstate.DirState.initialize(local_path)
 
1228
        state.unlock()
 
1229
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1230
                         branch,
 
1231
                         _format=self,
 
1232
                         _bzrdir=a_bzrdir,
 
1233
                         _control_files=control_files)
 
1234
        wt._new_tree()
 
1235
        wt.lock_tree_write()
 
1236
        state._validate()
 
1237
        try:
 
1238
            if revision_id in (None, NULL_REVISION):
 
1239
                wt._set_root_id(generate_ids.gen_root_id())
 
1240
                wt.flush()
 
1241
                wt.current_dirstate()._validate()
 
1242
            wt.set_last_revision(revision_id)
 
1243
            wt.flush()
 
1244
            basis = wt.basis_tree()
 
1245
            basis.lock_read()
 
1246
            # if the basis has a root id we have to use that; otherwise we use
 
1247
            # a new random one
 
1248
            basis_root_id = basis.get_root_id()
 
1249
            if basis_root_id is not None:
 
1250
                wt._set_root_id(basis_root_id)
 
1251
                wt.flush()
 
1252
            transform.build_tree(basis, wt)
 
1253
            basis.unlock()
 
1254
        finally:
 
1255
            control_files.unlock()
 
1256
            wt.unlock()
 
1257
        return wt
 
1258
 
 
1259
    def _open(self, a_bzrdir, control_files):
 
1260
        """Open the tree itself.
 
1261
 
 
1262
        :param a_bzrdir: the dir for the tree.
 
1263
        :param control_files: the control files for the tree.
 
1264
        """
 
1265
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1266
                           branch=a_bzrdir.open_branch(),
 
1267
                           _format=self,
 
1268
                           _bzrdir=a_bzrdir,
 
1269
                           _control_files=control_files)
 
1270
 
 
1271
    def __get_matchingbzrdir(self):
 
1272
        # please test against something that will let us do tree references
 
1273
        return bzrdir.format_registry.make_bzrdir(
 
1274
            'dirstate-with-subtree')
 
1275
 
 
1276
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1277
 
 
1278
 
 
1279
class DirStateRevisionTree(Tree):
 
1280
    """A revision tree pulling the inventory from a dirstate."""
 
1281
 
 
1282
    def __init__(self, dirstate, revision_id, repository):
 
1283
        self._dirstate = dirstate
 
1284
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1285
        self._repository = repository
 
1286
        self._inventory = None
 
1287
        self._locked = 0
 
1288
        self._dirstate_locked = False
 
1289
 
 
1290
    def __repr__(self):
 
1291
        return "<%s of %s in %s>" % \
 
1292
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1293
 
 
1294
    def annotate_iter(self, file_id):
 
1295
        """See Tree.annotate_iter"""
 
1296
        w = self._repository.weave_store.get_weave(file_id,
 
1297
                           self._repository.get_transaction())
 
1298
        return w.annotate_iter(self.inventory[file_id].revision)
 
1299
 
 
1300
    def _comparison_data(self, entry, path):
 
1301
        """See Tree._comparison_data."""
 
1302
        if entry is None:
 
1303
            return None, False, None
 
1304
        # trust the entry as RevisionTree does, but this may not be
 
1305
        # sensible: the entry might not have come from us?
 
1306
        return entry.kind, entry.executable, None
 
1307
 
 
1308
    def _file_size(self, entry, stat_value):
 
1309
        return entry.text_size
 
1310
 
 
1311
    def filter_unversioned_files(self, paths):
 
1312
        """Filter out paths that are not versioned.
 
1313
 
 
1314
        :return: set of paths.
 
1315
        """
 
1316
        pred = self.has_filename
 
1317
        return set((p for p in paths if not pred(p)))
 
1318
 
 
1319
    def get_root_id(self):
 
1320
        return self.path2id('')
 
1321
 
 
1322
    def _get_parent_index(self):
 
1323
        """Return the index in the dirstate referenced by this tree."""
 
1324
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1325
 
 
1326
    def _get_entry(self, file_id=None, path=None):
 
1327
        """Get the dirstate row for file_id or path.
 
1328
 
 
1329
        If either file_id or path is supplied, it is used as the key to lookup.
 
1330
        If both are supplied, the fastest lookup is used, and an error is
 
1331
        raised if they do not both point at the same row.
 
1332
        
 
1333
        :param file_id: An optional unicode file_id to be looked up.
 
1334
        :param path: An optional unicode path to be looked up.
 
1335
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1336
        """
 
1337
        if file_id is None and path is None:
 
1338
            raise errors.BzrError('must supply file_id or path')
 
1339
        file_id = osutils.safe_file_id(file_id)
 
1340
        if path is not None:
 
1341
            path = path.encode('utf8')
 
1342
        parent_index = self._get_parent_index()
 
1343
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1344
 
 
1345
    def _generate_inventory(self):
 
1346
        """Create and set self.inventory from the dirstate object.
 
1347
 
 
1348
        (So this is only called the first time the inventory is requested for
 
1349
        this tree; it then remains in memory until it's out of date.)
 
1350
 
 
1351
        This is relatively expensive: we have to walk the entire dirstate.
 
1352
        """
 
1353
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1354
            'dirstate revision tree'
 
1355
        # separate call for profiling - makes it clear where the costs are.
 
1356
        self._dirstate._read_dirblocks_if_needed()
 
1357
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1358
            'parent %s has disappeared from %s' % (
 
1359
            self._revision_id, self._dirstate.get_parent_ids())
 
1360
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1361
        # This is identical now to the WorkingTree _generate_inventory except
 
1362
        # for the tree index use.
 
