/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2007-10-30 21:46:13 UTC
  • mfrom: (2949.2.1 revert)
  • Revision ID: pqm@pqm.ubuntu.com-20071030214613-c0ojrpmnm21da068
(robertc) Stop reading the basis inventory from the repository twice during revert and manage basis tree locks better. (Robert Collins)

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
    @needs_read_lock
 
1217
    def _validate(self):
 
1218
        self._dirstate._validate()
 
1219
 
 
1220
    @needs_tree_write_lock
 
1221
    def _write_inventory(self, inv):
 
1222
        """Write inventory as the current inventory."""
 
1223
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1224
        self.current_dirstate().set_state_from_inventory(inv)
 
1225
        self._make_dirty(reset_inventory=False)
 
1226
        if self._inventory is not None:
 
1227
            self._inventory = inv
 
1228
        self.flush()
 
1229
 
 
1230
 
 
1231
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1232
    """The first consolidated dirstate working tree format.
 
1233
 
 
1234
    This format:
 
1235
        - exists within a metadir controlling .bzr
 
1236
        - includes an explicit version marker for the workingtree control
 
1237
          files, separate from the BzrDir format
 
1238
        - modifies the hash cache format
 
1239
        - is new in bzr 0.15
 
1240
        - uses a LockDir to guard access to it.
 
1241
    """
 
1242
 
 
1243
    upgrade_recommended = False
 
1244
 
 
1245
    def get_format_string(self):
 
1246
        """See WorkingTreeFormat.get_format_string()."""
 
1247
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1248
 
 
1249
    def get_format_description(self):
 
1250
        """See WorkingTreeFormat.get_format_description()."""
 
1251
        return "Working tree format 4"
 
1252
 
 
1253
    def initialize(self, a_bzrdir, revision_id=None):
 
1254
        """See WorkingTreeFormat.initialize().
 
1255
 
 
1256
        :param revision_id: allows creating a working tree at a different
 
1257
        revision than the branch is at.
 
1258
 
 
1259
        These trees get an initial random root id, if their repository supports
 
1260
        rich root data, TREE_ROOT otherwise.
 
1261
        """
 
1262
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1263
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1264
        transport = a_bzrdir.get_workingtree_transport(self)
 
1265
        control_files = self._open_control_files(a_bzrdir)
 
1266
        control_files.create_lock()
 
1267
        control_files.lock_write()
 
1268
        control_files.put_utf8('format', self.get_format_string())
 
1269
        branch = a_bzrdir.open_branch()
 
1270
        if revision_id is None:
 
1271
            revision_id = branch.last_revision()
 
1272
        local_path = transport.local_abspath('dirstate')
 
1273
        # write out new dirstate (must exist when we create the tree)
 
1274
        state = dirstate.DirState.initialize(local_path)
 
1275
        state.unlock()
 
1276
        del state
 
1277
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1278
                         branch,
 
1279
                         _format=self,
 
1280
                         _bzrdir=a_bzrdir,
 
1281
                         _control_files=control_files)
 
1282
        wt._new_tree()
 
1283
        wt.lock_tree_write()
 
1284
        try:
 
1285
            if revision_id in (None, NULL_REVISION):
 
1286
                if branch.repository.supports_rich_root():
 
1287
                    wt._set_root_id(generate_ids.gen_root_id())
 
1288
                else:
 
1289
                    wt._set_root_id(ROOT_ID)
 
1290
                wt.flush()
 
1291
            wt.set_last_revision(revision_id)
 
1292
            wt.flush()
 
1293
            basis = wt.basis_tree()
 
1294
            basis.lock_read()
 
1295
            # if the basis has a root id we have to use that; otherwise we use
 
1296
            # a new random one
 
1297
            basis_root_id = basis.get_root_id()
 
1298
            if basis_root_id is not None:
 
1299
                wt._set_root_id(basis_root_id)
 
1300
                wt.flush()
 
1301
            transform.build_tree(basis, wt)
 
1302
            basis.unlock()
 
1303
        finally:
 
1304
            control_files.unlock()
 
1305
            wt.unlock()
 
1306
        return wt
 
1307
 
 
1308
    def _open(self, a_bzrdir, control_files):
 
1309
        """Open the tree itself.
 
1310
 
 
1311
        :param a_bzrdir: the dir for the tree.
 
1312
        :param control_files: the control files for the tree.
 
1313
        """
 
1314
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1315
                           branch=a_bzrdir.open_branch(),
 
1316
                           _format=self,
 
1317
                           _bzrdir=a_bzrdir,
 
1318
                           _control_files=control_files)
 
1319
 
 
1320
    def __get_matchingbzrdir(self):
 
1321
        # please test against something that will let us do tree references
 
1322
        return bzrdir.format_registry.make_bzrdir(
 
1323
            'dirstate-with-subtree')
 
1324
 
 
1325
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1326
 
 
1327
 
 
1328
class DirStateRevisionTree(Tree):
 
1329
    """A revision tree pulling the inventory from a dirstate."""
 
1330
 
 
1331
    def __init__(self, dirstate, revision_id, repository):
 
1332
        self._dirstate = dirstate
 
1333
        self._revision_id = revision_id
 
1334
        self._repository = repository
 
1335
        self._inventory = None
 
1336
        self._locked = 0
 
1337
        self._dirstate_locked = False
 
1338
 
 
1339
    def __repr__(self):
 
1340
        return "<%s of %s in %s>" % \
 
1341
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1342
 
 
1343
    def annotate_iter(self, file_id,
 
1344
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1345
        """See Tree.annotate_iter"""
 
1346
        w = self._get_weave(file_id)
 
1347
        return w.annotate_iter(self.inventory[file_id].revision)
 
1348
 
 
1349
    def _get_ancestors(self, default_revision):
 
1350
        return set(self._repository.get_ancestry(self._revision_id,
 
1351
                                                 topo_sorted=False))
 
1352
    def _comparison_data(self, entry, path):
 
1353
        """See Tree._comparison_data."""
 
1354
        if entry is None:
 
1355
            return None, False, None
 
1356
        # trust the entry as RevisionTree does, but this may not be
 
1357
        # sensible: the entry might not have come from us?
 
1358
        return entry.kind, entry.executable, None
 
1359
 
 
1360
    def _file_size(self, entry, stat_value):
 
1361
        return entry.text_size
 
1362
 
 
1363
    def filter_unversioned_files(self, paths):
 
1364
        """Filter out paths that are not versioned.
 
1365
 
 
1366
        :return: set of paths.
 
