~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: John Arbash Meinel
  • Date: 2007-11-13 20:37:09 UTC
  • mto: This revision was merged to the branch mainline in revision 3001.
  • Revision ID: john@arbash-meinel.com-20071113203709-kysdte0emqv84pnj
Fix bug #162486, by having RemoteBranch properly initialize self._revision_id_to_revno_map.

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
    debug,
 
47
    delta,
 
48
    dirstate,
 
49
    errors,
 
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    trace,
 
59
    transform,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
 
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
 
67
""")
 
68
 
 
69
from bzrlib import symbol_versioning
 
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
 
76
from bzrlib.osutils import (
 
77
    file_kind,
 
78
    isdir,
 
79
    normpath,
 
80
    pathjoin,
 
81
    rand_chars,
 
82
    realpath,
 
83
    safe_unicode,
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
 
87
from bzrlib.transport.local import LocalTransport
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
 
120
    def __init__(self, basedir,
 
121
                 branch,
 
122
                 _control_files=None,
 
123
                 _format=None,
 
124
                 _bzrdir=None):
 
125
        """Construct a WorkingTree for basedir.
 
126
 
 
127
        If the branch is not supplied, it is opened automatically.
 
128
        If the branch is supplied, it must be the branch for this basedir.
 
129
        (branch.base is not cross checked, because for remote branches that
 
130
        would be meaningless).
 
131
        """
 
132
        self._format = _format
 
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
 
136
        basedir = safe_unicode(basedir)
 
137
        mutter("opening working tree %r", basedir)
 
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
 
141
        self.basedir = realpath(basedir)
 
142
        # if branch is at our basedir and is a format 6 or less
 
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
 
146
        self._control_files = _control_files
 
147
        self._dirty = None
 
148
        #-------------
 
149
        # during a read or write lock these objects are set, and are
 
150
        # None the rest of the time.
 
151
        self._dirstate = None
 
152
        self._inventory = None
 
153
        #-------------
 
154
 
 
155
    @needs_tree_write_lock
 
156
    def _add(self, files, ids, kinds):
 
157
        """See MutableTree._add."""
 
158
        state = self.current_dirstate()
 
159
        for f, file_id, kind in zip(files, ids, kinds):
 
160
            f = f.strip('/')
 
161
            if self.path2id(f):
 
162
                # special case tree root handling.
 
163
                if f == '' and self.path2id(f) == ROOT_ID:
 
164
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
165
                continue
 
166
            if file_id is None:
 
167
                file_id = generate_ids.gen_file_id(f)
 
168
            # deliberately add the file with no cached stat or sha1
 
169
            # - on the first access it will be gathered, and we can
 
170
            # always change this once tests are all passing.
 
171
            state.add(f, file_id, kind, None, '')
 
172
        self._make_dirty(reset_inventory=True)
 
173
 
 
174
    def _make_dirty(self, reset_inventory):
 
175
        """Make the tree state dirty.
 
176
 
 
177
        :param reset_inventory: True if the cached inventory should be removed
 
178
            (presuming there is one).
 
179
        """
 
180
        self._dirty = True
 
181
        if reset_inventory and self._inventory is not None:
 
182
            self._inventory = None
 
183
 
 
184
    @needs_tree_write_lock
 
185
    def add_reference(self, sub_tree):
 
186
        # use standard implementation, which calls back to self._add
 
187
        # 
 
188
        # So we don't store the reference_revision in the working dirstate,
 
189
        # it's just recorded at the moment of commit. 
 
190
        self._add_reference(sub_tree)
 
191
 
 
192
    def break_lock(self):
 
193
        """Break a lock if one is present from another instance.
 
194
 
 
195
        Uses the ui factory to ask for confirmation if the lock may be from
 
196
        an active process.
 
197
 
 
198
        This will probe the repository for its lock as well.
 
199
        """
 
200
        # if the dirstate is locked by an active process, reject the break lock
 
201
        # call.
 
202
        try:
 
203
            if self._dirstate is None:
 
204
                clear = True
 
205
            else:
 
206
                clear = False
 
207
            state = self._current_dirstate()
 
208
            if state._lock_token is not None:
 
209
                # we already have it locked. sheese, cant break our own lock.
 
210
                raise errors.LockActive(self.basedir)
 
211
            else:
 
212
                try:
 
213
                    # try for a write lock - need permission to get one anyhow
 
214
                    # to break locks.
 
215
                    state.lock_write()
 
216
                except errors.LockContention:
 
217
                    # oslocks fail when a process is still live: fail.
 
218
                    # TODO: get the locked lockdir info and give to the user to
 
219
                    # assist in debugging.
 
220
                    raise errors.LockActive(self.basedir)
 
221
                else:
 
222
                    state.unlock()
 
223
        finally:
 
224
            if clear:
 
225
                self._dirstate = None
 
226
        self._control_files.break_lock()
 
227
        self.branch.break_lock()
 
228
 
 
229
    def _comparison_data(self, entry, path):
 
230
        kind, executable, stat_value = \
 
231
            WorkingTree3._comparison_data(self, entry, path)
 
232
        # it looks like a plain directory, but it's really a reference -- see
 
233
        # also kind()
 
234
        if (self._repo_supports_tree_reference and
 
235
            kind == 'directory' and
 
236
            self._directory_is_tree_reference(path)):
 
237
            kind = 'tree-reference'
 
238
        return kind, executable, stat_value
 
239
 
 
240
    @needs_write_lock
 
241
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
242
        # mark the tree as dirty post commit - commit
 
243
        # can change the current versioned list by doing deletes.
 
244
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
245
        self._make_dirty(reset_inventory=True)
 
246
        return result
 
247
 
 
248
    def current_dirstate(self):
 
249
        """Return the current dirstate object.
 
250
 
 
251
        This is not part of the tree interface and only exposed for ease of
 
252
        testing.
 
253
 
 
254
        :raises errors.NotWriteLocked: when not in a lock.
 
255
        """
 
256
        self._must_be_locked()
 
257
        return self._current_dirstate()
 
258
 
 
259
    def _current_dirstate(self):
 
260
        """Internal function that does not check lock status.
 
261
 
 
262
        This is needed for break_lock which also needs the dirstate.
 
263
        """
 
264
        if self._dirstate is not None:
 
265
            return self._dirstate
 
266
        local_path = self.bzrdir.get_workingtree_transport(None
 
267
            ).local_abspath('dirstate')
 
268
        self._dirstate = dirstate.DirState.on_file(local_path)
 
269
        return self._dirstate
 
270
 
 
271
    def filter_unversioned_files(self, paths):
 
272
        """Filter out paths that are versioned.
 
273
 
 
274
        :return: set of paths.
 
275
        """
 
276
        # TODO: make a generic multi-bisect routine roughly that should list
 
277
        # the paths, then process one half at a time recursively, and feed the
 
278
        # results of each bisect in further still
 
279
        paths = sorted(paths)
 
280
        result = set()
 
281
        state = self.current_dirstate()
 
282
        # TODO we want a paths_to_dirblocks helper I think
 
283
        for path in paths:
 
284
            dirname, basename = os.path.split(path.encode('utf8'))
 
285
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
286
                dirname, basename, 0)
 
287
            if not path_is_versioned:
 
288
                result.add(path)
 
289
        return result
 
290
 
 
291
    def flush(self):
 
292
        """Write all cached data to disk."""
 
293
        if self._control_files._lock_mode != 'w':
 
294
            raise errors.NotWriteLocked(self)
 
295
        self.current_dirstate().save()
 
296
        self._inventory = None
 
297
        self._dirty = False
 
298
 
 
299
    @needs_tree_write_lock
 
300
    def _gather_kinds(self, files, kinds):
 
301
        """See MutableTree._gather_kinds."""
 
302
        for pos, f in enumerate(files):
 
303
            if kinds[pos] is None:
 
304
                kinds[pos] = self._kind(f)
 
305
 
 
306
    def _generate_inventory(self):
 
307
        """Create and set self.inventory from the dirstate object.
 
308
        
 
309
        This is relatively expensive: we have to walk the entire dirstate.
 
310
        Ideally we would not, and can deprecate this function.
 
311
        """
 
312
        #: uncomment to trap on inventory requests.
 
313
        # import pdb;pdb.set_trace()
 
314
        state = self.current_dirstate()
 
315
        state._read_dirblocks_if_needed()
 
316
        root_key, current_entry = self._get_entry(path='')
 
317
        current_id = root_key[2]
 
318
        assert current_entry[0][0] == 'd' # directory
 
319
        inv = Inventory(root_id=current_id)
 
320
        # Turn some things into local variables
 
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
322
        factory = entry_factory
 
323
        utf8_decode = cache_utf8._utf8_decode
 
324
        inv_byid = inv._byid
 
325
        # we could do this straight out of the dirstate; it might be fast
 
326
        # and should be profiled - RBC 20070216
 
327
        parent_ies = {'' : inv.root}
 
328
        for block in state._dirblocks[1:]: # skip the root
 
329
            dirname = block[0]
 
330
            try:
 
331
                parent_ie = parent_ies[dirname]
 
332
            except KeyError:
 
333
                # all the paths in this block are not versioned in this tree
 
334
                continue
 
335
            for key, entry in block[1]:
 
336
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
337
                if minikind in ('a', 'r'): # absent, relocated
 
338
                    # a parent tree only entry
 
339
                    continue
 
340
                name = key[1]
 
341
                name_unicode = utf8_decode(name)[0]
 
342
                file_id = key[2]
 
343
                kind = minikind_to_kind[minikind]
 
344
                inv_entry = factory[kind](file_id, name_unicode,
 
345
                                          parent_ie.file_id)
 
346
                if kind == 'file':
 
347
                    # This is only needed on win32, where this is the only way
 
348
                    # we know the executable bit.
 
349
                    inv_entry.executable = executable
 
350
                    # not strictly needed: working tree
 
351
                    #inv_entry.text_size = size
 
352
                    #inv_entry.text_sha1 = sha1
 
353
                elif kind == 'directory':
 
354
                    # add this entry to the parent map.
 
355
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
356
                elif kind == 'tree-reference':
 
357
                    assert self._repo_supports_tree_reference, \
 
358
                        "repository of %r " \
 
359
                        "doesn't support tree references " \
 
360
                        "required by entry %r" \
 
361
                        % (self, name)
 
362
                    inv_entry.reference_revision = link_or_sha1 or None
 
363
                elif kind != 'symlink':
 
364
                    raise AssertionError("unknown kind %r" % kind)
 
365
                # These checks cost us around 40ms on a 55k entry tree
 
366
                assert file_id not in inv_byid, ('file_id %s already in'
 
367
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                assert name_unicode not in parent_ie.children
 
369
                inv_byid[file_id] = inv_entry
 
370
                parent_ie.children[name_unicode] = inv_entry
 
371
        self._inventory = inv
 
372
 
 
373
    def _get_entry(self, file_id=None, path=None):
 
374
        """Get the dirstate row for file_id or path.
 
375
 
 
376
        If either file_id or path is supplied, it is used as the key to lookup.
 
377
        If both are supplied, the fastest lookup is used, and an error is
 
378
        raised if they do not both point at the same row.
 
379
        
 
380
        :param file_id: An optional unicode file_id to be looked up.
 
381
        :param path: An optional unicode path to be looked up.
 
382
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
383
        """
 
384
        if file_id is None and path is None:
 
385
            raise errors.BzrError('must supply file_id or path')
 
386
        state = self.current_dirstate()
 
387
        if path is not None:
 
388
            path = path.encode('utf8')
 
389
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
390
 
 
391
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
392
        # check file id is valid unconditionally.
 