1363
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1364
        current_id = root_key[2]
 
1365
        assert current_entry[parent_index][0] == 'd'
 
1366
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1367
        inv.root.revision = current_entry[parent_index][4]
 
1368
        # Turn some things into local variables
 
1369
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1370
        factory = entry_factory
 
1371
        utf8_decode = cache_utf8._utf8_decode
 
1372
        inv_byid = inv._byid
 
1373
        # we could do this straight out of the dirstate; it might be fast
 
1374
        # and should be profiled - RBC 20070216
 
1375
        parent_ies = {'' : inv.root}
 
1376
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1377
            dirname = block[0]
 
1378
            try:
 
1379
                parent_ie = parent_ies[dirname]
 
1380
            except KeyError:
 
1381
                # all the paths in this block are not versioned in this tree
 
1382
                continue
 
1383
            for key, entry in block[1]:
 
1384
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1385
                if minikind in ('a', 'r'): # absent, relocated
 
1386
                    # not this tree
 
1387
                    continue
 
1388
                name = key[1]
 
1389
                name_unicode = utf8_decode(name)[0]
 
1390
                file_id = key[2]
 
1391
                kind = minikind_to_kind[minikind]
 
1392
                inv_entry = factory[kind](file_id, name_unicode,
 
1393
                                          parent_ie.file_id)
 
1394
                inv_entry.revision = revid
 
1395
                if kind == 'file':
 
1396
                    inv_entry.executable = executable
 
1397
                    inv_entry.text_size = size
 
1398
                    inv_entry.text_sha1 = fingerprint
 
1399
                elif kind == 'directory':
 
1400
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1401
                elif kind == 'symlink':
 
1402
                    inv_entry.executable = False
 
1403
                    inv_entry.text_size = size
 
1404
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1405
                elif kind == 'tree-reference':
 
1406
                    inv_entry.reference_revision = fingerprint or None
 
1407
                else:
 
1408
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1409
                            % entry)
 
1410
                # These checks cost us around 40ms on a 55k entry tree
 
1411
                assert file_id not in inv_byid
 
1412
                assert name_unicode not in parent_ie.children
 
1413
                inv_byid[file_id] = inv_entry
 
1414
                parent_ie.children[name_unicode] = inv_entry
 
1415
        self._inventory = inv
 
1416
 
 
1417
    def get_file_mtime(self, file_id, path=None):
 
1418
        """Return the modification time for this record.
 
1419
 
 
1420
        We return the timestamp of the last-changed revision.
 
1421
        """
 
1422
        # Make sure the file exists
 
1423
        entry = self._get_entry(file_id, path=path)
 
1424
        if entry == (None, None): # do we raise?
 
1425
            return None
 
1426
        parent_index = self._get_parent_index()
 
1427
        last_changed_revision = entry[1][parent_index][4]
 
1428
        return self._repository.get_revision(last_changed_revision).timestamp
 
1429
 
 
1430
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1431
        entry = self._get_entry(file_id=file_id, path=path)
 
1432
        parent_index = self._get_parent_index()
 
1433
        parent_details = entry[1][parent_index]
 
1434
        if parent_details[0] == 'f':
 
1435
            return parent_details[1]
 
1436
        return None
 
1437
 
 
1438
    def get_file(self, file_id):
 
1439
        return StringIO(self.get_file_text(file_id))
 
1440
 
 
1441
    def get_file_lines(self, file_id):
 
1442
        ie = self.inventory[file_id]
 
1443
        return self._repository.weave_store.get_weave(file_id,
 
1444
                self._repository.get_transaction()).get_lines(ie.revision)
 
1445
 
 
1446
    def get_file_size(self, file_id):
 
1447
        return self.inventory[file_id].text_size
 
1448
 
 
1449
    def get_file_text(self, file_id):
 
1450
        return ''.join(self.get_file_lines(file_id))
 
1451
 
 
1452
    def get_reference_revision(self, file_id, path=None):
 
1453
        return self.inventory[file_id].reference_revision
 
1454
 
 
1455
    def get_symlink_target(self, file_id):
 
1456
        entry = self._get_entry(file_id=file_id)
 
1457
        parent_index = self._get_parent_index()
 
1458
        if entry[1][parent_index][0] != 'l':
 
1459
            return None
 
1460
        else:
 
1461
            # At present, none of the tree implementations supports non-ascii
 
1462
            # symlink targets. So we will just assume that the dirstate path is
 
1463
            # correct.
 
1464
            return entry[1][parent_index][1]
 
1465
 
 
1466
    def get_revision_id(self):
 
1467
        """Return the revision id for this tree."""
 
1468
        return self._revision_id
 
1469
 
 
1470
    def _get_inventory(self):
 
1471
        if self._inventory is not None:
 
1472
            return self._inventory
 
1473
        self._must_be_locked()
 
1474
        self._generate_inventory()
 
1475
        return self._inventory
 
1476
 
 
1477
    inventory = property(_get_inventory,
 
1478
                         doc="Inventory of this Tree")
 
1479
 
 
1480
    def get_parent_ids(self):
 
1481
        """The parents of a tree in the dirstate are not cached."""
 
1482
        return self._repository.get_revision(self._revision_id).parent_ids
 
1483
 
 
1484
    def has_filename(self, filename):
 
1485
        return bool(self.path2id(filename))
 
1486
 
 
1487
    def kind(self, file_id):
 
1488
        return self.inventory[file_id].kind
 
1489
 
 
1490
    def is_executable(self, file_id, path=None):
 
1491
        ie = self.inventory[file_id]
 
1492
        if ie.kind != "file":
 
1493
            return None
 
1494
        return ie.executable
 
1495
 
 
1496
    def list_files(self, include_root=False):
 
1497
        # We use a standard implementation, because DirStateRevisionTree is
 
1498
        # dealing with one of the parents of the current state
 
1499
        inv = self._get_inventory()
 
1500
        entries = inv.iter_entries()
 
1501
        if self.inventory.root is not None and not include_root:
 
1502
            entries.next()
 
1503
        for path, entry in entries:
 
1504
            yield path, 'V', entry.kind, entry.file_id, entry
 
1505
 
 
1506
    def lock_read(self):
 
1507
        """Lock the tree for a set of operations."""
 