1367
        """
 
1368
        pred = self.has_filename
 
1369
        return set((p for p in paths if not pred(p)))
 
1370
 
 
1371
    def get_root_id(self):
 
1372
        return self.path2id('')
 
1373
 
 
1374
    def _get_parent_index(self):
 
1375
        """Return the index in the dirstate referenced by this tree."""
 
1376
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1377
 
 
1378
    def _get_entry(self, file_id=None, path=None):
 
1379
        """Get the dirstate row for file_id or path.
 
1380
 
 
1381
        If either file_id or path is supplied, it is used as the key to lookup.
 
1382
        If both are supplied, the fastest lookup is used, and an error is
 
1383
        raised if they do not both point at the same row.
 
1384
        
 
1385
        :param file_id: An optional unicode file_id to be looked up.
 
1386
        :param path: An optional unicode path to be looked up.
 
1387
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1388
        """
 
1389
        if file_id is None and path is None:
 
1390
            raise errors.BzrError('must supply file_id or path')
 
1391
        if path is not None:
 
1392
            path = path.encode('utf8')
 
1393
        parent_index = self._get_parent_index()
 
1394
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1395
 
 
1396
    def _generate_inventory(self):
 
1397
        """Create and set self.inventory from the dirstate object.
 
1398
 
 
1399
        (So this is only called the first time the inventory is requested for
 
1400
        this tree; it then remains in memory until it's out of date.)
 
1401
 
 
1402
        This is relatively expensive: we have to walk the entire dirstate.
 
1403
        """
 
1404
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1405
            'dirstate revision tree'
 
1406
        # separate call for profiling - makes it clear where the costs are.
 
1407
        self._dirstate._read_dirblocks_if_needed()
 
1408
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1409
            'parent %s has disappeared from %s' % (
 
1410
            self._revision_id, self._dirstate.get_parent_ids())
 
1411
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1412
        # This is identical now to the WorkingTree _generate_inventory except
 
1413
        # for the tree index use.
 
1414
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1415
        current_id = root_key[2]
 
1416
        assert current_entry[parent_index][0] == 'd'
 
1417
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1418
        inv.root.revision = current_entry[parent_index][4]
 
1419
        # Turn some things into local variables
 
1420
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1421
        factory = entry_factory
 
1422
        utf8_decode = cache_utf8._utf8_decode
 
1423
        inv_byid = inv._byid
 
1424
        # we could do this straight out of the dirstate; it might be fast
 
1425
        # and should be profiled - RBC 20070216
 
1426
        parent_ies = {'' : inv.root}
 
1427
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1428
            dirname = block[0]
 
1429
            try:
 
1430
                parent_ie = parent_ies[dirname]
 
1431
            except KeyError:
 
1432
                # all the paths in this block are not versioned in this tree
 
1433
                continue
 
1434
            for key, entry in block[1]:
 
1435
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1436
                if minikind in ('a', 'r'): # absent, relocated
 
1437
                    # not this tree
 
1438
                    continue
 
1439
                name = key[1]
 
1440
                name_unicode = utf8_decode(name)[0]
 
1441
                file_id = key[2]
 
1442
                kind = minikind_to_kind[minikind]
 
1443
                inv_entry = factory[kind](file_id, name_unicode,
 
1444
                                          parent_ie.file_id)
 
1445
                inv_entry.revision = revid
 
1446
                if kind == 'file':
 
1447
                    inv_entry.executable = executable
 
1448
                    inv_entry.text_size = size
 
1449
                    inv_entry.text_sha1 = fingerprint
 
1450
                elif kind == 'directory':
 
1451
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1452
                elif kind == 'symlink':
 
1453
                    inv_entry.executable = False
 
1454
                    inv_entry.text_size = None
 
1455
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1456
                elif kind == 'tree-reference':
 
1457
                    inv_entry.reference_revision = fingerprint or None
 
1458
                else:
 
1459
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1460
                            % entry)
 
1461
                # These checks cost us around 40ms on a 55k entry tree
 
1462
                assert file_id not in inv_byid
 
1463
                assert name_unicode not in parent_ie.children
 
1464
                inv_byid[file_id] = inv_entry
 
1465
                parent_ie.children[name_unicode] = inv_entry
 
1466
        self._inventory = inv
 
1467
 
 
1468
    def get_file_mtime(self, file_id, path=None):
 
1469
        """Return the modification time for this record.
 
1470
 
 
1471
        We return the timestamp of the last-changed revision.
 
1472
        """
 
1473
        # Make sure the file exists
 
1474
        entry = self._get_entry(file_id, path=path)
 
1475
        if entry == (None, None): # do we raise?
 
1476
            return None
 
1477
        parent_index = self._get_parent_index()
 
1478
        last_changed_revision = entry[1][parent_index][4]
 
1479
        return self._repository.get_revision(last_changed_revision).timestamp
 
1480
 
 
1481
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1482
        entry = self._get_entry(file_id=file_id, path=path)
 
1483
        parent_index = self._get_parent_index()
 
1484
        parent_details = entry[1][parent_index]
 
1485
        if parent_details[0] == 'f':
 
1486
            return parent_details[1]
 
1487
        return None
 
1488
 
 
1489
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1490
    def get_weave(self, file_id):
 
1491
        return self._get_weave(file_id)
 
1492
 
 
1493
    def _get_weave(self, file_id):
 
1494
        return self._repository.weave_store.get_weave(file_id,
 
1495
                self._repository.get_transaction())
 
1496
 
 
1497
    def get_file(self, file_id, path=None):
 
1498
        return StringIO(self.get_file_text(file_id))
 
1499
 
 
1500
    def get_file_lines(self, file_id):
 
1501
        ie = self.inventory[file_id]
 
1502
        return self._get_weave(file_id).get_lines(ie.revision)
 
1503
 
 
1504
    def get_file_size(self, file_id):
 
1505
        return self.inventory[file_id].text_size
 
1506
 
 
1507
    def get_file_text(self, file_id):
 
1508
        return ''.join(self.get_file_lines(file_id))
 
1509
 
 
1510
    def get_reference_revision(self, file_id, path=None):
 
1511
        return self.inventory[file_id].reference_revision
 
1512
 
 
1513
    def iter_files_bytes(self, desired_files):
 
1514
        """See Tree.iter_files_bytes.
 
