/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: Aaron Bentley
  • Date: 2008-02-24 16:42:13 UTC
  • mfrom: (3234 +trunk)
  • mto: This revision was merged to the branch mainline in revision 3235.
  • Revision ID: aaron@aaronbentley.com-20080224164213-eza1lzru5bwuwmmj
Merge with bzr.dev

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
        self._setup_directory_is_tree_reference()
 
155
        self._detect_case_handling()
 
156
 
 
157
    @needs_tree_write_lock
 
158
    def _add(self, files, ids, kinds):
 
159
        """See MutableTree._add."""
 
160
        state = self.current_dirstate()
 
161
        for f, file_id, kind in zip(files, ids, kinds):
 
162
            f = f.strip('/')
 
163
            if self.path2id(f):
 
164
                # special case tree root handling.
 
165
                if f == '' and self.path2id(f) == ROOT_ID:
 
166
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
167
                continue
 
168
            if file_id is None:
 
169
                file_id = generate_ids.gen_file_id(f)
 
170
            # deliberately add the file with no cached stat or sha1
 
171
            # - on the first access it will be gathered, and we can
 
172
            # always change this once tests are all passing.
 
173
            state.add(f, file_id, kind, None, '')
 
174
        self._make_dirty(reset_inventory=True)
 
175
 
 
176
    def _make_dirty(self, reset_inventory):
 
177
        """Make the tree state dirty.
 
178
 
 
179
        :param reset_inventory: True if the cached inventory should be removed
 
180
            (presuming there is one).
 
181
        """
 
182
        self._dirty = True
 
183
        if reset_inventory and self._inventory is not None:
 
184
            self._inventory = None
 
185
 
 
186
    @needs_tree_write_lock
 
187
    def add_reference(self, sub_tree):
 
188
        # use standard implementation, which calls back to self._add
 
189
        # 
 
190
        # So we don't store the reference_revision in the working dirstate,
 
191
        # it's just recorded at the moment of commit. 
 
192
        self._add_reference(sub_tree)
 
193
 
 
194
    def break_lock(self):
 
195
        """Break a lock if one is present from another instance.
 
196
 
 
197
        Uses the ui factory to ask for confirmation if the lock may be from
 
198
        an active process.
 
199
 
 
200
        This will probe the repository for its lock as well.
 
201
        """
 
202
        # if the dirstate is locked by an active process, reject the break lock
 
203
        # call.
 
204
        try:
 
205
            if self._dirstate is None:
 
206
                clear = True
 
207
            else:
 
208
                clear = False
 
209
            state = self._current_dirstate()
 
210
            if state._lock_token is not None:
 
211
                # we already have it locked. sheese, cant break our own lock.
 
212
                raise errors.LockActive(self.basedir)
 
213
            else:
 
214
                try:
 
215
                    # try for a write lock - need permission to get one anyhow
 
216
                    # to break locks.
 
217
                    state.lock_write()
 
218
                except errors.LockContention:
 
219
                    # oslocks fail when a process is still live: fail.
 
220
                    # TODO: get the locked lockdir info and give to the user to
 
221
                    # assist in debugging.
 
222
                    raise errors.LockActive(self.basedir)
 
223
                else:
 
224
                    state.unlock()
 
225
        finally:
 
226
            if clear:
 
227
                self._dirstate = None
 
228
        self._control_files.break_lock()
 
229
        self.branch.break_lock()
 
230
 
 
231
    def _comparison_data(self, entry, path):
 
232
        kind, executable, stat_value = \
 
233
            WorkingTree3._comparison_data(self, entry, path)
 
234
        # it looks like a plain directory, but it's really a reference -- see
 
235
        # also kind()
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
 
239
            kind = 'tree-reference'
 
240
        return kind, executable, stat_value
 
241
 
 
242
    @needs_write_lock
 
243
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
244
        # mark the tree as dirty post commit - commit
 
245
        # can change the current versioned list by doing deletes.
 