1508
        if not self._locked:
 
1509
            self._repository.lock_read()
 
1510
            if self._dirstate._lock_token is None:
 
1511
                self._dirstate.lock_read()
 
1512
                self._dirstate_locked = True
 
1513
        self._locked += 1
 
1514
 
 
1515
    def _must_be_locked(self):
 
1516
        if not self._locked:
 
1517
            raise errors.ObjectNotLocked(self)
 
1518
 
 
1519
    @needs_read_lock
 
1520
    def path2id(self, path):
 
1521
        """Return the id for path in this tree."""
 
1522
        # lookup by path: faster than splitting and walking the ivnentory.
 
1523
        entry = self._get_entry(path=path)
 
1524
        if entry == (None, None):
 
1525
            return None
 
1526
        return entry[0][2]
 
1527
 
 
1528
    def unlock(self):
 
1529
        """Unlock, freeing any cache memory used during the lock."""
 
1530
        # outside of a lock, the inventory is suspect: release it.
 
1531
        self._locked -=1
 
1532
        if not self._locked:
 
1533
            self._inventory = None
 
1534
            self._locked = 0
 
1535
            if self._dirstate_locked:
 
1536
                self._dirstate.unlock()
 
1537
                self._dirstate_locked = False
 
1538
            self._repository.unlock()
 
1539
 
 
1540
    def walkdirs(self, prefix=""):
 
1541
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1542
        # implementation based on an inventory.  
 
1543
        # This should be cleaned up to use the much faster Dirstate code
 
1544
        # So for now, we just build up the parent inventory, and extract
 
1545
        # it the same way RevisionTree does.
 
1546
        _directory = 'directory'
 
1547
        inv = self._get_inventory()
 
1548
        top_id = inv.path2id(prefix)
 
1549
        if top_id is None:
 
1550
            pending = []
 
1551
        else:
 
1552
            pending = [(prefix, top_id)]
 
1553
        while pending:
 
1554
            dirblock = []
 
1555
            relpath, file_id = pending.pop()
 
1556
            # 0 - relpath, 1- file-id
 
1557
            if relpath:
 
1558
                relroot = relpath + '/'
 
1559
            else:
 
1560
                relroot = ""
 
1561
            # FIXME: stash the node in pending
 
1562
            entry = inv[file_id]
 
1563
            for name, child in entry.sorted_children():
 
1564
                toppath = relroot + name
 
1565
                dirblock.append((toppath, name, child.kind, None,
 
1566
                    child.file_id, child.kind
 
1567
                    ))
 
1568
            yield (relpath, entry.file_id), dirblock
 
1569
            # push the user specified dirs from dirblock
 
1570
            for dir in reversed(dirblock):
 
1571
                if dir[2] == _directory:
 
1572
                    pending.append((dir[0], dir[4]))
 
1573
 
 
1574
 
 
1575
class InterDirStateTree(InterTree):
 
1576
    """Fast path optimiser for changes_from with dirstate trees.
 
1577
    
 
1578
    This is used only when both trees are in the dirstate working file, and 
 
1579
    the source is any parent within the dirstate, and the destination is 
 
1580
    the current working tree of the same dirstate.
 
1581
    """
 
1582
    # this could be generalized to allow comparisons between any trees in the
 
1583
    # dirstate, and possibly between trees stored in different dirstates.
 
1584
 
 
1585
    def __init__(self, source, target):
 
1586
        super(InterDirStateTree, self).__init__(source, target)
 
1587
        if not InterDirStateTree.is_compatible(source, target):
 
1588
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1589
 
 
1590
    @staticmethod
 
1591
    def make_source_parent_tree(source, target):
 
1592
        """Change the source tree into a parent of the target."""
 
1593
        revid = source.commit('record tree')
 
1594
        target.branch.repository.fetch(source.branch.repository, revid)
 
1595
        target.set_parent_ids([revid])
 
1596
        return target.basis_tree(), target
 
1597
 
 
1598
    _matching_from_tree_format = WorkingTreeFormat4()
 
1599
    _matching_to_tree_format = WorkingTreeFormat4()
 
1600
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1601
 
 
1602
    def _iter_changes(self, include_unchanged=False,
 
1603
                      specific_files=None, pb=None, extra_trees=[],
 
1604
                      require_versioned=True, want_unversioned=False):
 
1605
        """Return the changes from source to target.
 
1606
 
 
1607
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1608
            for details.
 
1609
        :param specific_files: An optional list of file paths to restrict the
 
1610
            comparison to. When mapping filenames to ids, all matches in all
 
1611
            trees (including optional extra_trees) are used, and all children of
 
1612
            matched directories are included.
 
1613
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1614
            unchanged entries in the result.
 
1615
        :param extra_trees: An optional list of additional trees to use when
 
1616
            mapping the contents of specific_files (paths) to file_ids.
 
1617
        :param require_versioned: If True, all files in specific_files must be
 
1618
            versioned in one of source, target, extra_trees or
 
1619
            PathsNotVersionedError is raised.
 
1620
        :param want_unversioned: Should unversioned files be returned in the
 
1621
            output. An unversioned file is defined as one with (False, False)
 
1622
            for the versioned pair.
 
1623
        """
 
1624
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1625
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1626
        # NB: show_status depends on being able to pass in non-versioned files
 
1627
        # and report them as unknown
 
1628
        # TODO: handle extra trees in the dirstate.
 