1515
 
 
1516
        This version is implemented on top of Repository.iter_files_bytes"""
 
1517
        parent_index = self._get_parent_index()
 
1518
        repo_desired_files = []
 
1519
        for file_id, identifier in desired_files:
 
1520
            entry = self._get_entry(file_id)
 
1521
            if entry == (None, None):
 
1522
                raise errors.NoSuchId(self, file_id)
 
1523
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1524
                                       identifier))
 
1525
        return self._repository.iter_files_bytes(repo_desired_files)
 
1526
 
 
1527
    def get_symlink_target(self, file_id):
 
1528
        entry = self._get_entry(file_id=file_id)
 
1529
        parent_index = self._get_parent_index()
 
1530
        if entry[1][parent_index][0] != 'l':
 
1531
            return None
 
1532
        else:
 
1533
            # At present, none of the tree implementations supports non-ascii
 
1534
            # symlink targets. So we will just assume that the dirstate path is
 
1535
            # correct.
 
1536
            return entry[1][parent_index][1]
 
1537
 
 
1538
    def get_revision_id(self):
 
1539
        """Return the revision id for this tree."""
 
1540
        return self._revision_id
 
1541
 
 
1542
    def _get_inventory(self):
 
1543
        if self._inventory is not None:
 
1544
            return self._inventory
 
1545
        self._must_be_locked()
 
1546
        self._generate_inventory()
 
1547
        return self._inventory
 
1548
 
 
1549
    inventory = property(_get_inventory,
 
1550
                         doc="Inventory of this Tree")
 
1551
 
 
1552
    def get_parent_ids(self):
 
1553
        """The parents of a tree in the dirstate are not cached."""
 
1554
        return self._repository.get_revision(self._revision_id).parent_ids
 
1555
 
 
1556
    def has_filename(self, filename):
 
1557
        return bool(self.path2id(filename))
 
1558
 
 
1559
    def kind(self, file_id):
 
1560
        return self.inventory[file_id].kind
 
1561
 
 
1562
    def path_content_summary(self, path):
 
1563
        """See Tree.path_content_summary."""
 
1564
        id = self.inventory.path2id(path)
 
1565
        if id is None:
 
1566
            return ('missing', None, None, None)
 
1567
        entry = self._inventory[id]
 
1568
        kind = entry.kind
 
1569
        if kind == 'file':
 
1570
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1571
        elif kind == 'symlink':
 
1572
            return (kind, None, None, entry.symlink_target)
 
1573
        else:
 
1574
            return (kind, None, None, None)
 
1575
 
 
1576
    def is_executable(self, file_id, path=None):
 
1577
        ie = self.inventory[file_id]
 
1578
        if ie.kind != "file":
 
1579
            return None
 
1580
        return ie.executable
 
1581
 
 
1582
    def list_files(self, include_root=False):
 
1583
        # We use a standard implementation, because DirStateRevisionTree is
 
1584
        # dealing with one of the parents of the current state
 
1585
        inv = self._get_inventory()
 
1586
        entries = inv.iter_entries()
 
1587
        if self.inventory.root is not None and not include_root:
 
1588
            entries.next()
 
1589
        for path, entry in entries:
 
1590
            yield path, 'V', entry.kind, entry.file_id, entry
 
1591
 
 
1592
    def lock_read(self):
 
1593
        """Lock the tree for a set of operations."""
 
1594
        if not self._locked:
 
1595
            self._repository.lock_read()
 
1596
            if self._dirstate._lock_token is None:
 
1597
                self._dirstate.lock_read()
 
1598
                self._dirstate_locked = True
 
1599
        self._locked += 1
 
1600
 
 
1601
    def _must_be_locked(self):
 
1602
        if not self._locked:
 
1603
            raise errors.ObjectNotLocked(self)
 
1604
 
 
1605
    @needs_read_lock
 
1606
    def path2id(self, path):
 
1607
        """Return the id for path in this tree."""
 
1608
        # lookup by path: faster than splitting and walking the ivnentory.
 
1609
        entry = self._get_entry(path=path)
 
1610
        if entry == (None, None):
 
1611
            return None
 
1612
        return entry[0][2]
 
1613
 
 
1614
    def unlock(self):
 
1615
        """Unlock, freeing any cache memory used during the lock."""
 
1616
        # outside of a lock, the inventory is suspect: release it.
 
1617
        self._locked -=1
 
1618
        if not self._locked:
 
1619
            self._inventory = None
 
1620
            self._locked = 0
 
1621
            if self._dirstate_locked:
 
1622
                self._dirstate.unlock()
 
1623
                self._dirstate_locked = False
 
1624
            self._repository.unlock()
 
1625
 
 
1626
    def walkdirs(self, prefix=""):
 
1627
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1628
        # implementation based on an inventory.  
 
1629
        # This should be cleaned up to use the much faster Dirstate code
 
1630
        # So for now, we just build up the parent inventory, and extract
 
1631
        # it the same way RevisionTree does.
 
1632
        _directory = 'directory'
 
1633
        inv = self._get_inventory()
 
1634
        top_id = inv.path2id(prefix)
 
1635
        if top_id is None:
 
1636
            pending = []
 
1637
        else:
 
1638
            pending = [(prefix, top_id)]
 
1639
        while pending:
 
1640
            dirblock = []
 
1641
            relpath, file_id = pending.pop()
 
1642
            # 0 - relpath, 1- file-id
 
1643
            if relpath:
 
1644
                relroot = relpath + '/'
 
1645
            else:
 
1646
                relroot = ""
 
1647
            # FIXME: stash the node in pending
 
1648
            entry = inv[file_id]
 
1649
            for name, child in entry.sorted_children():
 
1650
                toppath = relroot + name
 
1651
                dirblock.append((toppath, name, child.kind, None,
 
1652
                    child.file_id, child.kind
 
1653
                    ))
 
1654
            yield (relpath, entry.file_id), dirblock
 
1655
            # push the user specified dirs from dirblock
 
1656
            for dir in reversed(dirblock):
 
1657
                if dir[2] == _directory:
 
1658
                    pending.append((dir[0], dir[4]))
 
1659
 
 
1660
 
 
1661
class InterDirStateTree(InterTree):
 
1662
    """Fast path optimiser for changes_from with dirstate trees.
 
1663
    
 
1664
    This is used only when both trees are in the dirstate working file, and 
 
1665
    the source is any parent within the dirstate, and the destination is 
 
1666
    the current working tree of the same dirstate.
 
1667
    """
 
1668
    # this could be generalized to allow comparisons between any trees in the
 
1669
    # dirstate, and possibly between trees stored in different dirstates.
 
1670
 
 
1671
    def __init__(self, source, target):
 
1672
        super(InterDirStateTree, self).__init__(source, target)
 
1673
        if not InterDirStateTree.is_compatible(source, target):
 
1674
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1675
 
 
1676
    @staticmethod
 
1677
    def make_source_parent_tree(source, target):
 
1678
        """Change the source tree into a parent of the target."""
 