393
        entry = self._get_entry(file_id=file_id, path=path)
 
394
        if entry[0] is None:
 
395
            raise errors.NoSuchId(self, file_id)
 
396
        if path is None:
 
397
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
398
 
 
399
        file_abspath = self.abspath(path)
 
400
        state = self.current_dirstate()
 
401
        if stat_value is None:
 
402
            try:
 
403
                stat_value = os.lstat(file_abspath)
 
404
            except OSError, e:
 
405
                if e.errno == errno.ENOENT:
 
406
                    return None
 
407
                else:
 
408
                    raise
 
409
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
410
                                          stat_value=stat_value)
 
411
        if entry[1][0][0] == 'f':
 
412
            return link_or_sha1
 
413
        return None
 
414
 
 
415
    def _get_inventory(self):
 
416
        """Get the inventory for the tree. This is only valid within a lock."""
 
417
        if 'evil' in debug.debug_flags:
 
418
            trace.mutter_callsite(2,
 
419
                "accessing .inventory forces a size of tree translation.")
 
420
        if self._inventory is not None:
 
421
            return self._inventory
 
422
        self._must_be_locked()
 
423
        self._generate_inventory()
 
424
        return self._inventory
 
425
 
 
426
    inventory = property(_get_inventory,
 
427
                         doc="Inventory of this Tree")
 
428
 
 
429
    @needs_read_lock
 
430
    def get_parent_ids(self):
 
431
        """See Tree.get_parent_ids.
 
432
        
 
433
        This implementation requests the ids list from the dirstate file.
 
434
        """
 
435
        return self.current_dirstate().get_parent_ids()
 
436
 
 
437
    def get_reference_revision(self, file_id, path=None):
 
438
        # referenced tree's revision is whatever's currently there
 
439
        return self.get_nested_tree(file_id, path).last_revision()
 
440
 
 
441
    def get_nested_tree(self, file_id, path=None):
 
442
        if path is None:
 
443
            path = self.id2path(file_id)
 
444
        # else: check file_id is at path?
 
445
        return WorkingTree.open(self.abspath(path))
 
446
 
 
447
    @needs_read_lock
 
448
    def get_root_id(self):
 
449
        """Return the id of this trees root"""
 
450
        return self._get_entry(path='')[0][2]
 
451
 
 
452
    def has_id(self, file_id):
 
453
        state = self.current_dirstate()
 
454
        row, parents = self._get_entry(file_id=file_id)
 
455
        if row is None:
 
456
            return False
 
457
        return osutils.lexists(pathjoin(
 
458
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
459
 
 
460
    @needs_read_lock
 
461
    def id2path(self, file_id):
 
462
        "Convert a file-id to a path."
 
463
        state = self.current_dirstate()
 
464
        entry = self._get_entry(file_id=file_id)
 
465
        if entry == (None, None):
 
466
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
467
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
468
        return path_utf8.decode('utf8')
 
469
 
 
470
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
471
        entry = self._get_entry(path=path)
 
472
        if entry == (None, None):
 
473
            return False # Missing entries are not executable
 
474
        return entry[1][0][3] # Executable?
 
475
 
 
476
    if not osutils.supports_executable():
 
477
        def is_executable(self, file_id, path=None):
 
478
            """Test if a file is executable or not.
 
479
 
 
480
            Note: The caller is expected to take a read-lock before calling this.
 
481
            """
 
482
            entry = self._get_entry(file_id=file_id, path=path)
 
483
            if entry == (None, None):
 
484
                return False
 
485
            return entry[1][0][3]
 
486
 
 
487
        _is_executable_from_path_and_stat = \
 
488
            _is_executable_from_path_and_stat_from_basis
 
489
    else:
 
490
        def is_executable(self, file_id, path=None):
 
491
            """Test if a file is executable or not.
 
492
 
 
493
            Note: The caller is expected to take a read-lock before calling this.
 
494
            """
 
495
            if not path:
 
496
                path = self.id2path(file_id)
 
497
            mode = os.lstat(self.abspath(path)).st_mode
 
498
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
499
 
 
500
    @needs_read_lock
 
501
    def __iter__(self):
 
502
        """Iterate through file_ids for this tree.
 
503
 
 
504
        file_ids are in a WorkingTree if they are in the working inventory
 
505
        and the working file exists.
 
506
        """
 
507
        result = []
 
508
        for key, tree_details in self.current_dirstate()._iter_entries():
 
509
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
510
                # not relevant to the working tree
 
511
                continue
 
512
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
513
            if osutils.lexists(path):
 
514
                result.append(key[2])
 
515
        return iter(result)
 
516
 
 
517
    def iter_references(self):
 
518
        for key, tree_details in self.current_dirstate()._iter_entries():
 
519
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
520
                # not relevant to the working tree
 
521
                continue
 
522
            if not key[1]:
 
523
                # the root is not a reference.
 
524
                continue
 
525
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
526
            try:
 
527
                if self._kind(path) == 'tree-reference':
 
528
                    yield path, key[2]
 
529
            except errors.NoSuchFile:
 
530
                # path is missing on disk.
 
531
                continue
 
532
 
 
533
    def kind(self, file_id):
 
534
        """Return the kind of a file.
 
535
 
 
536
        This is always the actual kind that's on disk, regardless of what it
 
537
        was added as.
 
538
 
 
539
        Note: The caller is expected to take a read-lock before calling this.
 
540
        """
 
541
        relpath = self.id2path(file_id)
 
542
        assert relpath != None, \
 
543
            "path for id {%s} is None!" % file_id
 
544
        return self._kind(relpath)
 
545
 
 
546
    def _kind(self, relpath):
 
547
        abspath = self.abspath(relpath)
 
548
        kind = file_kind(abspath)
 
549
        if (self._repo_supports_tree_reference and
 
550
            kind == 'directory' and
 
551
            self._directory_is_tree_reference(relpath)):
 
552
            kind = 'tree-reference'
 
553
        return kind
 
554
 
 
555
    @needs_read_lock
 
556
    def _last_revision(self):
 
557
        """See Mutable.last_revision."""
 
558
        parent_ids = self.current_dirstate().get_parent_ids()
 
559
        if parent_ids:
 
560
            return parent_ids[0]
 
561
        else:
 
562
            return _mod_revision.NULL_REVISION
 
563
 
 
564
    def lock_read(self):
 
565
        """See Branch.lock_read, and WorkingTree.unlock."""
 
566
        self.branch.lock_read()
 
567
        try:
 
568
            self._control_files.lock_read()
 
569
            try:
 
570
                state = self.current_dirstate()
 
571
                if not state._lock_token:
 
572
                    state.lock_read()
 
573
                # set our support for tree references from the repository in
 
574
                # use.
 
575
                self._repo_supports_tree_reference = getattr(
 
576
                    self.branch.repository._format, "supports_tree_reference",
 
577
                    False)
 
578
            except:
 
579
                self._control_files.unlock()
 
580
                raise
 
581
        except:
 
582
            self.branch.unlock()
 
583
            raise
 
584
 
 
585
    def _lock_self_write(self):
 
586
        """This should be called after the branch is locked."""
 
587
        try:
 
588
            self._control_files.lock_write()
 
589
            try:
 
590
                state = self.current_dirstate()
 
591
                if not state._lock_token:
 
592
                    state.lock_write()
 
593
                # set our support for tree references from the repository in
 
594
                # use.
 
595
                self._repo_supports_tree_reference = getattr(
 
596
                    self.branch.repository._format, "supports_tree_reference",
 
597
                    False)
 
598
            except:
 
599
                self._control_files.unlock()
 
600
                raise
 
601
        except:
 
602
            self.branch.unlock()
 
603
            raise
 
604
 
 
605
    def lock_tree_write(self):
 
606
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
607
        self.branch.lock_read()
 
608
        self._lock_self_write()
 
609
 
 
610
    def lock_write(self):
 
611
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
612
        self.branch.lock_write()
 
613
        self._lock_self_write()
 
614
 
 
615
    @needs_tree_write_lock
 
616
    def move(self, from_paths, to_dir, after=False):
 
617
        """See WorkingTree.move()."""
 
618
        result = []
 
619
        if not from_paths:
 
620
            return result
 
621
 
 
622
        state = self.current_dirstate()
 
623
 
 
624
        assert not isinstance(from_paths, basestring)
 
625
        to_dir_utf8 = to_dir.encode('utf8')
 
626
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
627
        id_index = state._get_id_index()
 
628
        # check destination directory
 
629
        # get the details for it
 
630
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
631
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
632
        if not entry_present:
 
633
            raise errors.BzrMoveFailedError('', to_dir,
 
634
                errors.NotVersionedError(to_dir))
 
635
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
636
        # get a handle on the block itself.
 
637
        to_block_index = state._ensure_block(
 
638
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
639
        to_block = state._dirblocks[to_block_index]
 
640
        to_abs = self.abspath(to_dir)
 
641
        if not isdir(to_abs):
 
642
            raise errors.BzrMoveFailedError('',to_dir,
 
643
                errors.NotADirectory(to_abs))
 
644
 
 
645
        if to_entry[1][0][0] != 'd':
 
646
            raise errors.BzrMoveFailedError('',to_dir,
 
647
                errors.NotADirectory(to_abs))
 
648
 
 
649
        if self._inventory is not None:
 
650
            update_inventory = True
 
651
            inv = self.inventory
 
652
            to_dir_ie = inv[to_dir_id]
 
653
            to_dir_id = to_entry[0][2]
 
654
        else:
 
655
            update_inventory = False
 
656
 
 
657
        rollbacks = []
 
658
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
659
                     fingerprint, packed_stat, size,
 
660
                     to_block, to_key, to_path_utf8):
 
661
            state._make_absent(old_entry)
 
662
            from_key = old_entry[0]
 
663
            rollbacks.append(
 
664
                lambda:state.update_minimal(from_key,
 
665
                    minikind,
 
666
                    executable=executable,
 
667
                    fingerprint=fingerprint,
 
668
                    packed_stat=packed_stat,
 
669
                    size=size,
 
670
                    path_utf8=from_path_utf8))
 
671
            state.update_minimal(to_key,
 
672
                    minikind,
 
673
                    executable=executable,
 
674
                    fingerprint=fingerprint,
 
675
                    packed_stat=packed_stat,
 
676
                    size=size,
 
677
                    path_utf8=to_path_utf8)
 
678
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
679
            new_entry = to_block[1][added_entry_index]
 
680
            rollbacks.append(lambda:state._make_absent(new_entry))
 
681
 
 
682
        for from_rel in from_paths:
 
683
            # from_rel is 'pathinroot/foo/bar'
 
684
            from_rel_utf8 = from_rel.encode('utf8')
 
685
            from_dirname, from_tail = osutils.split(from_rel)
 
686
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
687
            from_entry = self._get_entry(path=from_rel)
 
688
            if from_entry == (None, None):
 
689
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
690
                    errors.NotVersionedError(path=str(from_rel)))
 
691
 
 
692
            from_id = from_entry[0][2]
 
693
            to_rel = pathjoin(to_dir, from_tail)
 
694
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
695
            item_to_entry = self._get_entry(path=to_rel)
 
696
            if item_to_entry != (None, None):
 
697
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
698
                    "Target is already versioned.")
 
699
 
 
700
            if from_rel == to_rel:
 
701
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
702
                    "Source and target are identical.")
 
703
 
 
704
            from_missing = not self.has_filename(from_rel)
 
705
            to_missing = not self.has_filename(to_rel)
 
706
            if after:
 
707
                move_file = False
 
708
            else:
 
709
                move_file = True
 
710
            if to_missing:
 
711
                if not move_file:
 
712
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
713
                        errors.NoSuchFile(path=to_rel,
 
714
                        extra="New file has not been created yet"))
 
715
                elif from_missing:
 
716
                    # neither path exists
 
717
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
718
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
719
            else:
 
720
                if from_missing: # implicitly just update our path mapping
 
721
                    move_file = False
 
722
                elif not after:
 
723
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
724
                        extra="(Use --after to update the Bazaar id)")
 
725
 
 
726
            rollbacks = []
 
727
            def rollback_rename():
 
728
                """A single rename has failed, roll it back."""
 