1629
        # TODO: handle comparisons as an empty tree as a different special
 
1630
        # case? mbp 20070226
 
1631
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1632
            # we can't fast-path these cases (yet)
 
1633
            for f in super(InterDirStateTree, self)._iter_changes(
 
1634
                include_unchanged, specific_files, pb, extra_trees,
 
1635
                require_versioned, want_unversioned=want_unversioned):
 
1636
                yield f
 
1637
            return
 
1638
        parent_ids = self.target.get_parent_ids()
 
1639
        assert (self.source._revision_id in parent_ids), \
 
1640
                "revision {%s} is not stored in {%s}, but %s " \
 
1641
                "can only be used for trees stored in the dirstate" \
 
1642
                % (self.source._revision_id, self.target, self._iter_changes)
 
1643
        target_index = 0
 
1644
        if self.source._revision_id == NULL_REVISION:
 
1645
            source_index = None
 
1646
            indices = (target_index,)
 
1647
        else:
 
1648
            assert (self.source._revision_id in parent_ids), \
 
1649
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1650
                self.source._revision_id, parent_ids)
 
1651
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1652
            indices = (source_index,target_index)
 
1653
        # -- make all specific_files utf8 --
 
1654
        if specific_files:
 
1655
            specific_files_utf8 = set()
 
1656
            for path in specific_files:
 
1657
                specific_files_utf8.add(path.encode('utf8'))
 
1658
            specific_files = specific_files_utf8
 
1659
        else:
 
1660
            specific_files = set([''])
 
1661
        # -- specific_files is now a utf8 path set --
 
1662
        # -- get the state object and prepare it.
 
1663
        state = self.target.current_dirstate()
 
1664
        state._read_dirblocks_if_needed()
 
1665
        def _entries_for_path(path):
 
1666
            """Return a list with all the entries that match path for all ids.
 
1667
            """
 
1668
            dirname, basename = os.path.split(path)
 
1669
            key = (dirname, basename, '')
 
1670
            block_index, present = state._find_block_index_from_key(key)
 
1671
            if not present:
 
1672
                # the block which should contain path is absent.
 
1673
                return []
 
1674
            result = []
 
1675
            block = state._dirblocks[block_index][1]
 
1676
            entry_index, _ = state._find_entry_index(key, block)
 
1677
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1678
            while (entry_index < len(block) and
 
1679
                block[entry_index][0][0:2] == key[0:2]):
 
1680
                result.append(block[entry_index])
 
1681
                entry_index += 1
 
1682
            return result
 
1683
        if require_versioned:
 
1684
            # -- check all supplied paths are versioned in a search tree. --
 
1685
            all_versioned = True
 
1686
            for path in specific_files:
 
1687
                path_entries = _entries_for_path(path)
 
1688
                if not path_entries:
 
1689
                    # this specified path is not present at all: error
 
1690
                    all_versioned = False
 
1691
                    break
 
1692
                found_versioned = False
 
1693
                # for each id at this path
 
1694
                for entry in path_entries:
 
1695
                    # for each tree.
 
1696
                    for index in indices:
 
1697
                        if entry[1][index][0] != 'a': # absent
 
1698
                            found_versioned = True
 
1699
                            # all good: found a versioned cell
 
1700
                            break
 
1701
                if not found_versioned:
 
1702
                    # none of the indexes was not 'absent' at all ids for this
 
1703
                    # path.
 
1704
                    all_versioned = False
 
1705
                    break
 
1706
            if not all_versioned:
 
1707
                raise errors.PathsNotVersionedError(specific_files)
 
1708
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1709
        search_specific_files = set()
 
1710
        for path in specific_files:
 
1711
            other_specific_files = specific_files.difference(set([path]))
 
1712
            if not osutils.is_inside_any(other_specific_files, path):
 
1713
                # this is a top level path, we must check it.
 
1714
                search_specific_files.add(path)
 
1715
        # sketch: 
 
1716
        # compare source_index and target_index at or under each element of search_specific_files.
 
1717
        # follow the following comparison table. Note that we only want to do diff operations when
 
1718
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1719
        # for the target.
 
1720
        # cases:
 
1721
        # 
 
1722
        # Source | Target | disk | action
 
1723
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1724
        #        |        |      | diff check on source-target
 
1725
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1726
        #        |        |      | ???
 
1727
        #   r    |  a     |      | add source to search
 
1728
        #   r    |  a     |  a   | 
 
1729
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1730
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1731
        #   a    | fdlt   |      | add new id
 
1732
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1733
        #   a    |  a     |      | not present in either tree, skip
 
1734
        #   a    |  a     |  a   | not present in any tree, skip
 
1735
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1736
        #        |        |      | may not be selected by the users list of paths.
 
1737
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1738
        #        |        |      | may not be selected by the users list of paths.
 
1739
        #  fdlt  | fdlt   |      | content in both: diff them
 
1740
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1741
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1742
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1743
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1744
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1745
        #        |        |      | this id at the other path.
 
1746
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1747
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1748
        #        |        |      | this id at the other path.
 
1749
 
 
1750
        # for all search_indexs in each path at or under each element of
 
1751
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1752
        # relocated path as one to search if its not searched already. If the
 
1753
        # detail is not relocated, add the id.
 
1754
        searched_specific_files = set()
 
1755
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1756
        # Using a list so that we can access the values and change them in
 
1757
        # nested scope. Each one is [path, file_id, entry]
 
1758
        last_source_parent = [None, None, None]
 
1759
        last_target_parent = [None, None, None]
 
1760
 
 
1761
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1762
 
 
1763
        def _process_entry(entry, path_info):
 
1764
            """Compare an entry and real disk to generate delta information.
 
1765
 
 
1766
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1767
                the path of entry. If None, then the path is considered absent.
 
1768
                (Perhaps we should pass in a concrete entry for this ?)
 
1769
                Basename is returned as a utf8 string because we expect this
 
1770
                tuple will be ignored, and don't want to take the time to
 
1771
                decode.
 
1772
            """
 
1773
            if source_index is None:
 
1774
                source_details = NULL_PARENT_DETAILS
 
1775
            else:
 
1776
                source_details = entry[1][source_index]
 
1777
            target_details = entry[1][target_index]
 
1778
            target_minikind = target_details[0]
 
1779
            if path_info is not None and target_minikind in 'fdlt':
 
1780
                assert target_index == 0
 
1781
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1782
                                                  stat_value=path_info[3])
 
1783
                # The entry may have been modified by update_entry
 
1784
                target_details = entry[1][target_index]
 
1785
                target_minikind = target_details[0]
 
1786
            else:
 
1787
                link_or_sha1 = None
 
1788
            source_minikind = source_details[0]
 
1789
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1790
                # claimed content in both: diff
 
1791
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1792
                #        |        |      | diff check on source-target
 
1793
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1794
                #        |        |      | ???
 