1679
        revid = source.commit('record tree')
 
1680
        target.branch.repository.fetch(source.branch.repository, revid)
 
1681
        target.set_parent_ids([revid])
 
1682
        return target.basis_tree(), target
 
1683
 
 
1684
    _matching_from_tree_format = WorkingTreeFormat4()
 
1685
    _matching_to_tree_format = WorkingTreeFormat4()
 
1686
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1687
 
 
1688
    def _iter_changes(self, include_unchanged=False,
 
1689
                      specific_files=None, pb=None, extra_trees=[],
 
1690
                      require_versioned=True, want_unversioned=False):
 
1691
        """Return the changes from source to target.
 
1692
 
 
1693
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1694
            for details.
 
1695
        :param specific_files: An optional list of file paths to restrict the
 
1696
            comparison to. When mapping filenames to ids, all matches in all
 
1697
            trees (including optional extra_trees) are used, and all children of
 
1698
            matched directories are included.
 
1699
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1700
            unchanged entries in the result.
 
1701
        :param extra_trees: An optional list of additional trees to use when
 
1702
            mapping the contents of specific_files (paths) to file_ids.
 
1703
        :param require_versioned: If True, all files in specific_files must be
 
1704
            versioned in one of source, target, extra_trees or
 
1705
            PathsNotVersionedError is raised.
 
1706
        :param want_unversioned: Should unversioned files be returned in the
 
1707
            output. An unversioned file is defined as one with (False, False)
 
1708
            for the versioned pair.
 
1709
        """
 
1710
        utf8_decode = cache_utf8._utf8_decode
 
1711
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1712
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1713
        # NB: show_status depends on being able to pass in non-versioned files
 
1714
        # and report them as unknown
 
1715
        # TODO: handle extra trees in the dirstate.
 
1716
        # TODO: handle comparisons as an empty tree as a different special
 
1717
        # case? mbp 20070226
 
1718
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1719
            or specific_files == []):
 
1720
            # we can't fast-path these cases (yet)
 
1721
            for f in super(InterDirStateTree, self)._iter_changes(
 
1722
                include_unchanged, specific_files, pb, extra_trees,
 
1723
                require_versioned, want_unversioned=want_unversioned):
 
1724
                yield f
 
1725
            return
 
1726
        parent_ids = self.target.get_parent_ids()
 
1727
        assert (self.source._revision_id in parent_ids), \
 
1728
                "revision {%s} is not stored in {%s}, but %s " \
 
1729
                "can only be used for trees stored in the dirstate" \
 
1730
                % (self.source._revision_id, self.target, self._iter_changes)
 
1731
        target_index = 0
 
1732
        if self.source._revision_id == NULL_REVISION:
 
1733
            source_index = None
 
1734
            indices = (target_index,)
 
1735
        else:
 
1736
            assert (self.source._revision_id in parent_ids), \
 
1737
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1738
                self.source._revision_id, parent_ids)
 
1739
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1740
            indices = (source_index,target_index)
 
1741
        # -- make all specific_files utf8 --
 
1742
        if specific_files:
 
1743
            specific_files_utf8 = set()
 
1744
            for path in specific_files:
 
1745
                specific_files_utf8.add(path.encode('utf8'))
 
1746
            specific_files = specific_files_utf8
 
1747
        else:
 
1748
            specific_files = set([''])
 
1749
        # -- specific_files is now a utf8 path set --
 
1750
        # -- get the state object and prepare it.
 
1751
        state = self.target.current_dirstate()
 
1752
        state._read_dirblocks_if_needed()
 
1753
        def _entries_for_path(path):
 
1754
            """Return a list with all the entries that match path for all ids.
 
1755
            """
 
1756
            dirname, basename = os.path.split(path)
 
1757
            key = (dirname, basename, '')
 
1758
            block_index, present = state._find_block_index_from_key(key)
 
1759
            if not present:
 
1760
                # the block which should contain path is absent.
 
1761
                return []
 
1762
            result = []
 
1763
            block = state._dirblocks[block_index][1]
 
1764
            entry_index, _ = state._find_entry_index(key, block)
 
1765
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1766
            while (entry_index < len(block) and
 
1767
                block[entry_index][0][0:2] == key[0:2]):
 
1768
                result.append(block[entry_index])
 
1769
                entry_index += 1
 
1770
            return result
 
1771
        if require_versioned:
 
1772
            # -- check all supplied paths are versioned in a search tree. --
 
1773
            all_versioned = True
 
1774
            for path in specific_files:
 
1775
                path_entries = _entries_for_path(path)
 
1776
                if not path_entries:
 
1777
                    # this specified path is not present at all: error
 
1778
                    all_versioned = False
 
1779
                    break
 
1780
                found_versioned = False
 
1781
                # for each id at this path
 
1782
                for entry in path_entries:
 
1783
                    # for each tree.
 
1784
                    for index in indices:
 
1785
                        if entry[1][index][0] != 'a': # absent
 
1786
                            found_versioned = True
 
1787
                            # all good: found a versioned cell
 
1788
                            break
 
1789
                if not found_versioned:
 
1790
                    # none of the indexes was not 'absent' at all ids for this
 
1791
                    # path.
 
1792
                    all_versioned = False
 
1793
                    break
 
1794
            if not all_versioned:
 
1795
                raise errors.PathsNotVersionedError(specific_files)
 
1796
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1797
        search_specific_files = set()
 
1798
        for path in specific_files:
 
1799
            other_specific_files = specific_files.difference(set([path]))
 
1800
            if not osutils.is_inside_any(other_specific_files, path):
 
1801
                # this is a top level path, we must check it.
 
1802
                search_specific_files.add(path)
 
1803
        # sketch: 
 
1804
        # compare source_index and target_index at or under each element of search_specific_files.
 
1805
        # follow the following comparison table. Note that we only want to do diff operations when
 
1806
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1807
        # for the target.
 
1808
        # cases:
 
1809
        # 
 
1810
        # Source | Target | disk | action
 
1811
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1812
        #        |        |      | diff check on source-target
 
1813
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1814
        #        |        |      | ???
 
1815
        #   r    |  a     |      | add source to search
 
1816
        #   r    |  a     |  a   | 
 
1817
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1818
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1819
        #   a    | fdlt   |      | add new id
 
1820
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1821
        #   a    |  a     |      | not present in either tree, skip
 
1822
        #   a    |  a     |  a   | not present in any tree, skip
 
1823
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1824
        #        |        |      | may not be selected by the users list of paths.
 