729
                # roll back everything, even if we encounter trouble doing one
 
730
                # of them.
 
731
                #
 
732
                # TODO: at least log the other exceptions rather than just
 
733
                # losing them mbp 20070307
 
734
                exc_info = None
 
735
                for rollback in reversed(rollbacks):
 
736
                    try:
 
737
                        rollback()
 
738
                    except Exception, e:
 
739
                        exc_info = sys.exc_info()
 
740
                if exc_info:
 
741
                    raise exc_info[0], exc_info[1], exc_info[2]
 
742
 
 
743
            # perform the disk move first - its the most likely failure point.
 
744
            if move_file:
 
745
                from_rel_abs = self.abspath(from_rel)
 
746
                to_rel_abs = self.abspath(to_rel)
 
747
                try:
 
748
                    osutils.rename(from_rel_abs, to_rel_abs)
 
749
                except OSError, e:
 
750
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
751
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
752
            try:
 
753
                # perform the rename in the inventory next if needed: its easy
 
754
                # to rollback
 
755
                if update_inventory:
 
756
                    # rename the entry
 
757
                    from_entry = inv[from_id]
 
758
                    current_parent = from_entry.parent_id
 
759
                    inv.rename(from_id, to_dir_id, from_tail)
 
760
                    rollbacks.append(
 
761
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
762
                # finally do the rename in the dirstate, which is a little
 
763
                # tricky to rollback, but least likely to need it.
 
764
                old_block_index, old_entry_index, dir_present, file_present = \
 
765
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
766
                old_block = state._dirblocks[old_block_index][1]
 
767
                old_entry = old_block[old_entry_index]
 
768
                from_key, old_entry_details = old_entry
 
769
                cur_details = old_entry_details[0]
 
770
                # remove the old row
 
771
                to_key = ((to_block[0],) + from_key[1:3])
 
772
                minikind = cur_details[0]
 
773
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
774
                         minikind=minikind,
 
775
                         executable=cur_details[3],
 
776
                         fingerprint=cur_details[1],
 
777
                         packed_stat=cur_details[4],
 
778
                         size=cur_details[2],
 
779
                         to_block=to_block,
 
780
                         to_key=to_key,
 
781
                         to_path_utf8=to_rel_utf8)
 
782
 
 
783
                if minikind == 'd':
 
784
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
785
                        """Recursively update all entries in this dirblock."""
 
786
                        assert from_dir != '', "renaming root not supported"
 
787
                        from_key = (from_dir, '')
 
788
                        from_block_idx, present = \
 
789
                            state._find_block_index_from_key(from_key)
 
790
                        if not present:
 
791
                            # This is the old record, if it isn't present, then
 
792
                            # there is theoretically nothing to update.
 
793
                            # (Unless it isn't present because of lazy loading,
 
794
                            # but we don't do that yet)
 
795
                            return
 
796
                        from_block = state._dirblocks[from_block_idx]
 
797
                        to_block_index, to_entry_index, _, _ = \
 
798
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
799
                        to_block_index = state._ensure_block(
 
800
                            to_block_index, to_entry_index, to_dir_utf8)
 
801
                        to_block = state._dirblocks[to_block_index]
 
802
 
 
803
                        # Grab a copy since move_one may update the list.
 
804
                        for entry in from_block[1][:]:
 
805
                            assert entry[0][0] == from_dir
 
806
                            cur_details = entry[1][0]
 
807
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
808
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
809
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
810
                            minikind = cur_details[0]
 
811
                            if minikind in 'ar':
 
812
                                # Deleted children of a renamed directory
 
813
                                # Do not need to be updated.
 
814
                                # Children that have been renamed out of this
 
815
                                # directory should also not be updated
 
816
                                continue
 
817
                            move_one(entry, from_path_utf8=from_path_utf8,
 
818
                                     minikind=minikind,
 
819
                                     executable=cur_details[3],
 
820
                                     fingerprint=cur_details[1],
 
821
                                     packed_stat=cur_details[4],
 
822
                                     size=cur_details[2],
 
823
                                     to_block=to_block,
 
824
                                     to_key=to_key,
 
825
                                     to_path_utf8=to_path_utf8)
 
826
                            if minikind == 'd':
 
827
                                # We need to move all the children of this
 
828
                                # entry
 
829
                                update_dirblock(from_path_utf8, to_key,
 
830
                                                to_path_utf8)
 
831
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
832
            except:
 
833
                rollback_rename()
 
834
                raise
 
835
            result.append((from_rel, to_rel))
 
836
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
837
            self._make_dirty(reset_inventory=False)
 
838
 
 
839
        return result
 
840
 
 
841
    def _must_be_locked(self):
 
842
        if not self._control_files._lock_count:
 
843
            raise errors.ObjectNotLocked(self)
 
844
 
 
845
    def _new_tree(self):
 
846
        """Initialize the state in this tree to be a new tree."""
 
847
        self._dirty = True
 
848
 
 
849
    @needs_read_lock
 
850
    def path2id(self, path):
 
851
        """Return the id for path in this tree."""
 
852
        path = path.strip('/')
 
853
        entry = self._get_entry(path=path)
 
854
        if entry == (None, None):
 
855
            return None
 
856
        return entry[0][2]
 
857
 
 
858
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
859
        """See Tree.paths2ids().
 
860
 
 
861
        This specialisation fast-paths the case where all the trees are in the
 
862
        dirstate.
 
863
        """
 
864
        if paths is None:
 
865
            return None
 
866
        parents = self.get_parent_ids()
 
867
        for tree in trees:
 
868
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
869
                parents):
 
870
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
871
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
872
        # -- make all paths utf8 --
 
873
        paths_utf8 = set()
 
874
        for path in paths:
 
875
            paths_utf8.add(path.encode('utf8'))
 
876
        paths = paths_utf8
 
877
        # -- paths is now a utf8 path set --
 
878
        # -- get the state object and prepare it.
 
879
        state = self.current_dirstate()
 
880
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
881
            and '' not in paths):
 
882
            paths2ids = self._paths2ids_using_bisect
 
883
        else:
 
884
            paths2ids = self._paths2ids_in_memory
 
885
        return paths2ids(paths, search_indexes,
 
886
                         require_versioned=require_versioned)
 
887
 
 
888
    def _paths2ids_in_memory(self, paths, search_indexes,
 
889
                             require_versioned=True):
 
890
        state = self.current_dirstate()
 
891
        state._read_dirblocks_if_needed()
 
892
        def _entries_for_path(path):
 
893
            """Return a list with all the entries that match path for all ids.
 
894
            """
 
895
            dirname, basename = os.path.split(path)
 
896
            key = (dirname, basename, '')
 
897
            block_index, present = state._find_block_index_from_key(key)
 
898
            if not present:
 
899
                # the block which should contain path is absent.
 
900
                return []
 
901
            result = []
 
902
            block = state._dirblocks[block_index][1]
 
903
            entry_index, _ = state._find_entry_index(key, block)
 
904
            # we may need to look at multiple entries at this path: walk while the paths match.
 
905
            while (entry_index < len(block) and
 
906
                block[entry_index][0][0:2] == key[0:2]):
 
907
                result.append(block[entry_index])
 
908
                entry_index += 1
 
909
            return result
 
910
        if require_versioned:
 
911
            # -- check all supplied paths are versioned in a search tree. --
 
912
            all_versioned = True
 
913
            for path in paths:
 
914
                path_entries = _entries_for_path(path)
 
915
                if not path_entries:
 
916
                    # this specified path is not present at all: error
 
917
                    all_versioned = False
 
918
                    break
 
919
                found_versioned = False
 
920
                # for each id at this path
 
921
                for entry in path_entries:
 
922
                    # for each tree.
 
923
                    for index in search_indexes:
 
924
                        if entry[1][index][0] != 'a': # absent
 
925
                            found_versioned = True
 
926
                            # all good: found a versioned cell
 
927
                            break
 
928
                if not found_versioned:
 
929
                    # none of the indexes was not 'absent' at all ids for this
 
930
                    # path.
 
931
                    all_versioned = False
 
932
                    break
 
933
            if not all_versioned:
 
934
                raise errors.PathsNotVersionedError(paths)
 
935
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
936
        search_paths = osutils.minimum_path_selection(paths)
 
937
        # sketch: 
 
938
        # for all search_indexs in each path at or under each element of
 
939
        # search_paths, if the detail is relocated: add the id, and add the
 
940
        # relocated path as one to search if its not searched already. If the
 
941
        # detail is not relocated, add the id.
 
942
        searched_paths = set()
 
943
        found_ids = set()
 
944
        def _process_entry(entry):
 
945
            """Look at search_indexes within entry.
 
946
 
 
947
            If a specific tree's details are relocated, add the relocation
 
948
            target to search_paths if not searched already. If it is absent, do
 
949
            nothing. Otherwise add the id to found_ids.
 
950
            """
 
951
            for index in search_indexes:
 
952
                if entry[1][index][0] == 'r': # relocated
 
953
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
954
                        search_paths.add(entry[1][index][1])
 
955
                elif entry[1][index][0] != 'a': # absent
 
956
                    found_ids.add(entry[0][2])
 
957
        while search_paths:
 
958
            current_root = search_paths.pop()
 
959
            searched_paths.add(current_root)
 
960
            # process the entries for this containing directory: the rest will be
 
961
            # found by their parents recursively.
 
962
            root_entries = _entries_for_path(current_root)
 
963
            if not root_entries:
 
964
                # this specified path is not present at all, skip it.
 
965
                continue
 
966
            for entry in root_entries:
 
967
                _process_entry(entry)
 
968
            initial_key = (current_root, '', '')
 
969
            block_index, _ = state._find_block_index_from_key(initial_key)
 
970
            while (block_index < len(state._dirblocks) and
 
971
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
972
                for entry in state._dirblocks[block_index][1]:
 
973
                    _process_entry(entry)
 
974
                block_index += 1
 
975
        return found_ids
 
976
 
 
977
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
978
                                require_versioned=True):
 
979
        state = self.current_dirstate()
 
980
        found_ids = set()
 
981
 
 
982
        split_paths = sorted(osutils.split(p) for p in paths)
 
983
        found = state._bisect_recursive(split_paths)
 
984
 
 
985
        if require_versioned:
 
986
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
987
            for dir_name in split_paths:
 
988
                if dir_name not in found_dir_names:
 
989
                    raise errors.PathsNotVersionedError(paths)
 
990
 
 
991
        for dir_name_id, trees_info in found.iteritems():
 
992
            for index in search_indexes:
 
993
                if trees_info[index][0] not in ('r', 'a'):
 
994
                    found_ids.add(dir_name_id[2])
 
995
        return found_ids
 
996
 
 
997
    def read_working_inventory(self):
 
998
        """Read the working inventory.
 
999
        
 
1000
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1001
        """
 
1002
        return self.inventory
 
1003
 
 
1004
    @needs_read_lock
 
1005
    def revision_tree(self, revision_id):
 
1006
        """See Tree.revision_tree.
 
1007
 
 
1008
        WorkingTree4 supplies revision_trees for any basis tree.
 
1009
        """
 
1010
        dirstate = self.current_dirstate()
 
1011
        parent_ids = dirstate.get_parent_ids()
 
1012
        if revision_id not in parent_ids:
 
1013
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1014
        if revision_id in dirstate.get_ghosts():
 
1015
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1016
        return DirStateRevisionTree(dirstate, revision_id,
 
1017
            self.branch.repository)
 
1018
 
 
1019
    @needs_tree_write_lock
 
1020
    def set_last_revision(self, new_revision):
 
1021
        """Change the last revision in the working tree."""
 
1022
        parents = self.get_parent_ids()
 
1023
        if new_revision in (NULL_REVISION, None):
 
1024
            assert len(parents) < 2, (
 
1025
                "setting the last parent to none with a pending merge is "
 
1026
                "unsupported.")
 