1795
                if source_minikind in 'r':
 
1796
                    # add the source to the search path to find any children it
 
1797
                    # has.  TODO ? : only add if it is a container ?
 
1798
                    if not osutils.is_inside_any(searched_specific_files,
 
1799
                                                 source_details[1]):
 
1800
                        search_specific_files.add(source_details[1])
 
1801
                    # generate the old path; this is needed for stating later
 
1802
                    # as well.
 
1803
                    old_path = source_details[1]
 
1804
                    old_dirname, old_basename = os.path.split(old_path)
 
1805
                    path = pathjoin(entry[0][0], entry[0][1])
 
1806
                    old_entry = state._get_entry(source_index,
 
1807
                                                 path_utf8=old_path)
 
1808
                    # update the source details variable to be the real
 
1809
                    # location.
 
1810
                    source_details = old_entry[1][source_index]
 
1811
                    source_minikind = source_details[0]
 
1812
                else:
 
1813
                    old_dirname = entry[0][0]
 
1814
                    old_basename = entry[0][1]
 
1815
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1816
                if path_info is None:
 
1817
                    # the file is missing on disk, show as removed.
 
1818
                    content_change = True
 
1819
                    target_kind = None
 
1820
                    target_exec = False
 
1821
                else:
 
1822
                    # source and target are both versioned and disk file is present.
 
1823
                    target_kind = path_info[2]
 
1824
                    if target_kind == 'directory':
 
1825
                        if source_minikind != 'd':
 
1826
                            content_change = True
 
1827
                        else:
 
1828
                            # directories have no fingerprint
 
1829
                            content_change = False
 
1830
                        target_exec = False
 
1831
                    elif target_kind == 'file':
 
1832
                        if source_minikind != 'f':
 
1833
                            content_change = True
 
1834
                        else:
 
1835
                            # We could check the size, but we already have the
 
1836
                            # sha1 hash.
 
1837
                            content_change = (link_or_sha1 != source_details[1])
 
1838
                        # Target details is updated at update_entry time
 
1839
                        if use_filesystem_for_exec:
 
1840
                            # We don't need S_ISREG here, because we are sure
 
1841
                            # we are dealing with a file.
 
1842
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1843
                        else:
 
1844
                            target_exec = target_details[3]
 
1845
                    elif target_kind == 'symlink':
 
1846
                        if source_minikind != 'l':
 
1847
                            content_change = True
 
1848
                        else:
 
1849
                            content_change = (link_or_sha1 != source_details[1])
 
1850
                        target_exec = False
 
1851
                    elif target_kind == 'tree-reference':
 
1852
                        if source_minikind != 't':
 
1853
                            content_change = True
 
1854
                        else:
 
1855
                            content_change = False
 
1856
                        target_exec = False
 
1857
                    else:
 
1858
                        raise Exception, "unknown kind %s" % path_info[2]
 
1859
                # parent id is the entry for the path in the target tree
 
1860
                if old_dirname == last_source_parent[0]:
 
1861
                    source_parent_id = last_source_parent[1]
 
1862
                else:
 
1863
                    source_parent_entry = state._get_entry(source_index,
 
1864
                                                           path_utf8=old_dirname)
 
1865
                    source_parent_id = source_parent_entry[0][2]
 
1866
                    if source_parent_id == entry[0][2]:
 
1867
                        # This is the root, so the parent is None
 
1868
                        source_parent_id = None
 
1869
                    else:
 
1870
                        last_source_parent[0] = old_dirname
 
1871
                        last_source_parent[1] = source_parent_id
 
1872
                        last_source_parent[2] = source_parent_entry
 
1873
                new_dirname = entry[0][0]
 
1874
                if new_dirname == last_target_parent[0]:
 
1875
                    target_parent_id = last_target_parent[1]
 
1876
                else:
 
1877
                    # TODO: We don't always need to do the lookup, because the
 
1878
                    #       parent entry will be the same as the source entry.
 
1879
                    target_parent_entry = state._get_entry(target_index,
 
1880
                                                           path_utf8=new_dirname)
 
1881
                    target_parent_id = target_parent_entry[0][2]
 
1882
                    if target_parent_id == entry[0][2]:
 
1883
                        # This is the root, so the parent is None
 
1884
                        target_parent_id = None
 
1885
                    else:
 
1886
                        last_target_parent[0] = new_dirname
 
1887
                        last_target_parent[1] = target_parent_id
 
1888
                        last_target_parent[2] = target_parent_entry
 
1889
 
 
1890
                source_exec = source_details[3]
 
1891
                return ((entry[0][2], (old_path, path), content_change,
 
1892
                        (True, True),
 
1893
                        (source_parent_id, target_parent_id),
 
1894
                        (old_basename, entry[0][1]),
 
1895
                        (_minikind_to_kind[source_minikind], target_kind),
 
1896
                        (source_exec, target_exec)),)
 
1897
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1898
                # looks like a new file
 
1899
                if path_info is not None:
 
1900
                    path = pathjoin(entry[0][0], entry[0][1])
 
1901
                    # parent id is the entry for the path in the target tree
 
1902
                    # TODO: these are the same for an entire directory: cache em.
 
1903
                    parent_id = state._get_entry(target_index,
 
1904
                                                 path_utf8=entry[0][0])[0][2]
 
1905
                    if parent_id == entry[0][2]:
 
1906
                        parent_id = None
 
1907
                    if use_filesystem_for_exec:
 
1908
                        # We need S_ISREG here, because we aren't sure if this
 
1909
                        # is a file or not.
 