1825
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1826
        #        |        |      | may not be selected by the users list of paths.
 
1827
        #  fdlt  | fdlt   |      | content in both: diff them
 
1828
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1829
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1830
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1831
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1832
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1833
        #        |        |      | this id at the other path.
 
1834
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1835
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1836
        #        |        |      | this id at the other path.
 
1837
 
 
1838
        # for all search_indexs in each path at or under each element of
 
1839
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1840
        # relocated path as one to search if its not searched already. If the
 
1841
        # detail is not relocated, add the id.
 
1842
        searched_specific_files = set()
 
1843
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1844
        # Using a list so that we can access the values and change them in
 
1845
        # nested scope. Each one is [path, file_id, entry]
 
1846
        last_source_parent = [None, None]
 
1847
        last_target_parent = [None, None]
 
1848
 
 
1849
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1850
 
 
1851
        # Just a sentry, so that _process_entry can say that this
 
1852
        # record is handled, but isn't interesting to process (unchanged)
 
1853
        uninteresting = object()
 
1854
 
 
1855
 
 
1856
        old_dirname_to_file_id = {}
 
1857
        new_dirname_to_file_id = {}
 
1858
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1859
        #       keeping a cache of directories that we have seen.
 
1860
 
 
1861
        def _process_entry(entry, path_info):
 
1862
            """Compare an entry and real disk to generate delta information.
 
1863
 
 
1864
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1865
                the path of entry. If None, then the path is considered absent.
 
1866
                (Perhaps we should pass in a concrete entry for this ?)
 
1867
                Basename is returned as a utf8 string because we expect this
 
1868
                tuple will be ignored, and don't want to take the time to
 
1869
                decode.
 
1870
            :return: None if these don't match
 
1871
                     A tuple of information about the change, or
 
1872
                     the object 'uninteresting' if these match, but are
 
1873
                     basically identical.
 
1874
            """
 
1875
            if source_index is None:
 
1876
                source_details = NULL_PARENT_DETAILS
 
1877
            else:
 
1878
                source_details = entry[1][source_index]
 
1879
            target_details = entry[1][target_index]
 
1880
            target_minikind = target_details[0]
 
1881
            if path_info is not None and target_minikind in 'fdlt':
 
1882
                assert target_index == 0
 
1883
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1884
                                                  stat_value=path_info[3])
 
1885
                # The entry may have been modified by update_entry
 
1886
                target_details = entry[1][target_index]
 
1887
                target_minikind = target_details[0]
 
1888
            else:
 
1889
                link_or_sha1 = None
 
1890
            file_id = entry[0][2]
 
1891
            source_minikind = source_details[0]
 
1892
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1893
                # claimed content in both: diff
 
1894
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1895
                #        |        |      | diff check on source-target
 
1896
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1897
                #        |        |      | ???
 
1898
                if source_minikind in 'r':
 
1899
                    # add the source to the search path to find any children it
 
1900
                    # has.  TODO ? : only add if it is a container ?
 
1901
                    if not osutils.is_inside_any(searched_specific_files,
 
1902
                                                 source_details[1]):
 
1903
                        search_specific_files.add(source_details[1])
 
1904
                    # generate the old path; this is needed for stating later
 
1905
                    # as well.
 
1906
                    old_path = source_details[1]
 
1907
                    old_dirname, old_basename = os.path.split(old_path)
 
1908
                    path = pathjoin(entry[0][0], entry[0][1])
 
1909
                    old_entry = state._get_entry(source_index,
 
1910
                                                 path_utf8=old_path)
 
1911
                    # update the source details variable to be the real
 
1912
                    # location.
 
1913
                    source_details = old_entry[1][source_index]
 
1914
                    source_minikind = source_details[0]
 
1915
                else:
 
1916
                    old_dirname = entry[0][0]
 
1917
                    old_basename = entry[0][1]
 
1918
                    old_path = path = None
 
1919
                if path_info is None:
 
1920
                    # the file is missing on disk, show as removed.
 
1921
                    content_change = True
 
1922
                    target_kind = None
 
1923
                    target_exec = False
 
1924
                else:
 
1925
                    # source and target are both versioned and disk file is present.
 
1926
                    target_kind = path_info[2]
 
1927
                    if target_kind == 'directory':
 
1928
                        if path is None:
 
1929
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1930
                        new_dirname_to_file_id[path] = file_id
 
1931
                        if source_minikind != 'd':
 
1932
                            content_change = True
 
1933
                        else:
 
1934
                            # directories have no fingerprint
 
1935
                            content_change = False
 
1936
                        target_exec = False
 
1937
                    elif target_kind == 'file':
 
1938
                        if source_minikind != 'f':
 
1939
                            content_change = True
 
1940
                        else:
 
1941
                            # We could check the size, but we already have the
 
1942
                            # sha1 hash.
 
1943
                            content_change = (link_or_sha1 != source_details[1])
 
1944
                        # Target details is updated at update_entry time
 
1945
                        if use_filesystem_for_exec:
 
1946
                            # We don't need S_ISREG here, because we are sure
 
1947
                            # we are dealing with a file.
 
1948
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1949
                        else:
 
1950
                            target_exec = target_details[3]
 
1951
                    elif target_kind == 'symlink':
 
1952
                        if source_minikind != 'l':
 
1953
                            content_change = True
 
1954
                        else:
 
1955
                            content_change = (link_or_sha1 != source_details[1])
 
1956
                        target_exec = False
 
1957
                    elif target_kind == 'tree-reference':
 
1958
                        if source_minikind != 't':
 
1959
                            content_change = True
 
1960
                        else:
 
1961
                            content_change = False
 
1962
                        target_exec = False
 
1963
                    else:
 
1964
                        raise Exception, "unknown kind %s" % path_info[2]
 
1965
                if source_minikind == 'd':
 
1966
                    if path is None:
 
1967
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1968
                    old_dirname_to_file_id[old_path] = file_id
 
1969
                # parent id is the entry for the path in the target tree
 
1970
                if old_dirname == last_source_parent[0]:
 
1971
                    source_parent_id = last_source_parent[1]
 
1972
                else:
 
1973
                    try:
 
1974
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1975
                    except KeyError:
 
1976
                        source_parent_entry = state._get_entry(source_index,
 
1977
                                                               path_utf8=old_dirname)
 
1978
                        source_parent_id = source_parent_entry[0][2]
 
1979
                    if source_parent_id == entry[0][2]:
 
1980
                        # This is the root, so the parent is None
 
1981
                        source_parent_id = None
 
1982
                    else:
 
1983
                        last_source_parent[0] = old_dirname
 
1984
                        last_source_parent[1] = source_parent_id
 
1985
                new_dirname = entry[0][0]
 
1986
                if new_dirname == last_target_parent[0]:
 
1987
                    target_parent_id = last_target_parent[1]
 
1988
                else:
 
1989
                    try:
 
1990
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1991
                    except KeyError:
 
1992
                        # TODO: We don't always need to do the lookup, because the
 
1993
                        #       parent entry will be the same as the source entry.
 