1027
            self.set_parent_ids([])
 
1028
        else:
 
1029
            self.set_parent_ids([new_revision] + parents[1:],
 
1030
                allow_leftmost_as_ghost=True)
 
1031
 
 
1032
    @needs_tree_write_lock
 
1033
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1034
        """Set the parent ids to revision_ids.
 
1035
        
 
1036
        See also set_parent_trees. This api will try to retrieve the tree data
 
1037
        for each element of revision_ids from the trees repository. If you have
 
1038
        tree data already available, it is more efficient to use
 
1039
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1040
        an easier API to use.
 
1041
 
 
1042
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1043
            working tree. Any of these may be ghosts.
 
1044
        """
 
1045
        trees = []
 
1046
        for revision_id in revision_ids:
 
1047
            try:
 
1048
                revtree = self.branch.repository.revision_tree(revision_id)
 
1049
                # TODO: jam 20070213 KnitVersionedFile raises
 
1050
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1051
                #       given revision_id is not present. Should Repository be
 
1052
                #       catching it and re-raising NoSuchRevision?
 
1053
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1054
                revtree = None
 
1055
            trees.append((revision_id, revtree))
 
1056
        self.set_parent_trees(trees,
 
1057
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1058
 
 
1059
    @needs_tree_write_lock
 
1060
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1061
        """Set the parents of the working tree.
 
1062
 
 
1063
        :param parents_list: A list of (revision_id, tree) tuples.
 
1064
            If tree is None, then that element is treated as an unreachable
 
1065
            parent tree - i.e. a ghost.
 
1066
        """
 
1067
        dirstate = self.current_dirstate()
 
1068
        if len(parents_list) > 0:
 
1069
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1070
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1071
        real_trees = []
 
1072
        ghosts = []
 
1073
        # convert absent trees to the null tree, which we convert back to
 
1074
        # missing on access.
 
1075
        for rev_id, tree in parents_list:
 
1076
            _mod_revision.check_not_reserved_id(rev_id)
 
1077
            if tree is not None:
 
1078
                real_trees.append((rev_id, tree))
 
1079
            else:
 
1080
                real_trees.append((rev_id,
 
1081
                    self.branch.repository.revision_tree(None)))
 
1082
                ghosts.append(rev_id)
 
1083
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1084
        self._make_dirty(reset_inventory=False)
 
1085
 
 
1086
    def _set_root_id(self, file_id):
 
1087
        """See WorkingTree.set_root_id."""
 
1088
        state = self.current_dirstate()
 
1089
        state.set_path_id('', file_id)
 
1090
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1091
            self._make_dirty(reset_inventory=True)
 
1092
 
 
1093
    def _sha_from_stat(self, path, stat_result):
 
1094
        """Get a sha digest from the tree's stat cache.
 
1095
 
 
1096
        The default implementation assumes no stat cache is present.
 
1097
 
 
1098
        :param path: The path.
 
1099
        :param stat_result: The stat result being looked up.
 
1100
        """
 
1101
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1102
 
 
1103
    @needs_read_lock
 
1104
    def supports_tree_reference(self):
 
1105
        return self._repo_supports_tree_reference
 
1106
 
 
1107
    def unlock(self):
 
1108
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1109
        # do non-implementation specific cleanup
 
1110
        self._cleanup()
 
1111
 
 
1112
        if self._control_files._lock_count == 1:
 
1113
            # eventually we should do signature checking during read locks for
 
1114
            # dirstate updates.
 
1115
            if self._control_files._lock_mode == 'w':
 
1116
                if self._dirty:
 
1117
                    self.flush()
 
1118
            if self._dirstate is not None:
 
1119
                # This is a no-op if there are no modifications.
 
1120
                self._dirstate.save()
 
1121
                self._dirstate.unlock()
 
1122
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1123
            #       point. Instead, it could check if the header has been
 
1124
            #       modified when it is locked, and if not, it can hang on to
 
1125
            #       the data it has in memory.
 
1126
            self._dirstate = None
 
1127
            self._inventory = None
 
1128
        # reverse order of locking.
 
1129
        try:
 
1130
            return self._control_files.unlock()
 
1131
        finally:
 
1132
            self.branch.unlock()
 
1133
 
 
1134
    @needs_tree_write_lock
 
1135
    def unversion(self, file_ids):
 
1136
        """Remove the file ids in file_ids from the current versioned set.
 
1137
 
 
1138
        When a file_id is unversioned, all of its children are automatically
 
1139
        unversioned.
 
1140
 
 
1141
        :param file_ids: The file ids to stop versioning.
 
1142
        :raises: NoSuchId if any fileid is not currently versioned.
 
1143
        """
 
1144
        if not file_ids:
 
1145
            return
 
1146
        state = self.current_dirstate()
 
1147
        state._read_dirblocks_if_needed()
 
1148
        ids_to_unversion = set(file_ids)
 
1149
        paths_to_unversion = set()
 
1150
        # sketch:
 
1151
        # check if the root is to be unversioned, if so, assert for now.
 
1152
        # walk the state marking unversioned things as absent.
 
1153
        # if there are any un-unversioned ids at the end, raise
 
1154
        for key, details in state._dirblocks[0][1]:
 
1155
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1156
                key[2] in ids_to_unversion):
 
1157
                # I haven't written the code to unversion / yet - it should be
 
1158
                # supported.
 
1159
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1160
        block_index = 0
 
1161
        while block_index < len(state._dirblocks):
 
1162
            # process one directory at a time.
 
1163
            block = state._dirblocks[block_index]
 
1164
            # first check: is the path one to remove - it or its children
 
1165
            delete_block = False
 
1166
            for path in paths_to_unversion:
 
1167
                if (block[0].startswith(path) and
 
1168
                    (len(block[0]) == len(path) or
 
1169
                     block[0][len(path)] == '/')):
 
1170
                    # this entire block should be deleted - its the block for a
 
1171
                    # path to unversion; or the child of one
 
1172
                    delete_block = True
 
1173
                    break
 
1174
            # TODO: trim paths_to_unversion as we pass by paths
 
1175
            if delete_block:
 
1176
                # this block is to be deleted: process it.
 
1177
                # TODO: we can special case the no-parents case and
 
1178
                # just forget the whole block.
 
1179
                entry_index = 0
 
1180
                while entry_index < len(block[1]):
 
1181
                    # Mark this file id as having been removed
 
1182
                    entry = block[1][entry_index]
 
1183
                    ids_to_unversion.discard(entry[0][2])
 
1184
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1185
                                               # entries
 
1186
                        or not state._make_absent(entry)):
 
1187
                        entry_index += 1
 
1188
                # go to the next block. (At the moment we dont delete empty
 
1189
                # dirblocks)
 
1190
                block_index += 1
 
1191
                continue
 
1192
            entry_index = 0
 
1193
            while entry_index < len(block[1]):
 
1194
                entry = block[1][entry_index]
 
1195
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1196
                    # ^ some parent row.
 
1197
                    entry[0][2] not in ids_to_unversion):
 
1198
                    # ^ not an id to unversion
 
1199
                    entry_index += 1
 
1200
                    continue
 
1201
                if entry[1][0][0] == 'd':
 
1202
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1203
                if not state._make_absent(entry):
 
1204
                    entry_index += 1
 
1205
                # we have unversioned this id
 
1206
                ids_to_unversion.remove(entry[0][2])
 
1207
            block_index += 1
 
1208
        if ids_to_unversion:
 
1209
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1210
        self._make_dirty(reset_inventory=False)
 
1211
        # have to change the legacy inventory too.
 
1212
        if self._inventory is not None:
 
1213
            for file_id in file_ids:
 
1214
                self._inventory.remove_recursive_id(file_id)
 
1215
 
 
1216
    def update_basis_by_delta(self, new_revid, delta):
 
1217
        """See MutableTree.update_basis_by_delta."""
 
1218
        assert self.last_revision() != new_revid
 
1219
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1220
 
 
1221
    @needs_read_lock
 
1222
    def _validate(self):
 
1223
        self._dirstate._validate()
 
1224
 
 
1225
    @needs_tree_write_lock
 
1226
    def _write_inventory(self, inv):
 
1227
        """Write inventory as the current inventory."""
 
1228
        assert not self._dirty, ("attempting to write an inventory when the "
 
1229
            "dirstate is dirty will cause data loss")
 
1230
        self.current_dirstate().set_state_from_inventory(inv)
 
1231
        self._make_dirty(reset_inventory=False)
 
1232
        if self._inventory is not None:
 
1233
            self._inventory = inv
 
1234
        self.flush()
 
1235
 
 
1236
 
 
1237
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1238
    """The first consolidated dirstate working tree format.
 
1239
 
 
1240
    This format:
 
1241
        - exists within a metadir controlling .bzr
 
1242
        - includes an explicit version marker for the workingtree control
 
1243
          files, separate from the BzrDir format
 
1244
        - modifies the hash cache format
 
1245
        - is new in bzr 0.15
 
1246
        - uses a LockDir to guard access to it.
 
1247
    """
 
1248
 
 
1249
    upgrade_recommended = False
 
1250
 
 
1251
    def get_format_string(self):
 
1252
        """See WorkingTreeFormat.get_format_string()."""
 
1253
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1254
 
 
1255
    def get_format_description(self):
 
1256
        """See WorkingTreeFormat.get_format_description()."""
 
1257
        return "Working tree format 4"
 
1258
 
 
1259
    def initialize(self, a_bzrdir, revision_id=None):
 
1260
        """See WorkingTreeFormat.initialize().
 
1261
 
 
1262
        :param revision_id: allows creating a working tree at a different
 
1263
        revision than the branch is at.
 
1264
 
 
1265
        These trees get an initial random root id, if their repository supports
 
1266
        rich root data, TREE_ROOT otherwise.
 
1267
        """
 
1268
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1269
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1270
        transport = a_bzrdir.get_workingtree_transport(self)
 
1271
        control_files = self._open_control_files(a_bzrdir)
 
1272
        control_files.create_lock()
 
1273
        control_files.lock_write()
 
1274
        control_files.put_utf8('format', self.get_format_string())
 
1275
        branch = a_bzrdir.open_branch()
 
1276
        if revision_id is None:
 
1277
            revision_id = branch.last_revision()
 
1278
        local_path = transport.local_abspath('dirstate')
 
1279
        # write out new dirstate (must exist when we create the tree)
 
1280
        state = dirstate.DirState.initialize(local_path)
 
1281
        state.unlock()
 
1282
        del state
 
1283
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1284
                         branch,
 
1285
                         _format=self,
 
1286
                         _bzrdir=a_bzrdir,
 
1287
                         _control_files=control_files)
 
1288
        wt._new_tree()
 
1289
        wt.lock_tree_write()
 
1290
        try:
 
1291
            if revision_id in (None, NULL_REVISION):
 
1292
                if branch.repository.supports_rich_root():
 
1293
                    wt._set_root_id(generate_ids.gen_root_id())
 
1294
                else:
 
1295
                    wt._set_root_id(ROOT_ID)
 
1296
                wt.flush()
 
1297
            wt.set_last_revision(revision_id)
 
1298
            wt.flush()
 
1299
            basis = wt.basis_tree()
 
1300
            basis.lock_read()
 
1301
            # if the basis has a root id we have to use that; otherwise we use
 
1302
            # a new random one
 
1303
            basis_root_id = basis.get_root_id()
 
1304
            if basis_root_id is not None:
 
1305
                wt._set_root_id(basis_root_id)
 
1306
                wt.flush()
 
1307
            transform.build_tree(basis, wt)
 
1308
            basis.unlock()
 
1309
        finally:
 
1310
            control_files.unlock()
 
1311
            wt.unlock()
 
1312
        return wt
 
1313
 
 
1314
    def _open(self, a_bzrdir, control_files):
 
1315
        """Open the tree itself.
 
1316
 
 
1317
        :param a_bzrdir: the dir for the tree.
 
1318
        :param control_files: the control files for the tree.
 