1910
                        target_exec = bool(
 
1911
                            stat.S_ISREG(path_info[3].st_mode)
 
1912
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1913
                    else:
 
1914
                        target_exec = target_details[3]
 
1915
                    return ((entry[0][2], (None, path), True,
 
1916
                            (False, True),
 
1917
                            (None, parent_id),
 
1918
                            (None, entry[0][1]),
 
1919
                            (None, path_info[2]),
 
1920
                            (None, target_exec)),)
 
1921
                else:
 
1922
                    # but its not on disk: we deliberately treat this as just
 
1923
                    # never-present. (Why ?! - RBC 20070224)
 
1924
                    pass
 
1925
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1926
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1927
                # if its still on disk, *and* theres no other entry at this
 
1928
                # path [we dont know this in this routine at the moment -
 
1929
                # perhaps we should change this - then it would be an unknown.
 
1930
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1931
                # parent id is the entry for the path in the target tree
 
1932
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1933
                if parent_id == entry[0][2]:
 
1934
                    parent_id = None
 
1935
                return ((entry[0][2], (old_path, None), True,
 
1936
                        (True, False),
 
1937
                        (parent_id, None),
 
1938
                        (entry[0][1], None),
 
1939
                        (_minikind_to_kind[source_minikind], None),
 
1940
                        (source_details[3], None)),)
 
1941
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1942
                # a rename; could be a true rename, or a rename inherited from
 
1943
                # a renamed parent. TODO: handle this efficiently. Its not
 
1944
                # common case to rename dirs though, so a correct but slow
 
1945
                # implementation will do.
 
1946
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1947
                    search_specific_files.add(target_details[1])
 
1948
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
1949
                # neither of the selected trees contain this file,
 
1950
                # so skip over it. This is not currently directly tested, but
 
1951
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1952
                pass
 
1953
            else:
 
1954
                raise AssertionError("don't know how to compare "
 
1955
                    "source_minikind=%r, target_minikind=%r"
 
1956
                    % (source_minikind, target_minikind))
 
1957
                ## import pdb;pdb.set_trace()
 
1958
            return ()
 
1959
 
 
1960
        while search_specific_files:
 
1961
            # TODO: the pending list should be lexically sorted?  the
 
1962
            # interface doesn't require it.
 
1963
            current_root = search_specific_files.pop()
 
1964
            searched_specific_files.add(current_root)
 
1965
            # process the entries for this containing directory: the rest will be
 
1966
            # found by their parents recursively.
 
1967
            root_entries = _entries_for_path(current_root)
 
1968
            root_abspath = self.target.abspath(current_root)
 
1969
            try:
 
1970
                root_stat = os.lstat(root_abspath)
 
1971
            except OSError, e:
 
1972
                if e.errno == errno.ENOENT:
 
1973
                    # the path does not exist: let _process_entry know that.
 
1974
                    root_dir_info = None
 
1975
                else:
 
1976
                    # some other random error: hand it up.
 
1977
                    raise
 
1978
            else:
 
1979
                root_dir_info = ('', current_root,
 
1980
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1981
                    root_abspath)
 
1982
                if root_dir_info[2] == 'directory':
 
1983
                    if self.target._directory_is_tree_reference(
 
1984
                        current_root.decode('utf8')):
 
1985
                        root_dir_info = root_dir_info[:2] + \
 
1986
                            ('tree-reference',) + root_dir_info[3:]
 
1987
 
 
1988
            if not root_entries and not root_dir_info:
 
1989
                # this specified path is not present at all, skip it.
 
1990
                continue
 
1991
            path_handled = False
 
1992
            for entry in root_entries:
 
1993
                for result in _process_entry(entry, root_dir_info):
 
1994
                    # this check should probably be outside the loop: one
 
1995
                    # 'iterate two trees' api, and then _iter_changes filters
 
1996
                    # unchanged pairs. - RBC 20070226
 
1997
                    path_handled = True
 
1998
                    if (include_unchanged
 
1999
                        or result[2]                    # content change
 
2000
                        or result[3][0] != result[3][1] # versioned status
 
2001
                        or result[4][0] != result[4][1] # parent id
 
2002
                        or result[5][0] != result[5][1] # name
 
2003
                        or result[6][0] != result[6][1] # kind
 
2004
                        or result[7][0] != result[7][1] # executable
 
2005
                        ):
 
2006
                        result = (result[0],
 
2007
                            ((utf8_decode(result[1][0])[0]),
 
2008
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
2009
                        yield result
 
2010
            if want_unversioned and not path_handled:
 
2011
                new_executable = bool(
 
2012
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2013
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2014
                yield (None, (None, current_root), True, (False, False),
 
2015
                    (None, None),
 
2016
                    (None, splitpath(current_root)[-1]),
 
2017
                    (None, root_dir_info[2]), (None, new_executable))
 
2018
            initial_key = (current_root, '', '')
 
2019
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2020
            if block_index == 0:
 
2021
                # we have processed the total root already, but because the
 
2022
                # initial key matched it we should skip it here.
 
2023
                block_index +=1
 
2024
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2025
                current_dir_info = None
 
2026
            else:
 
2027
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2028
                try:
 
2029
                    current_dir_info = dir_iterator.next()
 
2030
                except OSError, e:
 
2031
                    if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
2032
                        # there may be directories in the inventory even though
 
2033
                        # this path is not a file on disk: so mark it as end of
 
2034
                        # iterator
 
2035
                        current_dir_info = None
 
2036
                    else:
 
2037
                        raise
 
2038
                else:
 
2039
                    if current_dir_info[0][0] == '':
 
2040
                        # remove .bzr from iteration
 
2041
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2042
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2043
                        del current_dir_info[1][bzr_index]
 
2044
            # walk until both the directory listing and the versioned metadata
 
2045
            # are exhausted. 
 