246
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
247
        self._make_dirty(reset_inventory=True)
 
248
        return result
 
249
 
 
250
    def current_dirstate(self):
 
251
        """Return the current dirstate object.
 
252
 
 
253
        This is not part of the tree interface and only exposed for ease of
 
254
        testing.
 
255
 
 
256
        :raises errors.NotWriteLocked: when not in a lock.
 
257
        """
 
258
        self._must_be_locked()
 
259
        return self._current_dirstate()
 
260
 
 
261
    def _current_dirstate(self):
 
262
        """Internal function that does not check lock status.
 
263
 
 
264
        This is needed for break_lock which also needs the dirstate.
 
265
        """
 
266
        if self._dirstate is not None:
 
267
            return self._dirstate
 
268
        local_path = self.bzrdir.get_workingtree_transport(None
 
269
            ).local_abspath('dirstate')
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
 
271
        return self._dirstate
 
272
 
 
273
    def filter_unversioned_files(self, paths):
 
274
        """Filter out paths that are versioned.
 
275
 
 
276
        :return: set of paths.
 
277
        """
 
278
        # TODO: make a generic multi-bisect routine roughly that should list
 
279
        # the paths, then process one half at a time recursively, and feed the
 
280
        # results of each bisect in further still
 
281
        paths = sorted(paths)
 
282
        result = set()
 
283
        state = self.current_dirstate()
 
284
        # TODO we want a paths_to_dirblocks helper I think
 
285
        for path in paths:
 
286
            dirname, basename = os.path.split(path.encode('utf8'))
 
287
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
288
                dirname, basename, 0)
 
289
            if not path_is_versioned:
 
290
                result.add(path)
 
291
        return result
 
292
 
 
293
    def flush(self):
 
294
        """Write all cached data to disk."""
 
295
        if self._control_files._lock_mode != 'w':
 
296
            raise errors.NotWriteLocked(self)
 
297
        self.current_dirstate().save()
 
298
        self._inventory = None
 
299
        self._dirty = False
 
300
 
 
301
    @needs_tree_write_lock
 
302
    def _gather_kinds(self, files, kinds):
 
303
        """See MutableTree._gather_kinds."""
 
304
        for pos, f in enumerate(files):
 
305
            if kinds[pos] is None:
 
306
                kinds[pos] = self._kind(f)
 
307
 
 
308
    def _generate_inventory(self):
 
309
        """Create and set self.inventory from the dirstate object.
 
310
        
 
311
        This is relatively expensive: we have to walk the entire dirstate.
 
312
        Ideally we would not, and can deprecate this function.
 
313
        """
 
314
        #: uncomment to trap on inventory requests.
 
315
        # import pdb;pdb.set_trace()
 
316
        state = self.current_dirstate()
 
317
        state._read_dirblocks_if_needed()
 
318
        root_key, current_entry = self._get_entry(path='')
 
319
        current_id = root_key[2]
 
320
        assert current_entry[0][0] == 'd' # directory
 
321
        inv = Inventory(root_id=current_id)
 
322
        # Turn some things into local variables
 
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
324
        factory = entry_factory
 
325
        utf8_decode = cache_utf8._utf8_decode
 
326
        inv_byid = inv._byid
 
327
        # we could do this straight out of the dirstate; it might be fast
 
328
        # and should be profiled - RBC 20070216
 
329
        parent_ies = {'' : inv.root}
 
330
        for block in state._dirblocks[1:]: # skip the root
 
331
            dirname = block[0]
 
332
            try:
 
333
                parent_ie = parent_ies[dirname]
 
334
            except KeyError:
 
335
                # all the paths in this block are not versioned in this tree
 
336
                continue
 
337
            for key, entry in block[1]:
 
338
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
339
                if minikind in ('a', 'r'): # absent, relocated
 
340
                    # a parent tree only entry
 
341
                    continue
 
342
                name = key[1]
 
343
                name_unicode = utf8_decode(name)[0]
 
344
                file_id = key[2]
 
345
                kind = minikind_to_kind[minikind]
 
346
                inv_entry = factory[kind](file_id, name_unicode,
 
347
                                          parent_ie.file_id)
 
348
                if kind == 'file':
 
349
                    # This is only needed on win32, where this is the only way
 
350
                    # we know the executable bit.
 