1319
        """
 
1320
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1321
                           branch=a_bzrdir.open_branch(),
 
1322
                           _format=self,
 
1323
                           _bzrdir=a_bzrdir,
 
1324
                           _control_files=control_files)
 
1325
 
 
1326
    def __get_matchingbzrdir(self):
 
1327
        # please test against something that will let us do tree references
 
1328
        return bzrdir.format_registry.make_bzrdir(
 
1329
            'dirstate-with-subtree')
 
1330
 
 
1331
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1332
 
 
1333
 
 
1334
class DirStateRevisionTree(Tree):
 
1335
    """A revision tree pulling the inventory from a dirstate."""
 
1336
 
 
1337
    def __init__(self, dirstate, revision_id, repository):
 
1338
        self._dirstate = dirstate
 
1339
        self._revision_id = revision_id
 
1340
        self._repository = repository
 
1341
        self._inventory = None
 
1342
        self._locked = 0
 
1343
        self._dirstate_locked = False
 
1344
 
 
1345
    def __repr__(self):
 
1346
        return "<%s of %s in %s>" % \
 
1347
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1348
 
 
1349
    def annotate_iter(self, file_id,
 
1350
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1351
        """See Tree.annotate_iter"""
 
1352
        w = self._get_weave(file_id)
 
1353
        return w.annotate_iter(self.inventory[file_id].revision)
 
1354
 
 
1355
    def _get_ancestors(self, default_revision):
 
1356
        return set(self._repository.get_ancestry(self._revision_id,
 
1357
                                                 topo_sorted=False))
 
1358
    def _comparison_data(self, entry, path):
 
1359
        """See Tree._comparison_data."""
 
1360
        if entry is None:
 
1361
            return None, False, None
 
1362
        # trust the entry as RevisionTree does, but this may not be
 
1363
        # sensible: the entry might not have come from us?
 
1364
        return entry.kind, entry.executable, None
 
1365
 
 
1366
    def _file_size(self, entry, stat_value):
 
1367
        return entry.text_size
 
1368
 
 
1369
    def filter_unversioned_files(self, paths):
 
1370
        """Filter out paths that are not versioned.
 
1371
 
 
1372
        :return: set of paths.
 
1373
        """
 
1374
        pred = self.has_filename
 
1375
        return set((p for p in paths if not pred(p)))
 
1376
 
 
1377
    def get_root_id(self):
 
1378
        return self.path2id('')
 
1379
 
 
1380
    def _get_parent_index(self):
 
1381
        """Return the index in the dirstate referenced by this tree."""
 
1382
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1383
 
 
1384
    def _get_entry(self, file_id=None, path=None):
 
1385
        """Get the dirstate row for file_id or path.
 
1386
 
 
1387
        If either file_id or path is supplied, it is used as the key to lookup.
 
1388
        If both are supplied, the fastest lookup is used, and an error is
 
1389
        raised if they do not both point at the same row.
 
1390
        
 
1391
        :param file_id: An optional unicode file_id to be looked up.
 
1392
        :param path: An optional unicode path to be looked up.
 
1393
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1394
        """
 
1395
        if file_id is None and path is None:
 
1396
            raise errors.BzrError('must supply file_id or path')
 
1397
        if path is not None:
 
1398
            path = path.encode('utf8')
 
1399
        parent_index = self._get_parent_index()
 
1400
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1401
 
 
1402
    def _generate_inventory(self):
 
1403
        """Create and set self.inventory from the dirstate object.
 
1404
 
 
1405
        (So this is only called the first time the inventory is requested for
 
1406
        this tree; it then remains in memory until it's out of date.)
 
1407
 
 
1408
        This is relatively expensive: we have to walk the entire dirstate.
 
1409
        """
 
1410
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1411
            'dirstate revision tree'
 
1412
        # separate call for profiling - makes it clear where the costs are.
 
1413
        self._dirstate._read_dirblocks_if_needed()
 
1414
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1415
            'parent %s has disappeared from %s' % (
 
1416
            self._revision_id, self._dirstate.get_parent_ids())
 
1417
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1418
        # This is identical now to the WorkingTree _generate_inventory except
 
1419
        # for the tree index use.
 
1420
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1421
        current_id = root_key[2]
 
1422
        assert current_entry[parent_index][0] == 'd'
 
1423
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1424
        inv.root.revision = current_entry[parent_index][4]
 
1425
        # Turn some things into local variables
 
1426
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1427
        factory = entry_factory
 
1428
        utf8_decode = cache_utf8._utf8_decode
 
1429
        inv_byid = inv._byid
 
1430
        # we could do this straight out of the dirstate; it might be fast
 
1431
        # and should be profiled - RBC 20070216
 
1432
        parent_ies = {'' : inv.root}
 
1433
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1434
            dirname = block[0]
 
1435
            try:
 
1436
                parent_ie = parent_ies[dirname]
 
1437
            except KeyError:
 
1438
                # all the paths in this block are not versioned in this tree
 
1439
                continue
 
1440
            for key, entry in block[1]:
 
1441
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1442
                if minikind in ('a', 'r'): # absent, relocated
 
1443
                    # not this tree
 
1444
                    continue
 
1445
                name = key[1]
 
1446
                name_unicode = utf8_decode(name)[0]
 
1447
                file_id = key[2]
 
1448
                kind = minikind_to_kind[minikind]
 
1449
                inv_entry = factory[kind](file_id, name_unicode,
 
1450
                                          parent_ie.file_id)
 
1451
                inv_entry.revision = revid
 
1452
                if kind == 'file':
 
1453
                    inv_entry.executable = executable
 
1454
                    inv_entry.text_size = size
 
1455
                    inv_entry.text_sha1 = fingerprint
 
1456
                elif kind == 'directory':
 
1457
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1458
                elif kind == 'symlink':
 
1459
                    inv_entry.executable = False
 
1460
                    inv_entry.text_size = None
 
1461
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1462
                elif kind == 'tree-reference':
 
1463
                    inv_entry.reference_revision = fingerprint or None
 
1464
                else:
 
1465
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1466
                            % entry)
 
1467
                # These checks cost us around 40ms on a 55k entry tree
 
1468
                assert file_id not in inv_byid
 
1469
                assert name_unicode not in parent_ie.children
 
1470
                inv_byid[file_id] = inv_entry
 
1471
                parent_ie.children[name_unicode] = inv_entry
 
1472
        self._inventory = inv
 
1473
 
 
1474
    def get_file_mtime(self, file_id, path=None):
 
1475
        """Return the modification time for this record.
 
1476
 
 
1477
        We return the timestamp of the last-changed revision.
 
1478
        """
 
1479
        # Make sure the file exists
 
1480
        entry = self._get_entry(file_id, path=path)
 
1481
        if entry == (None, None): # do we raise?
 
1482
            return None
 
1483
        parent_index = self._get_parent_index()
 
1484
        last_changed_revision = entry[1][parent_index][4]
 
1485
        return self._repository.get_revision(last_changed_revision).timestamp
 
1486
 
 
1487
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1488
        entry = self._get_entry(file_id=file_id, path=path)
 
1489
        parent_index = self._get_parent_index()
 
1490
        parent_details = entry[1][parent_index]
 
1491
        if parent_details[0] == 'f':
 
1492
            return parent_details[1]
 
1493
        return None
 
1494
 
 
1495
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1496
    def get_weave(self, file_id):
 
1497
        return self._get_weave(file_id)
 
1498
 
 
1499
    def _get_weave(self, file_id):
 
1500
        return self._repository.weave_store.get_weave(file_id,
 
1501
                self._repository.get_transaction())
 
1502
 
 
1503
    def get_file(self, file_id, path=None):
 
1504
        return StringIO(self.get_file_text(file_id))
 
1505
 
 
1506
    def get_file_lines(self, file_id):
 
1507
        ie = self.inventory[file_id]
 
1508
        return self._get_weave(file_id).get_lines(ie.revision)
 
1509
 
 
1510
    def get_file_size(self, file_id):
 
1511
        return self.inventory[file_id].text_size
 
1512
 
 
1513
    def get_file_text(self, file_id):
 
1514
        return ''.join(self.get_file_lines(file_id))
 
1515
 
 
1516
    def get_reference_revision(self, file_id, path=None):
 
1517
        return self.inventory[file_id].reference_revision
 
1518
 
 
1519
    def iter_files_bytes(self, desired_files):
 
1520
        """See Tree.iter_files_bytes.
 
1521
 
 
1522
        This version is implemented on top of Repository.iter_files_bytes"""
 
1523
        parent_index = self._get_parent_index()
 
1524
        repo_desired_files = []
 
1525
        for file_id, identifier in desired_files:
 
1526
            entry = self._get_entry(file_id)
 
1527
            if entry == (None, None):
 
1528
                raise errors.NoSuchId(self, file_id)
 
1529
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1530
                                       identifier))
 
1531
        return self._repository.iter_files_bytes(repo_desired_files)
 
1532
 
 
1533
    def get_symlink_target(self, file_id):
 
1534
        entry = self._get_entry(file_id=file_id)
 
1535
        parent_index = self._get_parent_index()
 
1536
        if entry[1][parent_index][0] != 'l':
 
1537
            return None
 
1538
        else:
 
1539
            # At present, none of the tree implementations supports non-ascii
 
1540
            # symlink targets. So we will just assume that the dirstate path is
 
1541
            # correct.
 
1542
            return entry[1][parent_index][1]
 
1543
 
 
1544
    def get_revision_id(self):
 
1545
        """Return the revision id for this tree."""
 
1546
        return self._revision_id
 
1547
 
 
1548
    def _get_inventory(self):
 
1549
        if self._inventory is not None:
 
1550
            return self._inventory
 
1551
        self._must_be_locked()
 
1552
        self._generate_inventory()
 
1553
        return self._inventory
 
1554
 
 
1555
    inventory = property(_get_inventory,
 
1556
                         doc="Inventory of this Tree")
 
1557
 
 
1558
    def get_parent_ids(self):
 
1559
        """The parents of a tree in the dirstate are not cached."""
 
1560
        return self._repository.get_revision(self._revision_id).parent_ids
 
1561
 
 
1562
    def has_filename(self, filename):
 
1563
        return bool(self.path2id(filename))
 
1564
 
 
1565
    def kind(self, file_id):
 
1566
        return self.inventory[file_id].kind
 
1567
 
 
1568
    def path_content_summary(self, path):
 
1569
        """See Tree.path_content_summary."""
 
1570
        id = self.inventory.path2id(path)
 
1571
        if id is None:
 
1572
            return ('missing', None, None, None)
 
1573
        entry = self._inventory[id]
 
1574
        kind = entry.kind
 
1575
        if kind == 'file':
 
1576
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1577
        elif kind == 'symlink':
 
1578
            return (kind, None, None, entry.symlink_target)
 
1579
        else:
 
1580
            return (kind, None, None, None)
 
1581
 
 
1582
    def is_executable(self, file_id, path=None):
 
1583
        ie = self.inventory[file_id]
 
1584
        if ie.kind != "file":
 
1585
            return None
 
1586
        return ie.executable
 
1587
 
 
1588
    def list_files(self, include_root=False):
 
1589
        # We use a standard implementation, because DirStateRevisionTree is
 
1590
        # dealing with one of the parents of the current state
 
1591
        inv = self._get_inventory()
 
1592
        entries = inv.iter_entries()
 
1593
        if self.inventory.root is not None and not include_root:
 
1594
            entries.next()
 
1595
        for path, entry in entries:
 
1596
            yield path, 'V', entry.kind, entry.file_id, entry
 
1597
 
 
1598
    def lock_read(self):
 
1599
        """Lock the tree for a set of operations."""
 