2046
            if (block_index < len(state._dirblocks) and
 
2047
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2048
                current_block = state._dirblocks[block_index]
 
2049
            else:
 
2050
                current_block = None
 
2051
            while (current_dir_info is not None or
 
2052
                   current_block is not None):
 
2053
                if (current_dir_info and current_block
 
2054
                    and current_dir_info[0][0] != current_block[0]):
 
2055
                    if current_dir_info[0][0] < current_block[0] :
 
2056
                        # filesystem data refers to paths not covered by the dirblock.
 
2057
                        # this has two possibilities:
 
2058
                        # A) it is versioned but empty, so there is no block for it
 
2059
                        # B) it is not versioned.
 
2060
                        # in either case it was processed by the containing directories walk:
 
2061
                        # if it is root/foo, when we walked root we emitted it,
 
2062
                        # or if we ere given root/foo to walk specifically, we
 
2063
                        # emitted it when checking the walk-root entries
 
2064
                        # advance the iterator and loop - we dont need to emit it.
 
2065
                        try:
 
2066
                            current_dir_info = dir_iterator.next()
 
2067
                        except StopIteration:
 
2068
                            current_dir_info = None
 
2069
                    else:
 
2070
                        # We have a dirblock entry for this location, but there
 
2071
                        # is no filesystem path for this. This is most likely
 
2072
                        # because a directory was removed from the disk.
 
2073
                        # We don't have to report the missing directory,
 
2074
                        # because that should have already been handled, but we
 
2075
                        # need to handle all of the files that are contained
 
2076
                        # within.
 
2077
                        for current_entry in current_block[1]:
 
2078
                            # entry referring to file not present on disk.
 
2079
                            # advance the entry only, after processing.
 
2080
                            for result in _process_entry(current_entry, None):
 
2081
                                # this check should probably be outside the loop: one
 
2082
                                # 'iterate two trees' api, and then _iter_changes filters
 
2083
                                # unchanged pairs. - RBC 20070226
 
2084
                                if (include_unchanged
 
2085
                                    or result[2]                    # content change
 
2086
                                    or result[3][0] != result[3][1] # versioned status
 
2087
                                    or result[4][0] != result[4][1] # parent id
 
2088
                                    or result[5][0] != result[5][1] # name
 
2089
                                    or result[6][0] != result[6][1] # kind
 
2090
                                    or result[7][0] != result[7][1] # executable
 
2091
                                    ):
 
2092
                                    result = (result[0],
 
2093
                                        ((utf8_decode(result[1][0])[0]),
 
2094
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2095
                                    yield result
 
2096
                        block_index +=1
 
2097
                        if (block_index < len(state._dirblocks) and
 
2098
                            osutils.is_inside(current_root,
 
2099
                                              state._dirblocks[block_index][0])):
 
2100
                            current_block = state._dirblocks[block_index]
 
2101
                        else:
 
2102
                            current_block = None
 
2103
                    continue
 
2104
                entry_index = 0
 
2105
                if current_block and entry_index < len(current_block[1]):
 
2106
                    current_entry = current_block[1][entry_index]
 
2107
                else:
 
2108
                    current_entry = None
 
2109
                advance_entry = True
 
2110
                path_index = 0
 
2111
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2112
                    current_path_info = current_dir_info[1][path_index]
 
2113
                    if current_path_info[2] == 'directory':
 
2114
                        if self.target._directory_is_tree_reference(
 
2115
                            current_path_info[0].decode('utf8')):
 
2116
                            current_path_info = current_path_info[:2] + \
 
2117
                                ('tree-reference',) + current_path_info[3:]
 
2118
                else:
 
2119
                    current_path_info = None
 
2120
                advance_path = True
 
2121
                path_handled = False
 
2122
                while (current_entry is not None or
 
2123
                    current_path_info is not None):
 
2124
                    if current_entry is None:
 
2125
                        # the check for path_handled when the path is adnvaced
 
2126
                        # will yield this path if needed.
 
2127
                        pass
 
2128
                    elif current_path_info is None:
 
2129
                        # no path is fine: the per entry code will handle it.
 
2130
                        for result in _process_entry(current_entry, current_path_info):
 
2131
                            # this check should probably be outside the loop: one
 
2132
                            # 'iterate two trees' api, and then _iter_changes filters
 
2133
                            # unchanged pairs. - RBC 20070226
 
2134
                            if (include_unchanged
 
2135
                                or result[2]                    # content change
 
2136
                                or result[3][0] != result[3][1] # versioned status
 
2137
                                or result[4][0] != result[4][1] # parent id
 
2138
                                or result[5][0] != result[5][1] # name
 
2139
                                or result[6][0] != result[6][1] # kind
 
2140
                                or result[7][0] != result[7][1] # executable
 
2141
                                ):
 
2142
                                result = (result[0],
 
2143
                                    ((utf8_decode(result[1][0])[0]),
 
2144
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2145
                                yield result
 
2146
                    elif current_entry[0][1] != current_path_info[1]:
 
2147
                        if current_path_info[1] < current_entry[0][1]:
 
2148
                            # extra file on disk: pass for now, but only
 
2149
                            # increment the path, not the entry
 
2150
                            advance_entry = False
 
2151
                        else:
 
2152
                            # entry referring to file not present on disk.
 
2153
                            # advance the entry only, after processing.
 