351
                    inv_entry.executable = executable
 
352
                    # not strictly needed: working tree
 
353
                    #inv_entry.text_size = size
 
354
                    #inv_entry.text_sha1 = sha1
 
355
                elif kind == 'directory':
 
356
                    # add this entry to the parent map.
 
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
358
                elif kind == 'tree-reference':
 
359
                    assert self._repo_supports_tree_reference, \
 
360
                        "repository of %r " \
 
361
                        "doesn't support tree references " \
 
362
                        "required by entry %r" \
 
363
                        % (self, name)
 
364
                    inv_entry.reference_revision = link_or_sha1 or None
 
365
                elif kind != 'symlink':
 
366
                    raise AssertionError("unknown kind %r" % kind)
 
367
                # These checks cost us around 40ms on a 55k entry tree
 
368
                assert file_id not in inv_byid, ('file_id %s already in'
 
369
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
370
                assert name_unicode not in parent_ie.children
 
371
                inv_byid[file_id] = inv_entry
 
372
                parent_ie.children[name_unicode] = inv_entry
 
373
        self._inventory = inv
 
374
 
 
375
    def _get_entry(self, file_id=None, path=None):
 
376
        """Get the dirstate row for file_id or path.
 
377
 
 
378
        If either file_id or path is supplied, it is used as the key to lookup.
 
379
        If both are supplied, the fastest lookup is used, and an error is
 
380
        raised if they do not both point at the same row.
 
381
        
 
382
        :param file_id: An optional unicode file_id to be looked up.
 
383
        :param path: An optional unicode path to be looked up.
 
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
385
        """
 
386
        if file_id is None and path is None:
 
387
            raise errors.BzrError('must supply file_id or path')
 
388
        state = self.current_dirstate()
 
389
        if path is not None:
 
390
            path = path.encode('utf8')
 
391
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
392
 
 
393
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
394
        # check file id is valid unconditionally.
 
395
        entry = self._get_entry(file_id=file_id, path=path)
 
396
        if entry[0] is None:
 
397
            raise errors.NoSuchId(self, file_id)
 
398
        if path is None:
 
399
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
400
 
 
401
        file_abspath = self.abspath(path)
 
402
        state = self.current_dirstate()
 
403
        if stat_value is None:
 
404
            try:
 
405
                stat_value = os.lstat(file_abspath)
 
406
            except OSError, e:
 
407
                if e.errno == errno.ENOENT:
 
408
                    return None
 
409
                else:
 
410
                    raise
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
 
413
        if entry[1][0][0] == 'f':
 
414
            return link_or_sha1
 
415
        return None
 
416
 
 
417
    def _get_inventory(self):
 
418
        """Get the inventory for the tree. This is only valid within a lock."""
 
419
        if 'evil' in debug.debug_flags:
 
420
            trace.mutter_callsite(2,
 
421
                "accessing .inventory forces a size of tree translation.")
 
422
        if self._inventory is not None:
 
423
            return self._inventory
 
424
        self._must_be_locked()
 
425
        self._generate_inventory()
 
426
        return self._inventory
 
427
 
 
428
    inventory = property(_get_inventory,
 
429
                         doc="Inventory of this Tree")
 
430
 
 
431
    @needs_read_lock
 
432
    def get_parent_ids(self):
 
433
        """See Tree.get_parent_ids.
 
434
        
 
435
        This implementation requests the ids list from the dirstate file.
 
436
        """
 
437
        return self.current_dirstate().get_parent_ids()
 
438
 
 
439
    def get_reference_revision(self, file_id, path=None):
 
440
        # referenced tree's revision is whatever's currently there
 
441
        return self.get_nested_tree(file_id, path).last_revision()
 
442
 
 
443
    def get_nested_tree(self, file_id, path=None):
 
444
        if path is None:
 
445
            path = self.id2path(file_id)
 
446
        # else: check file_id is at path?
 