1600
        if not self._locked:
 
1601
            self._repository.lock_read()
 
1602
            if self._dirstate._lock_token is None:
 
1603
                self._dirstate.lock_read()
 
1604
                self._dirstate_locked = True
 
1605
        self._locked += 1
 
1606
 
 
1607
    def _must_be_locked(self):
 
1608
        if not self._locked:
 
1609
            raise errors.ObjectNotLocked(self)
 
1610
 
 
1611
    @needs_read_lock
 
1612
    def path2id(self, path):
 
1613
        """Return the id for path in this tree."""
 
1614
        # lookup by path: faster than splitting and walking the ivnentory.
 
1615
        entry = self._get_entry(path=path)
 
1616
        if entry == (None, None):
 
1617
            return None
 
1618
        return entry[0][2]
 
1619
 
 
1620
    def unlock(self):
 
1621
        """Unlock, freeing any cache memory used during the lock."""
 
1622
        # outside of a lock, the inventory is suspect: release it.
 
1623
        self._locked -=1
 
1624
        if not self._locked:
 
1625
            self._inventory = None
 
1626
            self._locked = 0
 
1627
            if self._dirstate_locked:
 
1628
                self._dirstate.unlock()
 
1629
                self._dirstate_locked = False
 
1630
            self._repository.unlock()
 
1631
 
 
1632
    def walkdirs(self, prefix=""):
 
1633
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1634
        # implementation based on an inventory.  
 
1635
        # This should be cleaned up to use the much faster Dirstate code
 
1636
        # So for now, we just build up the parent inventory, and extract
 
1637
        # it the same way RevisionTree does.
 
1638
        _directory = 'directory'
 
1639
        inv = self._get_inventory()
 
1640
        top_id = inv.path2id(prefix)
 
1641
        if top_id is None:
 
1642
            pending = []
 
1643
        else:
 
1644
            pending = [(prefix, top_id)]
 
1645
        while pending:
 
1646
            dirblock = []
 
1647
            relpath, file_id = pending.pop()
 
1648
            # 0 - relpath, 1- file-id
 
1649
            if relpath:
 
1650
                relroot = relpath + '/'
 
1651
            else:
 
1652
                relroot = ""
 
1653
            # FIXME: stash the node in pending
 
1654
            entry = inv[file_id]
 
1655
            for name, child in entry.sorted_children():
 
1656
                toppath = relroot + name
 
1657
                dirblock.append((toppath, name, child.kind, None,
 
1658
                    child.file_id, child.kind
 
1659
                    ))
 
1660
            yield (relpath, entry.file_id), dirblock
 
1661
            # push the user specified dirs from dirblock
 
1662
            for dir in reversed(dirblock):
 
1663
                if dir[2] == _directory:
 
1664
                    pending.append((dir[0], dir[4]))
 
1665
 
 
1666
 
 
1667
class InterDirStateTree(InterTree):
 
1668
    """Fast path optimiser for changes_from with dirstate trees.
 
1669
    
 
1670
    This is used only when both trees are in the dirstate working file, and 
 
1671
    the source is any parent within the dirstate, and the destination is 
 
1672
    the current working tree of the same dirstate.
 
1673
    """
 
1674
    # this could be generalized to allow comparisons between any trees in the
 
1675
    # dirstate, and possibly between trees stored in different dirstates.
 
1676
 
 
1677
    def __init__(self, source, target):
 
1678
        super(InterDirStateTree, self).__init__(source, target)
 
1679
        if not InterDirStateTree.is_compatible(source, target):
 
1680
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1681
 
 
1682
    @staticmethod
 
1683
    def make_source_parent_tree(source, target):
 
1684
        """Change the source tree into a parent of the target."""
 
1685
        revid = source.commit('record tree')
 
1686
        target.branch.repository.fetch(source.branch.repository, revid)
 
1687
        target.set_parent_ids([revid])
 
1688
        return target.basis_tree(), target
 
1689
 
 
1690
    _matching_from_tree_format = WorkingTreeFormat4()
 
1691
    _matching_to_tree_format = WorkingTreeFormat4()
 
1692
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1693
 
 
1694
    def _iter_changes(self, include_unchanged=False,
 
1695
                      specific_files=None, pb=None, extra_trees=[],
 
1696
                      require_versioned=True, want_unversioned=False):
 
1697
        """Return the changes from source to target.
 
1698
 
 
1699
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1700
            for details.
 
1701
        :param specific_files: An optional list of file paths to restrict the
 
1702
            comparison to. When mapping filenames to ids, all matches in all
 
1703
            trees (including optional extra_trees) are used, and all children of
 
1704
            matched directories are included.
 
1705
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1706
            unchanged entries in the result.
 
1707
        :param extra_trees: An optional list of additional trees to use when
 
1708
            mapping the contents of specific_files (paths) to file_ids.
 
1709
        :param require_versioned: If True, all files in specific_files must be
 
1710
            versioned in one of source, target, extra_trees or
 
1711
            PathsNotVersionedError is raised.
 
1712
        :param want_unversioned: Should unversioned files be returned in the
 
1713
            output. An unversioned file is defined as one with (False, False)
 
1714
            for the versioned pair.
 
1715
        """
 
1716
        utf8_decode = cache_utf8._utf8_decode
 
1717
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1718
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1719
        # NB: show_status depends on being able to pass in non-versioned files
 
1720
        # and report them as unknown
 
1721
        # TODO: handle extra trees in the dirstate.
 
1722
        # TODO: handle comparisons as an empty tree as a different special
 
1723
        # case? mbp 20070226
 
1724
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1725
            or specific_files == []):
 
1726
            # we can't fast-path these cases (yet)
 
1727
            for f in super(InterDirStateTree, self)._iter_changes(
 
1728
                include_unchanged, specific_files, pb, extra_trees,
 
1729
                require_versioned, want_unversioned=want_unversioned):
 
1730
                yield f
 
1731
            return
 
1732
        parent_ids = self.target.get_parent_ids()
 
1733
        assert (self.source._revision_id in parent_ids), \
 
1734
                "revision {%s} is not stored in {%s}, but %s " \
 
1735
                "can only be used for trees stored in the dirstate" \
 
1736
                % (self.source._revision_id, self.target, self._iter_changes)
 
1737
        target_index = 0
 
1738
        if self.source._revision_id == NULL_REVISION:
 
1739
            source_index = None
 
1740
            indices = (target_index,)
 
1741
        else:
 
1742
            assert (self.source._revision_id in parent_ids), \
 
1743
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1744
                self.source._revision_id, parent_ids)
 
1745
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1746
            indices = (source_index,target_index)
 
1747
        # -- make all specific_files utf8 --
 
1748
        if specific_files:
 
1749
            specific_files_utf8 = set()
 
1750
            for path in specific_files:
 
1751
                specific_files_utf8.add(path.encode('utf8'))
 
1752
            specific_files = specific_files_utf8
 
1753
        else:
 
1754
            specific_files = set([''])
 
1755
        # -- specific_files is now a utf8 path set --
 
1756
        # -- get the state object and prepare it.
 
1757
        state = self.target.current_dirstate()
 
1758
        state._read_dirblocks_if_needed()
 
1759
        def _entries_for_path(path):
 
1760
            """Return a list with all the entries that match path for all ids.
 
1761
            """
 
1762
            dirname, basename = os.path.split(path)
 
1763
            key = (dirname, basename, '')
 
1764
            block_index, present = state._find_block_index_from_key(key)
 
1765
            if not present:
 
1766
                # the block which should contain path is absent.
 
1767
                return []
 
1768
            result = []
 
1769
            block = state._dirblocks[block_index][1]
 
1770
            entry_index, _ = state._find_entry_index(key, block)
 
1771
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1772
            while (entry_index < len(block) and
 
1773
                block[entry_index][0][0:2] == key[0:2]):
 
1774
                result.append(block[entry_index])
 
1775
                entry_index += 1
 
1776
            return result
 
1777
        if require_versioned:
 
1778
            # -- check all supplied paths are versioned in a search tree. --
 
1779
            all_versioned = True
 
1780
            for path in specific_files:
 
1781
                path_entries = _entries_for_path(path)
 
1782
                if not path_entries:
 
1783
                    # this specified path is not present at all: error
 
1784
                    all_versioned = False
 
1785
                    break
 
1786
                found_versioned = False
 
1787
                # for each id at this path
 
1788
                for entry in path_entries:
 
1789
                    # for each tree.
 
1790
                    for index in indices:
 
1791
                        if entry[1][index][0] != 'a': # absent
 
1792
                            found_versioned = True
 
1793
                            # all good: found a versioned cell
 
1794
                            break
 
1795
                if not found_versioned:
 
1796
                    # none of the indexes was not 'absent' at all ids for this
 
1797
                    # path.
 
1798
                    all_versioned = False
 
1799
                    break
 
1800
            if not all_versioned:
 
1801
                raise errors.PathsNotVersionedError(specific_files)
 
1802
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1803
        search_specific_files = set()
 
1804
        for path in specific_files:
 
1805
            other_specific_files = specific_files.difference(set([path]))
 
1806
            if not osutils.is_inside_any(other_specific_files, path):
 
1807
                # this is a top level path, we must check it.
 
1808
                search_specific_files.add(path)
 
1809
        # sketch: 
 
1810
        # compare source_index and target_index at or under each element of search_specific_files.
 
1811
        # follow the following comparison table. Note that we only want to do diff operations when
 
1812
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1813
        # for the target.
 
1814
        # cases:
 
1815
        # 
 
1816
        # Source | Target | disk | action
 
1817
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1818
        #        |        |      | diff check on source-target
 
1819
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1820
        #        |        |      | ???
 
1821
        #   r    |  a     |      | add source to search
 
1822
        #   r    |  a     |  a   | 
 
1823
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1824
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1825
        #   a    | fdlt   |      | add new id
 
1826
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1827
        #   a    |  a     |      | not present in either tree, skip
 
1828
        #   a    |  a     |  a   | not present in any tree, skip
 
1829
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1830
        #        |        |      | may not be selected by the users list of paths.
 
1831
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1832
        #        |        |      | may not be selected by the users list of paths.
 
1833
        #  fdlt  | fdlt   |      | content in both: diff them
 
1834
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1835
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1836
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1837
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1838
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1839
        #        |        |      | this id at the other path.
 
1840
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1841
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1842
        #        |        |      | this id at the other path.
 
1843
 
 
1844
        # for all search_indexs in each path at or under each element of
 
1845
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1846
        # relocated path as one to search if its not searched already. If the
 
1847
        # detail is not relocated, add the id.
 
1848
        searched_specific_files = set()
 
1849
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1850
        # Using a list so that we can access the values and change them in
 
1851
        # nested scope. Each one is [path, file_id, entry]
 
1852
        last_source_parent = [None, None]
 
1853
        last_target_parent = [None, None]
 
1854
 
 
1855
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1856
 
 
1857
        # Just a sentry, so that _process_entry can say that this
 
1858
        # record is handled, but isn't interesting to process (unchanged)
 
1859
        uninteresting = object()
 
1860
 
 
1861
 
 
1862
        old_dirname_to_file_id = {}
 
1863
        new_dirname_to_file_id = {}
 
1864
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1865
        #       keeping a cache of directories that we have seen.
 
1866
 
 
1867
        def _process_entry(entry, path_info):
 
1868
            """Compare an entry and real disk to generate delta information.
 
1869
 
 
1870
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1871
                the path of entry. If None, then the path is considered absent.
 
1872
                (Perhaps we should pass in a concrete entry for this ?)
 
1873
                Basename is returned as a utf8 string because we expect this
 
1874
                tuple will be ignored, and don't want to take the time to
 
1875
                decode.
 
1876
            :return: None if these don't match
 
1877
                     A tuple of information about the change, or
 
1878
                     the object 'uninteresting' if these match, but are
 
1879
                     basically identical.
 