1994
                        target_parent_entry = state._get_entry(target_index,
 
1995
                                                               path_utf8=new_dirname)
 
1996
                        assert target_parent_entry != (None, None), (
 
1997
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1998
                            % (new_dirname, entry))
 
1999
                        target_parent_id = target_parent_entry[0][2]
 
2000
                    if target_parent_id == entry[0][2]:
 
2001
                        # This is the root, so the parent is None
 
2002
                        target_parent_id = None
 
2003
                    else:
 
2004
                        last_target_parent[0] = new_dirname
 
2005
                        last_target_parent[1] = target_parent_id
 
2006
 
 
2007
                source_exec = source_details[3]
 
2008
                if (include_unchanged
 
2009
                    or content_change
 
2010
                    or source_parent_id != target_parent_id
 
2011
                    or old_basename != entry[0][1]
 
2012
                    or source_exec != target_exec
 
2013
                    ):
 
2014
                    if old_path is None:
 
2015
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2016
                        old_path_u = utf8_decode(old_path)[0]
 
2017
                        path_u = old_path_u
 
2018
                    else:
 
2019
                        old_path_u = utf8_decode(old_path)[0]
 
2020
                        if old_path == path:
 
2021
                            path_u = old_path_u
 
2022
                        else:
 
2023
                            path_u = utf8_decode(path)[0]
 
2024
                    source_kind = _minikind_to_kind[source_minikind]
 
2025
                    return (entry[0][2],
 
2026
                           (old_path_u, path_u),
 
2027
                           content_change,
 
2028
                           (True, True),
 
2029
                           (source_parent_id, target_parent_id),
 
2030
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2031
                           (source_kind, target_kind),
 
2032
                           (source_exec, target_exec))
 
2033
                else:
 
2034
                    return uninteresting
 
2035
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2036
                # looks like a new file
 
2037
                if path_info is not None:
 
2038
                    path = pathjoin(entry[0][0], entry[0][1])
 
2039
                    # parent id is the entry for the path in the target tree
 
2040
                    # TODO: these are the same for an entire directory: cache em.
 
2041
                    parent_id = state._get_entry(target_index,
 
2042
                                                 path_utf8=entry[0][0])[0][2]
 
2043
                    if parent_id == entry[0][2]:
 
2044
                        parent_id = None
 
2045
                    if use_filesystem_for_exec:
 
2046
                        # We need S_ISREG here, because we aren't sure if this
 
2047
                        # is a file or not.
 
2048
                        target_exec = bool(
 
2049
                            stat.S_ISREG(path_info[3].st_mode)
 
2050
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2051
                    else:
 
2052
                        target_exec = target_details[3]
 
2053
                    return (entry[0][2],
 
2054
                           (None, utf8_decode(path)[0]),
 
2055
                           True,
 
2056
                           (False, True),
 
2057
                           (None, parent_id),
 
2058
                           (None, utf8_decode(entry[0][1])[0]),
 
2059
                           (None, path_info[2]),
 
2060
                           (None, target_exec))
 
2061
                else:
 
2062
                    # but its not on disk: we deliberately treat this as just
 
2063
                    # never-present. (Why ?! - RBC 20070224)
 
2064
                    pass
 
2065
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2066
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2067
                # if its still on disk, *and* theres no other entry at this
 
2068
                # path [we dont know this in this routine at the moment -
 
2069
                # perhaps we should change this - then it would be an unknown.
 
2070
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2071
                # parent id is the entry for the path in the target tree
 
2072
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2073
                if parent_id == entry[0][2]:
 
2074
                    parent_id = None
 
2075
                return (entry[0][2],
 
2076
                       (utf8_decode(old_path)[0], None),
 
2077
                       True,
 
2078
                       (True, False),
 
2079
                       (parent_id, None),
 
2080
                       (utf8_decode(entry[0][1])[0], None),
 
2081
                       (_minikind_to_kind[source_minikind], None),
 
2082
                       (source_details[3], None))
 
2083
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2084
                # a rename; could be a true rename, or a rename inherited from
 
2085
                # a renamed parent. TODO: handle this efficiently. Its not
 
2086
                # common case to rename dirs though, so a correct but slow
 
2087
                # implementation will do.
 
2088
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2089
                    search_specific_files.add(target_details[1])
 
2090
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2091
                # neither of the selected trees contain this file,
 
2092
                # so skip over it. This is not currently directly tested, but
 
2093
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2094
                pass
 
2095
            else:
 
2096
                raise AssertionError("don't know how to compare "
 
2097
                    "source_minikind=%r, target_minikind=%r"
 
2098
                    % (source_minikind, target_minikind))
 
2099
                ## import pdb;pdb.set_trace()
 
2100
            return None
 
2101
 
 
2102
        while search_specific_files:
 
2103
            # TODO: the pending list should be lexically sorted?  the
 
2104
            # interface doesn't require it.
 
2105
            current_root = search_specific_files.pop()
 
2106
            current_root_unicode = current_root.decode('utf8')
 
2107
            searched_specific_files.add(current_root)
 
2108
            # process the entries for this containing directory: the rest will be
 
2109
            # found by their parents recursively.
 
2110
            root_entries = _entries_for_path(current_root)
 
2111
            root_abspath = self.target.abspath(current_root_unicode)
 
2112
            try:
 
2113
                root_stat = os.lstat(root_abspath)
 
2114
            except OSError, e:
 
2115
                if e.errno == errno.ENOENT:
 
2116
                    # the path does not exist: let _process_entry know that.
 
2117
                    root_dir_info = None
 
2118
                else:
 
2119
                    # some other random error: hand it up.
 
2120
                    raise
 
2121
            else:
 
2122
                root_dir_info = ('', current_root,
 
2123
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2124
                    root_abspath)
 
2125
                if root_dir_info[2] == 'directory':
 
2126
                    if self.target._directory_is_tree_reference(
 
2127
                        current_root.decode('utf8')):
 
2128
                        root_dir_info = root_dir_info[:2] + \
 
2129
                            ('tree-reference',) + root_dir_info[3:]
 
2130
 
 
2131
            if not root_entries and not root_dir_info:
 
2132
                # this specified path is not present at all, skip it.
 