447
        return WorkingTree.open(self.abspath(path))
 
448
 
 
449
    @needs_read_lock
 
450
    def get_root_id(self):
 
451
        """Return the id of this trees root"""
 
452
        return self._get_entry(path='')[0][2]
 
453
 
 
454
    def has_id(self, file_id):
 
455
        state = self.current_dirstate()
 
456
        row, parents = self._get_entry(file_id=file_id)
 
457
        if row is None:
 
458
            return False
 
459
        return osutils.lexists(pathjoin(
 
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
461
 
 
462
    @needs_read_lock
 
463
    def id2path(self, file_id):
 
464
        "Convert a file-id to a path."
 
465
        state = self.current_dirstate()
 
466
        entry = self._get_entry(file_id=file_id)
 
467
        if entry == (None, None):
 
468
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
469
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
470
        return path_utf8.decode('utf8')
 
471
 
 
472
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
473
        entry = self._get_entry(path=path)
 
474
        if entry == (None, None):
 
475
            return False # Missing entries are not executable
 
476
        return entry[1][0][3] # Executable?
 
477
 
 
478
    if not osutils.supports_executable():
 
479
        def is_executable(self, file_id, path=None):
 
480
            """Test if a file is executable or not.
 
481
 
 
482
            Note: The caller is expected to take a read-lock before calling this.
 
483
            """
 
484
            entry = self._get_entry(file_id=file_id, path=path)
 
485
            if entry == (None, None):
 
486
                return False
 
487
            return entry[1][0][3]
 
488
 
 
489
        _is_executable_from_path_and_stat = \
 
490
            _is_executable_from_path_and_stat_from_basis
 
491
    else:
 
492
        def is_executable(self, file_id, path=None):
 
493
            """Test if a file is executable or not.
 
494
 
 
495
            Note: The caller is expected to take a read-lock before calling this.
 
496
            """
 
497
            self._must_be_locked()
 
498
            if not path:
 
499
                path = self.id2path(file_id)
 
500
            mode = os.lstat(self.abspath(path)).st_mode
 
501
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
502
 
 
503
    @needs_read_lock
 
504
    def __iter__(self):
 
505
        """Iterate through file_ids for this tree.
 
506
 
 
507
        file_ids are in a WorkingTree if they are in the working inventory
 
508
        and the working file exists.
 
509
        """
 
510
        result = []
 
511
        for key, tree_details in self.current_dirstate()._iter_entries():
 
512
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
513
                # not relevant to the working tree
 
514
                continue
 
515
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
516
            if osutils.lexists(path):
 
517
                result.append(key[2])
 
518
        return iter(result)
 
519
 
 
520
    def iter_references(self):
 
521
        for key, tree_details in self.current_dirstate()._iter_entries():
 
522
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
523
                # not relevant to the working tree
 
524
                continue
 
525
            if not key[1]:
 
526
                # the root is not a reference.
 
527
                continue
 
528
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
529
            try:
 
530
                if self._kind(path) == 'tree-reference':
 
531
                    yield path, key[2]
 
532
            except errors.NoSuchFile:
 
533
                # path is missing on disk.
 
534
                continue
 
535
 
 
536
    def kind(self, file_id):
 
537
        """Return the kind of a file.
 
538
 
 
539
        This is always the actual kind that's on disk, regardless of what it
 
540
        was added as.
 
541
 
 
542
        Note: The caller is expected to take a read-lock before calling this.
 
543
        """
 
544
        relpath = self.id2path(file_id)
 
545
        assert relpath != None, \
 
546
            "path for id {%s} is None!" % file_id
 
547
        return self._kind(relpath)
 
548
 
 
549
    def _kind(self, relpath):
 
550
        abspath = self.abspath(relpath)
 
551
        kind = file_kind(abspath)
 
552
        if (self._repo_supports_tree_reference and
 
553
            kind == 'directory' and
 
554
            self._directory_is_tree_reference(relpath)):
 
555
            kind = 'tree-reference'
 
556
        return kind
 
557
 
 
558
    @needs_read_lock
 
559
    def _last_revision(self):
 
560
        """See Mutable.last_revision."""
 