1880
            """
 
1881
            if source_index is None:
 
1882
                source_details = NULL_PARENT_DETAILS
 
1883
            else:
 
1884
                source_details = entry[1][source_index]
 
1885
            target_details = entry[1][target_index]
 
1886
            target_minikind = target_details[0]
 
1887
            if path_info is not None and target_minikind in 'fdlt':
 
1888
                assert target_index == 0
 
1889
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1890
                                                  stat_value=path_info[3])
 
1891
                # The entry may have been modified by update_entry
 
1892
                target_details = entry[1][target_index]
 
1893
                target_minikind = target_details[0]
 
1894
            else:
 
1895
                link_or_sha1 = None
 
1896
            file_id = entry[0][2]
 
1897
            source_minikind = source_details[0]
 
1898
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1899
                # claimed content in both: diff
 
1900
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1901
                #        |        |      | diff check on source-target
 
1902
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1903
                #        |        |      | ???
 
1904
                if source_minikind in 'r':
 
1905
                    # add the source to the search path to find any children it
 
1906
                    # has.  TODO ? : only add if it is a container ?
 
1907
                    if not osutils.is_inside_any(searched_specific_files,
 
1908
                                                 source_details[1]):
 
1909
                        search_specific_files.add(source_details[1])
 
1910
                    # generate the old path; this is needed for stating later
 
1911
                    # as well.
 
1912
                    old_path = source_details[1]
 
1913
                    old_dirname, old_basename = os.path.split(old_path)
 
1914
                    path = pathjoin(entry[0][0], entry[0][1])
 
1915
                    old_entry = state._get_entry(source_index,
 
1916
                                                 path_utf8=old_path)
 
1917
                    # update the source details variable to be the real
 
1918
                    # location.
 
1919
                    source_details = old_entry[1][source_index]
 
1920
                    source_minikind = source_details[0]
 
1921
                else:
 
1922
                    old_dirname = entry[0][0]
 
1923
                    old_basename = entry[0][1]
 
1924
                    old_path = path = None
 
1925
                if path_info is None:
 
1926
                    # the file is missing on disk, show as removed.
 
1927
                    content_change = True
 
1928
                    target_kind = None
 
1929
                    target_exec = False
 
1930
                else:
 
1931
                    # source and target are both versioned and disk file is present.
 
1932
                    target_kind = path_info[2]
 
1933
                    if target_kind == 'directory':
 
1934
                        if path is None:
 
1935
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1936
                        new_dirname_to_file_id[path] = file_id
 
1937
                        if source_minikind != 'd':
 
1938
                            content_change = True
 
1939
                        else:
 
1940
                            # directories have no fingerprint
 
1941
                            content_change = False
 
1942
                        target_exec = False
 
1943
                    elif target_kind == 'file':
 
1944
                        if source_minikind != 'f':
 
1945
                            content_change = True
 
1946
                        else:
 
1947
                            # We could check the size, but we already have the
 
1948
                            # sha1 hash.
 
1949
                            content_change = (link_or_sha1 != source_details[1])
 
1950
                        # Target details is updated at update_entry time
 
1951
                        if use_filesystem_for_exec:
 
1952
                            # We don't need S_ISREG here, because we are sure
 
1953
                            # we are dealing with a file.
 
1954
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1955
                        else:
 
1956
                            target_exec = target_details[3]
 
1957
                    elif target_kind == 'symlink':
 
1958
                        if source_minikind != 'l':
 
1959
                            content_change = True
 
1960
                        else:
 
1961
                            content_change = (link_or_sha1 != source_details[1])
 
1962
                        target_exec = False
 
1963
                    elif target_kind == 'tree-reference':
 
1964
                        if source_minikind != 't':
 
1965
                            content_change = True
 
1966
                        else:
 
1967
                            content_change = False
 
1968
                        target_exec = False
 
1969
                    else:
 
1970
                        raise Exception, "unknown kind %s" % path_info[2]
 
1971
                if source_minikind == 'd':
 
1972
                    if path is None:
 
1973
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1974
                    old_dirname_to_file_id[old_path] = file_id
 
1975
                # parent id is the entry for the path in the target tree
 
1976
                if old_dirname == last_source_parent[0]:
 
1977
                    source_parent_id = last_source_parent[1]
 
1978
                else:
 
1979
                    try:
 
1980
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1981
                    except KeyError:
 
1982
                        source_parent_entry = state._get_entry(source_index,
 
1983
                                                               path_utf8=old_dirname)
 
1984
                        source_parent_id = source_parent_entry[0][2]
 
1985
                    if source_parent_id == entry[0][2]:
 
1986
                        # This is the root, so the parent is None
 
1987
                        source_parent_id = None
 
1988
                    else:
 
1989
                        last_source_parent[0] = old_dirname
 
1990
                        last_source_parent[1] = source_parent_id
 
1991
                new_dirname = entry[0][0]
 
1992
                if new_dirname == last_target_parent[0]:
 
1993
                    target_parent_id = last_target_parent[1]
 
1994
                else:
 
1995
                    try:
 
1996
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1997
                    except KeyError:
 
1998
                        # TODO: We don't always need to do the lookup, because the
 
1999
                        #       parent entry will be the same as the source entry.
 
2000
                        target_parent_entry = state._get_entry(target_index,
 
2001
                                                               path_utf8=new_dirname)
 
2002
                        assert target_parent_entry != (None, None), (
 
2003
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2004
                            % (new_dirname, entry))
 
2005
                        target_parent_id = target_parent_entry[0][2]
 
2006
                    if target_parent_id == entry[0][2]:
 
2007
                        # This is the root, so the parent is None
 
2008
                        target_parent_id = None
 
2009
                    else:
 
2010
                        last_target_parent[0] = new_dirname
 
2011
                        last_target_parent[1] = target_parent_id
 
2012
 
 
2013
                source_exec = source_details[3]
 
2014
                if (include_unchanged
 
2015
                    or content_change
 
2016
                    or source_parent_id != target_parent_id
 
2017
                    or old_basename != entry[0][1]
 
2018
                    or source_exec != target_exec
 
2019
                    ):
 
2020
                    if old_path is None:
 
2021
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2022
                        old_path_u = utf8_decode(old_path)[0]
 
2023
                        path_u = old_path_u
 
2024
                    else:
 
2025
                        old_path_u = utf8_decode(old_path)[0]
 
2026
                        if old_path == path:
 
2027
                            path_u = old_path_u
 
2028
                        else:
 
2029
                            path_u = utf8_decode(path)[0]
 
2030
                    source_kind = _minikind_to_kind[source_minikind]
 
2031
                    return (entry[0][2],
 
2032
                           (old_path_u, path_u),
 
2033
                           content_change,
 
2034
                           (True, True),
 
2035
                           (source_parent_id, target_parent_id),
 
2036
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2037
                           (source_kind, target_kind),
 
2038
                           (source_exec, target_exec))
 
2039
                else:
 
2040
                    return uninteresting
 
2041
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2042
                # looks like a new file
 
2043
                if path_info is not None:
 
2044
                    path = pathjoin(entry[0][0], entry[0][1])
 
2045
                    # parent id is the entry for the path in the target tree
 
2046
                    # TODO: these are the same for an entire directory: cache em.
 
2047
                    parent_id = state._get_entry(target_index,
 
2048
                                                 path_utf8=entry[0][0])[0][2]
 
2049
                    if parent_id == entry[0][2]:
 
2050
                        parent_id = None
 
2051
                    if use_filesystem_for_exec:
 
2052
                        # We need S_ISREG here, because we aren't sure if this
 
2053
                        # is a file or not.
 
2054
                        target_exec = bool(
 
2055
                            stat.S_ISREG(path_info[3].st_mode)
 
2056
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2057
                    else:
 
2058
                        target_exec = target_details[3]
 
2059
                    return (entry[0][2],
 
2060
                           (None, utf8_decode(path)[0]),
 
2061
                           True,
 
2062
                           (False, True),
 
2063
                           (None, parent_id),
 
2064
                           (None, utf8_decode(entry[0][1])[0]),
 
2065
                           (None, path_info[2]),
 
2066
                           (None, target_exec))
 
2067
                else:
 
2068
                    # but its not on disk: we deliberately treat this as just
 
2069
                    # never-present. (Why ?! - RBC 20070224)
 
2070
                    pass
 
2071
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2072
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2073
                # if its still on disk, *and* theres no other entry at this
 
2074
                # path [we dont know this in this routine at the moment -
 
2075
                # perhaps we should change this - then it would be an unknown.
 
2076
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2077
                # parent id is the entry for the path in the target tree
 
2078
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2079
                if parent_id == entry[0][2]:
 
2080
                    parent_id = None
 
2081
                return (entry[0][2],
 
2082
                       (utf8_decode(old_path)[0], None),
 
2083
                       True,
 
2084
                       (True, False),
 
2085
                       (parent_id, None),
 
2086
                       (utf8_decode(entry[0][1])[0], None),
 
2087
                       (_minikind_to_kind[source_minikind], None),
 
2088
                       (source_details[3], None))
 
2089
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2090
                # a rename; could be a true rename, or a rename inherited from
 
2091
                # a renamed parent. TODO: handle this efficiently. Its not
 
2092
                # common case to rename dirs though, so a correct but slow
 
2093
                # implementation will do.
 
2094
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2095
                    search_specific_files.add(target_details[1])
 
2096
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2097
                # neither of the selected trees contain this file,
 
2098
                # so skip over it. This is not currently directly tested, but
 
2099
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2100
                pass
 
2101
            else:
 
2102
                raise AssertionError("don't know how to compare "
 
2103
                    "source_minikind=%r, target_minikind=%r"
 
2104
                    % (source_minikind, target_minikind))
 
2105
                ## import pdb;pdb.set_trace()
 
2106
            return None
 
2107
 
 
2108
        while search_specific_files:
 
2109
            # TODO: the pending list should be lexically sorted?  the
 
2110
            # interface doesn't require it.
 
2111
            current_root = search_specific_files.pop()
 
2112
            current_root_unicode = current_root.decode('utf8')
 
2113
            searched_specific_files.add(current_root)
 
2114
            # process the entries for this containing directory: the rest will be
 
2115
            # found by their parents recursively.
 
2116
            root_entries = _entries_for_path(current_root)
 
2117
            root_abspath = self.target.abspath(current_root_unicode)
 
2118
            try:
 
2119
                root_stat = os.lstat(root_abspath)
 
2120
            except OSError, e:
 
2121
                if e.errno == errno.ENOENT:
 
2122
                    # the path does not exist: let _process_entry know that.
 
2123
                    root_dir_info = None
 
2124
                else:
 
2125
                    # some other random error: hand it up.
 
2126
                    raise
 
2127
            else:
 
2128
                root_dir_info = ('', current_root,
 
2129
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2130
                    root_abspath)
 
2131
                if root_dir_info[2] == 'directory':
 
2132
                    if self.target._directory_is_tree_reference(
 
2133
                        current_root.decode('utf8')):
 
2134
                        root_dir_info = root_dir_info[:2] + \
 
2135
                            ('tree-reference',) + root_dir_info[3:]
 
2136
 
 
2137
            if not root_entries and not root_dir_info:
 
2138
                # this specified path is not present at all, skip it.
 
2139
                continue
 
2140
            path_handled = False
 
2141
            for entry in root_entries:
 
2142
                result = _process_entry(entry, root_dir_info)
 
2143
                if result is not None:
 
2144
                    path_handled = True
 
2145
                    if result is not uninteresting:
 
2146
                        yield result
 
2147
            if want_unversioned and not path_handled and root_dir_info:
 
2148
                new_executable = bool(
 
2149
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2150
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2151
                yield (None,
 
2152
                       (None, current_root_unicode),
 
2153
                       True,
 
2154
                       (False, False),
 
2155
                       (None, None),
 
2156
                       (None, splitpath(current_root_unicode)[-1]),
 
2157
                       (None, root_dir_info[2]),
 
2158
                       (None, new_executable)
 
2159
                      )
 
2160
            initial_key = (current_root, '', '')
 
2161
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2162
            if block_index == 0:
 
2163
                # we have processed the total root already, but because the
 
2164
                # initial key matched it we should skip it here.
 