2154
                            for result in _process_entry(current_entry, None):
 
2155
                                # this check should probably be outside the loop: one
 
2156
                                # 'iterate two trees' api, and then _iter_changes filters
 
2157
                                # unchanged pairs. - RBC 20070226
 
2158
                                path_handled = True
 
2159
                                if (include_unchanged
 
2160
                                    or result[2]                    # content change
 
2161
                                    or result[3][0] != result[3][1] # versioned status
 
2162
                                    or result[4][0] != result[4][1] # parent id
 
2163
                                    or result[5][0] != result[5][1] # name
 
2164
                                    or result[6][0] != result[6][1] # kind
 
2165
                                    or result[7][0] != result[7][1] # executable
 
2166
                                    ):
 
2167
                                    result = (result[0],
 
2168
                                        ((utf8_decode(result[1][0])[0]),
 
2169
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2170
                                    yield result
 
2171
                            advance_path = False
 
2172
                    else:
 
2173
                        for result in _process_entry(current_entry, current_path_info):
 
2174
                            # this check should probably be outside the loop: one
 
2175
                            # 'iterate two trees' api, and then _iter_changes filters
 
2176
                            # unchanged pairs. - RBC 20070226
 
2177
                            path_handled = True
 
2178
                            if (include_unchanged
 
2179
                                or result[2]                    # content change
 
2180
                                or result[3][0] != result[3][1] # versioned status
 
2181
                                or result[4][0] != result[4][1] # parent id
 
2182
                                or result[5][0] != result[5][1] # name
 
2183
                                or result[6][0] != result[6][1] # kind
 
2184
                                or result[7][0] != result[7][1] # executable
 
2185
                                ):
 
2186
                                result = (result[0],
 
2187
                                    ((utf8_decode(result[1][0])[0]),
 
2188
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2189
                                yield result
 
2190
                    if advance_entry and current_entry is not None:
 
2191
                        entry_index += 1
 
2192
                        if entry_index < len(current_block[1]):
 
2193
                            current_entry = current_block[1][entry_index]
 
2194
                        else:
 
2195
                            current_entry = None
 
2196
                    else:
 
2197
                        advance_entry = True # reset the advance flaga
 
2198
                    if advance_path and current_path_info is not None:
 
2199
                        if not path_handled:
 
2200
                            # unversioned in all regards
 
2201
                            if want_unversioned:
 
2202
                                new_executable = bool(
 
2203
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2204
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2205
                                if want_unversioned:
 
2206
                                    yield (None, (None, current_path_info[0]),
 
2207
                                        True,
 
2208
                                        (False, False),
 
2209
                                        (None, None),
 
2210
                                        (None, current_path_info[1]),
 
2211
                                        (None, current_path_info[2]),
 
2212
                                        (None, new_executable))
 
2213
                            # dont descend into this unversioned path if it is
 
2214
                            # a dir
 
2215
                            if current_path_info[2] in ('directory'):
 
2216
                                del current_dir_info[1][path_index]
 
2217
                                path_index -= 1
 
2218
                        # dont descend the disk iterator into any tree 
 
2219
                        # paths.
 
2220
                        if current_path_info[2] == 'tree-reference':
 
2221
                            del current_dir_info[1][path_index]
 
2222
                            path_index -= 1
 
2223
                        path_index += 1
 
2224
                        if path_index < len(current_dir_info[1]):
 
2225
                            current_path_info = current_dir_info[1][path_index]
 
2226
                            if current_path_info[2] == 'directory':
 
2227
                                if self.target._directory_is_tree_reference(
 
2228
                                    current_path_info[0].decode('utf8')):
 
2229
                                    current_path_info = current_path_info[:2] + \
 
2230
                                        ('tree-reference',) + current_path_info[3:]
 
2231
                        else:
 
2232
                            current_path_info = None
 
2233
                        path_handled = False
 
2234
                    else:
 
2235
                        advance_path = True # reset the advance flagg.
 
2236
                if current_block is not None:
 
2237
                    block_index += 1
 
2238
                    if (block_index < len(state._dirblocks) and
 
2239
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2240
                        current_block = state._dirblocks[block_index]
 
2241
                    else:
 
2242
                        current_block = None
 
2243
                if current_dir_info is not None:
 
2244
                    try:
 
2245
                        current_dir_info = dir_iterator.next()
 
2246
                    except StopIteration:
 
2247
                        current_dir_info = None
 
2248
 
 
2249
 
 
2250
    @staticmethod
 
2251
    def is_compatible(source, target):
 
2252
        # the target must be a dirstate working tree
 
2253
        if not isinstance(target, WorkingTree4):
 
2254
            return False
 
2255
        # the source must be a revtreee or dirstate rev tree.
 
2256
        if not isinstance(source,
 
2257
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2258
            return False
 
2259
        # the source revid must be in the target dirstate
 
2260
        if not (source._revision_id == NULL_REVISION or
 
2261
            source._revision_id in target.get_parent_ids()):
 
2262
            # TODO: what about ghosts? it may well need to 
 
2263
            # check for them explicitly.
 
2264
            return False
 
2265
        return True
 
2266
 
 
2267
InterTree.register_optimiser(InterDirStateTree)
 
2268
 
 
2269
 
 
2270
class Converter3to4(object):
 
2271
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2272
 
 
2273
    def __init__(self):
 
2274
        self.target_format = WorkingTreeFormat4()
 
2275
 
 
2276
    def convert(self, tree):
 
2277
        # lock the control files not the tree, so that we dont get tree
 
2278
        # on-unlock behaviours, and so that noone else diddles with the 
 
2279
        # tree during upgrade.
 
2280
        tree._control_files.lock_write()
 
2281
        try:
 
2282
            self.create_dirstate_data(tree)
 
2283
            self.update_format(tree)
 
2284
            self.remove_xml_files(tree)
 
2285
        finally:
 
2286
            tree._control_files.unlock()
 
2287
 
 
2288
    def create_dirstate_data(self, tree):
 
2289
        """Create the dirstate based data for tree."""
 
2290
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2291
            ).local_abspath('dirstate')
 
2292
        state = dirstate.DirState.from_tree(tree, local_path)
 
2293
        state.save()
 
2294
        state.unlock()
 
2295
 
 
2296
    def remove_xml_files(self, tree):
 
2297
        """Remove the oldformat 3 data."""
 
2298
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2299
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2300
            'pending-merges', 'stat-cache']:
 
2301
            try:
 
2302
                transport.delete(path)
 
2303
            except errors.NoSuchFile:
 
2304
                # some files are optional - just deal.
 
2305
                pass
 
2306
 
 
2307
    def update_format(self, tree):
 
2308
        """Change the format marker."""
 
2309
        tree._control_files.put_utf8('format',
 
2310
            self.target_format.get_format_string())