2133
                continue
 
2134
            path_handled = False
 
2135
            for entry in root_entries:
 
2136
                result = _process_entry(entry, root_dir_info)
 
2137
                if result is not None:
 
2138
                    path_handled = True
 
2139
                    if result is not uninteresting:
 
2140
                        yield result
 
2141
            if want_unversioned and not path_handled and root_dir_info:
 
2142
                new_executable = bool(
 
2143
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2144
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2145
                yield (None,
 
2146
                       (None, current_root_unicode),
 
2147
                       True,
 
2148
                       (False, False),
 
2149
                       (None, None),
 
2150
                       (None, splitpath(current_root_unicode)[-1]),
 
2151
                       (None, root_dir_info[2]),
 
2152
                       (None, new_executable)
 
2153
                      )
 
2154
            initial_key = (current_root, '', '')
 
2155
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2156
            if block_index == 0:
 
2157
                # we have processed the total root already, but because the
 
2158
                # initial key matched it we should skip it here.
 
2159
                block_index +=1
 
2160
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2161
                current_dir_info = None
 
2162
            else:
 
2163
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2164
                try:
 
2165
                    current_dir_info = dir_iterator.next()
 
2166
                except OSError, e:
 
2167
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2168
                    # python 2.5 has e.errno == EINVAL,
 
2169
                    #            and e.winerror == ERROR_DIRECTORY
 
2170
                    e_winerror = getattr(e, 'winerror', None)
 
2171
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2172
                    # there may be directories in the inventory even though
 
2173
                    # this path is not a file on disk: so mark it as end of
 
2174
                    # iterator
 
2175
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2176
                        current_dir_info = None
 
2177
                    elif (sys.platform == 'win32'
 
2178
                          and (e.errno in win_errors
 
2179
                               or e_winerror in win_errors)):
 
2180
                        current_dir_info = None
 
2181
                    else:
 
2182
                        raise
 
2183
                else:
 
2184
                    if current_dir_info[0][0] == '':
 
2185
                        # remove .bzr from iteration
 
2186
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2187
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2188
                        del current_dir_info[1][bzr_index]
 
2189
            # walk until both the directory listing and the versioned metadata
 
2190
            # are exhausted. 
 
2191
            if (block_index < len(state._dirblocks) and
 
2192
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2193
                current_block = state._dirblocks[block_index]
 
2194
            else:
 
2195
                current_block = None
 
2196
            while (current_dir_info is not None or
 
2197
                   current_block is not None):
 
2198
                if (current_dir_info and current_block
 
2199
                    and current_dir_info[0][0] != current_block[0]):
 
2200
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2201
                        # filesystem data refers to paths not covered by the dirblock.
 
2202
                        # this has two possibilities:
 
2203
                        # A) it is versioned but empty, so there is no block for it
 
2204
                        # B) it is not versioned.
 
2205
 
 
2206
                        # if (A) then we need to recurse into it to check for
 
2207
                        # new unknown files or directories.
 
2208
                        # if (B) then we should ignore it, because we don't
 
2209
                        # recurse into unknown directories.
 
2210
                        path_index = 0
 
2211
                        while path_index < len(current_dir_info[1]):
 
2212
                                current_path_info = current_dir_info[1][path_index]
 
2213
                                if want_unversioned:
 
2214
                                    if current_path_info[2] == 'directory':
 
2215
                                        if self.target._directory_is_tree_reference(
 
2216
                                            current_path_info[0].decode('utf8')):
 
2217
                                            current_path_info = current_path_info[:2] + \
 
2218
                                                ('tree-reference',) + current_path_info[3:]
 
2219
                                    new_executable = bool(
 
2220
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2221
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2222
                                    yield (None,
 
2223
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2224
                                        True,
 
2225
                                        (False, False),
 
2226
                                        (None, None),
 
2227
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2228
                                        (None, current_path_info[2]),
 
2229
                                        (None, new_executable))
 
2230
                                # dont descend into this unversioned path if it is
 
2231
                                # a dir
 
2232
                                if current_path_info[2] in ('directory',
 
2233
                                                            'tree-reference'):
 
2234
                                    del current_dir_info[1][path_index]
 
2235
                                    path_index -= 1
 
2236
                                path_index += 1
 
2237
 
 
2238
                        # This dir info has been handled, go to the next
 
2239
                        try:
 
2240
                            current_dir_info = dir_iterator.next()
 
2241
                        except StopIteration:
 
2242
                            current_dir_info = None
 
2243
                    else:
 
2244
                        # We have a dirblock entry for this location, but there
 
2245
                        # is no filesystem path for this. This is most likely
 
2246
                        # because a directory was removed from the disk.
 
2247
                        # We don't have to report the missing directory,
 
2248
                        # because that should have already been handled, but we
 
2249
                        # need to handle all of the files that are contained
 
2250
                        # within.
 
2251
                        for current_entry in current_block[1]:
 
2252
                            # entry referring to file not present on disk.
 
2253
                            # advance the entry only, after processing.
 
2254
                            result = _process_entry(current_entry, None)
 
2255
                            if result is not None:
 
2256
                                if result is not uninteresting:
 
2257
                                    yield result
 
2258
                        block_index +=1
 
2259
                        if (block_index < len(state._dirblocks) and
 
2260
                            osutils.is_inside(current_root,
 
2261
                                              state._dirblocks[block_index][0])):
 
2262
                            current_block = state._dirblocks[block_index]
 
2263
                        else:
 
2264
                            current_block = None
 
2265
                    continue
 
2266
                entry_index = 0
 
2267
                if current_block and entry_index < len(current_block[1]):
 
2268
                    current_entry = current_block[1][entry_index]
 
2269
                else:
 
2270
                    current_entry = None
 
2271
                advance_entry = True
 
2272
                path_index = 0
 
2273
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2274
                    current_path_info = current_dir_info[1][path_index]
 
2275
                    if current_path_info[2] == 'directory':
 
2276
                        if self.target._directory_is_tree_reference(
 
2277
                            current_path_info[0].decode('utf8')):
 
2278
                            current_path_info = current_path_info[:2] + \
 
2279
                                ('tree-reference',) + current_path_info[3:]
 
2280
                else:
 
2281
                    current_path_info = None
 
2282
                advance_path = True
 
2283
                path_handled = False
 
2284
                while (current_entry is not None or
 
2285
                    current_path_info is not None):
 
2286
                    if current_entry is None:
 
2287
                        # the check for path_handled when the path is adnvaced
 
2288
                        # will yield this path if needed.
 