561
        parent_ids = self.current_dirstate().get_parent_ids()
 
562
        if parent_ids:
 
563
            return parent_ids[0]
 
564
        else:
 
565
            return _mod_revision.NULL_REVISION
 
566
 
 
567
    def lock_read(self):
 
568
        """See Branch.lock_read, and WorkingTree.unlock."""
 
569
        self.branch.lock_read()
 
570
        try:
 
571
            self._control_files.lock_read()
 
572
            try:
 
573
                state = self.current_dirstate()
 
574
                if not state._lock_token:
 
575
                    state.lock_read()
 
576
                # set our support for tree references from the repository in
 
577
                # use.
 
578
                self._repo_supports_tree_reference = getattr(
 
579
                    self.branch.repository._format, "supports_tree_reference",
 
580
                    False)
 
581
            except:
 
582
                self._control_files.unlock()
 
583
                raise
 
584
        except:
 
585
            self.branch.unlock()
 
586
            raise
 
587
 
 
588
    def _lock_self_write(self):
 
589
        """This should be called after the branch is locked."""
 
590
        try:
 
591
            self._control_files.lock_write()
 
592
            try:
 
593
                state = self.current_dirstate()
 
594
                if not state._lock_token:
 
595
                    state.lock_write()
 
596
                # set our support for tree references from the repository in
 
597
                # use.
 
598
                self._repo_supports_tree_reference = getattr(
 
599
                    self.branch.repository._format, "supports_tree_reference",
 
600
                    False)
 
601
            except:
 
602
                self._control_files.unlock()
 
603
                raise
 
604
        except:
 
605
            self.branch.unlock()
 
606
            raise
 
607
 
 
608
    def lock_tree_write(self):
 
609
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
610
        self.branch.lock_read()
 
611
        self._lock_self_write()
 
612
 
 
613
    def lock_write(self):
 
614
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
615
        self.branch.lock_write()
 
616
        self._lock_self_write()
 
617
 
 
618
    @needs_tree_write_lock
 
619
    def move(self, from_paths, to_dir, after=False):
 
620
        """See WorkingTree.move()."""
 
621
        result = []
 
622
        if not from_paths:
 
623
            return result
 
624
 
 
625
        state = self.current_dirstate()
 
626
 
 
627
        assert not isinstance(from_paths, basestring)
 
628
        to_dir_utf8 = to_dir.encode('utf8')
 
629
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
630
        id_index = state._get_id_index()
 
631
        # check destination directory
 
632
        # get the details for it
 
633
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
634
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
635
        if not entry_present:
 
636
            raise errors.BzrMoveFailedError('', to_dir,
 
637
                errors.NotVersionedError(to_dir))
 
638
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
639
        # get a handle on the block itself.
 
640
        to_block_index = state._ensure_block(
 
641
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
642
        to_block = state._dirblocks[to_block_index]
 
643
        to_abs = self.abspath(to_dir)
 
644
        if not isdir(to_abs):
 
645
            raise errors.BzrMoveFailedError('',to_dir,
 
646
                errors.NotADirectory(to_abs))
 
647
 
 
648
        if to_entry[1][0][0] != 'd':
 
649
            raise errors.BzrMoveFailedError('',to_dir,
 
650
                errors.NotADirectory(to_abs))
 
651
 
 
652
        if self._inventory is not None:
 
653
            update_inventory = True
 
654
            inv = self.inventory
 
655
            to_dir_ie = inv[to_dir_id]
 
656
            to_dir_id = to_entry[0][2]
 
657
        else:
 
658
            update_inventory = False
 
659
 
 
660
        rollbacks = []
 
661
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
662
                     fingerprint, packed_stat, size,
 
663
                     to_block, to_key, to_path_utf8):
 
664
            state._make_absent(old_entry)
 
665
            from_key = old_entry[0]
 