2165
                block_index +=1
 
2166
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2167
                current_dir_info = None
 
2168
            else:
 
2169
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2170
                try:
 
2171
                    current_dir_info = dir_iterator.next()
 
2172
                except OSError, e:
 
2173
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2174
                    # python 2.5 has e.errno == EINVAL,
 
2175
                    #            and e.winerror == ERROR_DIRECTORY
 
2176
                    e_winerror = getattr(e, 'winerror', None)
 
2177
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2178
                    # there may be directories in the inventory even though
 
2179
                    # this path is not a file on disk: so mark it as end of
 
2180
                    # iterator
 
2181
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2182
                        current_dir_info = None
 
2183
                    elif (sys.platform == 'win32'
 
2184
                          and (e.errno in win_errors
 
2185
                               or e_winerror in win_errors)):
 
2186
                        current_dir_info = None
 
2187
                    else:
 
2188
                        raise
 
2189
                else:
 
2190
                    if current_dir_info[0][0] == '':
 
2191
                        # remove .bzr from iteration
 
2192
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2193
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2194
                        del current_dir_info[1][bzr_index]
 
2195
            # walk until both the directory listing and the versioned metadata
 
2196
            # are exhausted. 
 
2197
            if (block_index < len(state._dirblocks) and
 
2198
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2199
                current_block = state._dirblocks[block_index]
 
2200
            else:
 
2201
                current_block = None
 
2202
            while (current_dir_info is not None or
 
2203
                   current_block is not None):
 
2204
                if (current_dir_info and current_block
 
2205
                    and current_dir_info[0][0] != current_block[0]):
 
2206
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2207
                        # filesystem data refers to paths not covered by the dirblock.
 
2208
                        # this has two possibilities:
 
2209
                        # A) it is versioned but empty, so there is no block for it
 
2210
                        # B) it is not versioned.
 
2211
 
 
2212
                        # if (A) then we need to recurse into it to check for
 
2213
                        # new unknown files or directories.
 
2214
                        # if (B) then we should ignore it, because we don't
 
2215
                        # recurse into unknown directories.
 
2216
                        path_index = 0
 
2217
                        while path_index < len(current_dir_info[1]):
 
2218
                                current_path_info = current_dir_info[1][path_index]
 
2219
                                if want_unversioned:
 
2220
                                    if current_path_info[2] == 'directory':
 
2221
                                        if self.target._directory_is_tree_reference(
 
2222
                                            current_path_info[0].decode('utf8')):
 
2223
                                            current_path_info = current_path_info[:2] + \
 
2224
                                                ('tree-reference',) + current_path_info[3:]
 
2225
                                    new_executable = bool(
 
2226
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2227
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2228
                                    yield (None,
 
2229
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2230
                                        True,
 
2231
                                        (False, False),
 
2232
                                        (None, None),
 
2233
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2234
                                        (None, current_path_info[2]),
 
2235
                                        (None, new_executable))
 
2236
                                # dont descend into this unversioned path if it is
 
2237
                                # a dir
 
2238
                                if current_path_info[2] in ('directory',
 
2239
                                                            'tree-reference'):
 
2240
                                    del current_dir_info[1][path_index]
 
2241
                                    path_index -= 1
 
2242
                                path_index += 1
 
2243
 
 
2244
                        # This dir info has been handled, go to the next
 
2245
                        try:
 
2246
                            current_dir_info = dir_iterator.next()
 
2247
                        except StopIteration:
 
2248
                            current_dir_info = None
 
2249
                    else:
 
2250
                        # We have a dirblock entry for this location, but there
 
2251
                        # is no filesystem path for this. This is most likely
 
2252
                        # because a directory was removed from the disk.
 
2253
                        # We don't have to report the missing directory,
 
2254
                        # because that should have already been handled, but we
 
2255
                        # need to handle all of the files that are contained
 
2256
                        # within.
 
2257
                        for current_entry in current_block[1]:
 
2258
                            # entry referring to file not present on disk.
 
2259
                            # advance the entry only, after processing.
 
2260
                            result = _process_entry(current_entry, None)
 
2261
                            if result is not None:
 
2262
                                if result is not uninteresting:
 
2263
                                    yield result
 
2264
                        block_index +=1
 
2265
                        if (block_index < len(state._dirblocks) and
 
2266
                            osutils.is_inside(current_root,
 
2267
                                              state._dirblocks[block_index][0])):
 
2268
                            current_block = state._dirblocks[block_index]
 
2269
                        else:
 
2270
                            current_block = None
 
2271
                    continue
 
2272
                entry_index = 0
 
2273
                if current_block and entry_index < len(current_block[1]):
 
2274
                    current_entry = current_block[1][entry_index]
 
2275
                else:
 
2276
                    current_entry = None
 
2277
                advance_entry = True
 
2278
                path_index = 0
 
2279
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2280
                    current_path_info = current_dir_info[1][path_index]
 
2281
                    if current_path_info[2] == 'directory':
 
2282
                        if self.target._directory_is_tree_reference(
 
2283
                            current_path_info[0].decode('utf8')):
 
2284
                            current_path_info = current_path_info[:2] + \
 
2285
                                ('tree-reference',) + current_path_info[3:]
 
2286
                else:
 
2287
                    current_path_info = None
 
2288
                advance_path = True
 
2289
                path_handled = False
 
2290
                while (current_entry is not None or
 
2291
                    current_path_info is not None):
 
2292
                    if current_entry is None:
 
2293
                        # the check for path_handled when the path is adnvaced
 
2294
                        # will yield this path if needed.
 
2295
                        pass
 
2296
                    elif current_path_info is None:
 
2297
                        # no path is fine: the per entry code will handle it.
 
2298
                        result = _process_entry(current_entry, current_path_info)
 
2299
                        if result is not None:
 
2300
                            if result is not uninteresting:
 
2301
                                yield result
 
2302
                    elif (current_entry[0][1] != current_path_info[1]
 
2303
                          or current_entry[1][target_index][0] in 'ar'):
 
2304
                        # The current path on disk doesn't match the dirblock
 
2305
                        # record. Either the dirblock is marked as absent, or
 
2306
                        # the file on disk is not present at all in the
 
2307
                        # dirblock. Either way, report about the dirblock
 
2308
                        # entry, and let other code handle the filesystem one.
 
2309
 
 
2310
                        # Compare the basename for these files to determine
 
2311
                        # which comes first
 
2312
                        if current_path_info[1] < current_entry[0][1]:
 
2313
                            # extra file on disk: pass for now, but only
 
2314
                            # increment the path, not the entry
 
2315
                            advance_entry = False
 
2316
                        else:
 
2317
                            # entry referring to file not present on disk.
 
2318
                            # advance the entry only, after processing.
 
2319
                            result = _process_entry(current_entry, None)
 
2320
                            if result is not None:
 
2321
                                if result is not uninteresting:
 
2322
                                    yield result
 
2323
                            advance_path = False
 
2324
                    else:
 
2325
                        result = _process_entry(current_entry, current_path_info)
 
2326
                        if result is not None:
 
2327
                            path_handled = True
 
2328
                            if result is not uninteresting:
 
2329
                                yield result
 
2330
                    if advance_entry and current_entry is not None:
 
2331
                        entry_index += 1
 
2332
                        if entry_index < len(current_block[1]):
 
2333
                            current_entry = current_block[1][entry_index]
 
2334
                        else:
 
2335
                            current_entry = None
 
2336
                    else:
 
2337
                        advance_entry = True # reset the advance flaga
 
2338
                    if advance_path and current_path_info is not None:
 
2339
                        if not path_handled:
 
2340
                            # unversioned in all regards
 
2341
                            if want_unversioned:
 
2342
                                new_executable = bool(
 
2343
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2344
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2345
                                yield (None,
 
2346
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2347
                                    True,
 
2348
                                    (False, False),
 
2349
                                    (None, None),
 
2350
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2351
                                    (None, current_path_info[2]),
 
2352
                                    (None, new_executable))
 
2353
                            # dont descend into this unversioned path if it is
 
2354
                            # a dir
 
2355
                            if current_path_info[2] in ('directory'):
 
2356
                                del current_dir_info[1][path_index]
 
2357
                                path_index -= 1
 
2358
                        # dont descend the disk iterator into any tree 
 
2359
                        # paths.
 
2360
                        if current_path_info[2] == 'tree-reference':
 
2361
                            del current_dir_info[1][path_index]
 
2362
                            path_index -= 1
 
2363
                        path_index += 1
 
2364
                        if path_index < len(current_dir_info[1]):
 
2365
                            current_path_info = current_dir_info[1][path_index]
 
2366
                            if current_path_info[2] == 'directory':
 
2367
                                if self.target._directory_is_tree_reference(
 
2368
                                    current_path_info[0].decode('utf8')):
 
2369
                                    current_path_info = current_path_info[:2] + \
 
2370
                                        ('tree-reference',) + current_path_info[3:]
 
2371
                        else:
 
2372
                            current_path_info = None
 
2373
                        path_handled = False
 
2374
                    else:
 
2375
                        advance_path = True # reset the advance flagg.
 
2376
                if current_block is not None:
 
2377
                    block_index += 1
 
2378
                    if (block_index < len(state._dirblocks) and
 
2379
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2380
                        current_block = state._dirblocks[block_index]
 
2381
                    else:
 
2382
                        current_block = None
 
2383
                if current_dir_info is not None:
 
2384
                    try:
 
2385
                        current_dir_info = dir_iterator.next()
 
2386
                    except StopIteration:
 
2387
                        current_dir_info = None
 
2388
 
 
2389
 
 
2390
    @staticmethod
 
2391
    def is_compatible(source, target):
 
2392
        # the target must be a dirstate working tree
 
2393
        if not isinstance(target, WorkingTree4):
 
2394
            return False
 
2395
        # the source must be a revtreee or dirstate rev tree.
 
2396
        if not isinstance(source,
 
2397
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2398
            return False
 
2399
        # the source revid must be in the target dirstate
 
2400
        if not (source._revision_id == NULL_REVISION or
 
2401
            source._revision_id in target.get_parent_ids()):
 
2402
            # TODO: what about ghosts? it may well need to 
 
2403
            # check for them explicitly.
 
2404
            return False
 
2405
        return True
 
2406
 
 
2407
InterTree.register_optimiser(InterDirStateTree)
 
2408
 
 
2409
 
 
2410
class Converter3to4(object):
 
2411
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2412
 
 
2413
    def __init__(self):
 
2414
        self.target_format = WorkingTreeFormat4()
 
2415
 
 
2416
    def convert(self, tree):
 
2417
        # lock the control files not the tree, so that we dont get tree
 
2418
        # on-unlock behaviours, and so that noone else diddles with the 
 
2419
        # tree during upgrade.
 
2420
        tree._control_files.lock_write()
 
2421
        try:
 
2422
            tree.read_working_inventory()
 
2423
            self.create_dirstate_data(tree)
 
2424
            self.update_format(tree)
 
2425
            self.remove_xml_files(tree)
 
2426
        finally:
 
2427
            tree._control_files.unlock()
 
2428
 
 
2429
    def create_dirstate_data(self, tree):
 
2430
        """Create the dirstate based data for tree."""
 
2431
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2432
            ).local_abspath('dirstate')
 
2433
        state = dirstate.DirState.from_tree(tree, local_path)
 
2434
        state.save()
 
2435
        state.unlock()
 
2436
 
 
2437
    def remove_xml_files(self, tree):
 
2438
        """Remove the oldformat 3 data."""
 
2439
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2440
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2441
            'pending-merges', 'stat-cache']:
 
2442
            try:
 
2443
                transport.delete(path)
 
2444
            except errors.NoSuchFile:
 
2445
                # some files are optional - just deal.
 
2446
                pass
 
2447
 
 
2448
    def update_format(self, tree):
 
2449
        """Change the format marker."""
 
2450
        tree._control_files.put_utf8('format',
 
2451
            self.target_format.get_format_string())