2289
                        pass
 
2290
                    elif current_path_info is None:
 
2291
                        # no path is fine: the per entry code will handle it.
 
2292
                        result = _process_entry(current_entry, current_path_info)
 
2293
                        if result is not None:
 
2294
                            if result is not uninteresting:
 
2295
                                yield result
 
2296
                    elif (current_entry[0][1] != current_path_info[1]
 
2297
                          or current_entry[1][target_index][0] in 'ar'):
 
2298
                        # The current path on disk doesn't match the dirblock
 
2299
                        # record. Either the dirblock is marked as absent, or
 
2300
                        # the file on disk is not present at all in the
 
2301
                        # dirblock. Either way, report about the dirblock
 
2302
                        # entry, and let other code handle the filesystem one.
 
2303
 
 
2304
                        # Compare the basename for these files to determine
 
2305
                        # which comes first
 
2306
                        if current_path_info[1] < current_entry[0][1]:
 
2307
                            # extra file on disk: pass for now, but only
 
2308
                            # increment the path, not the entry
 
2309
                            advance_entry = False
 
2310
                        else:
 
2311
                            # entry referring to file not present on disk.
 
2312
                            # advance the entry only, after processing.
 
2313
                            result = _process_entry(current_entry, None)
 
2314
                            if result is not None:
 
2315
                                if result is not uninteresting:
 
2316
                                    yield result
 
2317
                            advance_path = False
 
2318
                    else:
 
2319
                        result = _process_entry(current_entry, current_path_info)
 
2320
                        if result is not None:
 
2321
                            path_handled = True
 
2322
                            if result is not uninteresting:
 
2323
                                yield result
 
2324
                    if advance_entry and current_entry is not None:
 
2325
                        entry_index += 1
 
2326
                        if entry_index < len(current_block[1]):
 
2327
                            current_entry = current_block[1][entry_index]
 
2328
                        else:
 
2329
                            current_entry = None
 
2330
                    else:
 
2331
                        advance_entry = True # reset the advance flaga
 
2332
                    if advance_path and current_path_info is not None:
 
2333
                        if not path_handled:
 
2334
                            # unversioned in all regards
 
2335
                            if want_unversioned:
 
2336
                                new_executable = bool(
 
2337
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2338
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2339
                                yield (None,
 
2340
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2341
                                    True,
 
2342
                                    (False, False),
 
2343
                                    (None, None),
 
2344
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2345
                                    (None, current_path_info[2]),
 
2346
                                    (None, new_executable))
 
2347
                            # dont descend into this unversioned path if it is
 
2348
                            # a dir
 
2349
                            if current_path_info[2] in ('directory'):
 
2350
                                del current_dir_info[1][path_index]
 
2351
                                path_index -= 1
 
2352
                        # dont descend the disk iterator into any tree 
 
2353
                        # paths.
 
2354
                        if current_path_info[2] == 'tree-reference':
 
2355
                            del current_dir_info[1][path_index]
 
2356
                            path_index -= 1
 
2357
                        path_index += 1
 
2358
                        if path_index < len(current_dir_info[1]):
 
2359
                            current_path_info = current_dir_info[1][path_index]
 
2360
                            if current_path_info[2] == 'directory':
 
2361
                                if self.target._directory_is_tree_reference(
 
2362
                                    current_path_info[0].decode('utf8')):
 
2363
                                    current_path_info = current_path_info[:2] + \
 
2364
                                        ('tree-reference',) + current_path_info[3:]
 
2365
                        else:
 
2366
                            current_path_info = None
 
2367
                        path_handled = False
 
2368
                    else:
 
2369
                        advance_path = True # reset the advance flagg.
 
2370
                if current_block is not None:
 
2371
                    block_index += 1
 
2372
                    if (block_index < len(state._dirblocks) and
 
2373
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2374
                        current_block = state._dirblocks[block_index]
 
2375
                    else:
 
2376
                        current_block = None
 
2377
                if current_dir_info is not None:
 
2378
                    try:
 
2379
                        current_dir_info = dir_iterator.next()
 
2380
                    except StopIteration:
 
2381
                        current_dir_info = None
 
2382
 
 
2383
 
 
2384
    @staticmethod
 
2385
    def is_compatible(source, target):
 
2386
        # the target must be a dirstate working tree
 
2387
        if not isinstance(target, WorkingTree4):
 
2388
            return False
 
2389
        # the source must be a revtreee or dirstate rev tree.
 
2390
        if not isinstance(source,
 
2391
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2392
            return False
 
2393
        # the source revid must be in the target dirstate
 
2394
        if not (source._revision_id == NULL_REVISION or
 
2395
            source._revision_id in target.get_parent_ids()):
 
2396
            # TODO: what about ghosts? it may well need to 
 
2397
            # check for them explicitly.
 
2398
            return False
 
2399
        return True
 
2400
 
 
2401
InterTree.register_optimiser(InterDirStateTree)
 
2402
 
 
2403
 
 
2404
class Converter3to4(object):
 
2405
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2406
 
 
2407
    def __init__(self):
 
2408
        self.target_format = WorkingTreeFormat4()
 
2409
 
 
2410
    def convert(self, tree):
 
2411
        # lock the control files not the tree, so that we dont get tree
 
2412
        # on-unlock behaviours, and so that noone else diddles with the 
 
2413
        # tree during upgrade.
 
2414
        tree._control_files.lock_write()
 
2415
        try:
 
2416
            tree.read_working_inventory()
 
2417
            self.create_dirstate_data(tree)
 
2418
            self.update_format(tree)
 
2419
            self.remove_xml_files(tree)
 
2420
        finally:
 
2421
            tree._control_files.unlock()
 
2422
 
 
2423
    def create_dirstate_data(self, tree):
 
2424
        """Create the dirstate based data for tree."""
 
2425
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2426
            ).local_abspath('dirstate')
 
2427
        state = dirstate.DirState.from_tree(tree, local_path)
 
2428
        state.save()
 
2429
        state.unlock()
 
2430
 
 
2431
    def remove_xml_files(self, tree):
 
2432
        """Remove the oldformat 3 data."""
 
2433
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2434
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2435
            'pending-merges', 'stat-cache']:
 
2436
            try:
 
2437
                transport.delete(path)
 
2438
            except errors.NoSuchFile:
 
2439
                # some files are optional - just deal.
 
2440
                pass
 
2441
 
 
2442
    def update_format(self, tree):
 
2443
        """Change the format marker."""
 
2444
        tree._control_files.put_utf8('format',
 
2445
            self.target_format.get_format_string())