666
            rollbacks.append(
 
667
                lambda:state.update_minimal(from_key,
 
668
                    minikind,
 
669
                    executable=executable,
 
670
                    fingerprint=fingerprint,
 
671
                    packed_stat=packed_stat,
 
672
                    size=size,
 
673
                    path_utf8=from_path_utf8))
 
674
            state.update_minimal(to_key,
 
675
                    minikind,
 
676
                    executable=executable,
 
677
                    fingerprint=fingerprint,
 
678
                    packed_stat=packed_stat,
 
679
                    size=size,
 
680
                    path_utf8=to_path_utf8)
 
681
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
682
            new_entry = to_block[1][added_entry_index]
 
683
            rollbacks.append(lambda:state._make_absent(new_entry))
 
684
 
 
685
        for from_rel in from_paths:
 
686
            # from_rel is 'pathinroot/foo/bar'
 
687
            from_rel_utf8 = from_rel.encode('utf8')
 
688
            from_dirname, from_tail = osutils.split(from_rel)
 
689
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
690
            from_entry = self._get_entry(path=from_rel)
 
691
            if from_entry == (None, None):
 
692
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
693
                    errors.NotVersionedError(path=str(from_rel)))
 
694
 
 
695
            from_id = from_entry[0][2]
 
696
            to_rel = pathjoin(to_dir, from_tail)
 
697
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
698
            item_to_entry = self._get_entry(path=to_rel)
 
699
            if item_to_entry != (None, None):
 
700
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
701
                    "Target is already versioned.")
 
702
 
 
703
            if from_rel == to_rel:
 
704
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
705
                    "Source and target are identical.")
 
706
 
 
707
            from_missing = not self.has_filename(from_rel)
 
708
            to_missing = not self.has_filename(to_rel)
 
709
            if after:
 
710
                move_file = False
 
711
            else:
 
712
                move_file = True
 
713
            if to_missing:
 
714
                if not move_file:
 
715
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
716
                        errors.NoSuchFile(path=to_rel,
 
717
                        extra="New file has not been created yet"))
 
718
                elif from_missing:
 
719
                    # neither path exists
 
720
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
721
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
722
            else:
 
723
                if from_missing: # implicitly just update our path mapping
 
724
                    move_file = False
 
725
                elif not after:
 
726
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
727
 
 
728
            rollbacks = []
 
729
            def rollback_rename():
 
730
                """A single rename has failed, roll it back."""
 
731
                # roll back everything, even if we encounter trouble doing one
 
732
                # of them.
 
733
                #
 
734
                # TODO: at least log the other exceptions rather than just
 
735
                # losing them mbp 20070307
 
736
                exc_info = None
 
737
                for rollback in reversed(rollbacks):
 
738
                    try:
 
739
                        rollback()
 
740
                    except Exception, e:
 
741
                        exc_info = sys.exc_info()
 
742
                if exc_info:
 
743
                    raise exc_info[0], exc_info[1], exc_info[2]
 
744
 
 
745
            # perform the disk move first - its the most likely failure point.
 
746
            if move_file:
 
747
                from_rel_abs = self.abspath(from_rel)
 
748
                to_rel_abs = self.abspath(to_rel)
 
749
                try:
 
750
                    osutils.rename(from_rel_abs, to_rel_abs)
 
751
                except OSError, e:
 
752
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
753
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
754
            try:
 
755
                # perform the rename in the inventory next if needed: its easy
 
756
                # to rollback
 
757
                if update_inventory:
 
758
                    # rename the entry
 
759
                    from_entry = inv[from_id]
 
760
                    current_parent = from_entry.parent_id
 
761
                    inv.rename(from_id, to_dir_id, from_tail)
 
762
                    rollbacks.append(
 
763
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
764
                # finally do the rename in the dirstate, which is a little
 
765
                # tricky to rollback, but least likely to need it.
 
766
                old_block_index, old_entry_index, dir_present, file_present = \
 
767
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
768
                old_block = state._dirblocks[old_block_index][1]
 
769
                old_entry = old_block[old_entry_index]
 
770
                from_key, old_entry_details = old_entry
 
771
                cur_details = old_entry_details[0]
 
772
                # remove the old row
 
773
                to_key = ((to_block[0],) + from_key[1:3])
 
774
                minikind = cur_details[0]
 
775
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
776
                         minikind=minikind,
 
777
                         executable=cur_details[3],
 
778
                         fingerprint=cur_details[1],
 
779
                         packed_stat=cur_details[4],
 
780
                         size=cur_details[2],
 
781
                         to_block=to_block,
 
782
                         to_key=to_key,
 
783
                         to_path_utf8=to_rel_utf8)
 
784
 
 
785
                if minikind == 'd':
 
786
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
787
                        """Recursively update all entries in this dirblock."""
 
788
                        assert from_dir != '', "renaming root not supported"
 
789
                        from_key = (from_dir, '')
 
790
                        from_block_idx, present = \
 
791
                            state._find_block_index_from_key(from_key)
 
792
                        if not present:
 
793
                            # This is the old record, if it isn't present, then
 
794
                            # there is theoretically nothing to update.
 
795
                            # (Unless it isn't present because of lazy loading,
 
796
                            # but we don't do that yet)
 
797
                            return
 
798
                        from_block = state._dirblocks[from_block_idx]
 
799
                        to_block_index, to_entry_index, _, _ = \
 
800
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
801
                        to_block_index = state._ensure_block(
 
802
                            to_block_index, to_entry_index, to_dir_utf8)
 
803
                        to_block = state._dirblocks[to_block_index]
 
804
 
 
805
                        # Grab a copy since move_one may update the list.
 
806
                        for entry in from_block[1][:]:
 
807
                            assert entry[0][0] == from_dir
 
808
                            cur_details = entry[1][0]
 
809
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
810
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
811
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
812
                            minikind = cur_details[0]
 
813
                            if minikind in 'ar':
 
814
                                # Deleted children of a renamed directory
 
815
                                # Do not need to be updated.
 
816
                                # Children that have been renamed out of this
 
817
                                # directory should also not be updated
 
818
                                continue
 
819
                            move_one(entry, from_path_utf8=from_path_utf8,
 
820
                                     minikind=minikind,
 
821
                                     executable=cur_details[3],
 
822
                                     fingerprint=cur_details[1],
 
823
                                     packed_stat=cur_details[4],
 
824
                                     size=cur_details[2],
 
825
                                     to_block=to_block,
 
826
                                     to_key=to_key,
 
827
                                     to_path_utf8=to_path_utf8)
 
828
                            if minikind == 'd':
 
829
                                # We need to move all the children of this
 
830
                                # entry
 
831
                                update_dirblock(from_path_utf8, to_key,
 
832
                                                to_path_utf8)
 
833
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
834
            except:
 
835
                rollback_rename()
 
836
                raise
 
837
            result.append((from_rel, to_rel))
 
838
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
839
            self._make_dirty(reset_inventory=False)
 
840
 
 
841
        return result
 
842
 
 
843
    def _must_be_locked(self):
 
844
        if not self._control_files._lock_count:
 
845
            raise errors.ObjectNotLocked(self)
 
846
 
 
847
    def _new_tree(self):
 
848
        """Initialize the state in this tree to be a new tree."""
 
849
        self._dirty = True
 
850
 
 
851
    @needs_read_lock
 
852
    def path2id(self, path):
 
853
        """Return the id for path in this tree."""
 
854
        path = path.strip('/')
 
855
        entry = self._get_entry(path=path)
 
856
        if entry == (None, None):
 
857
            return None
 
858
        return entry[0][2]
 
859
 
 
860
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
861
        """See Tree.paths2ids().
 
862
 
 
863
        This specialisation fast-paths the case where all the trees are in the
 
864
        dirstate.
 
865
        """
 
866
        if paths is None:
 
867
            return None
 
868
        parents = self.get_parent_ids()
 
869
        for tree in trees:
 
870
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
871
                parents):
 
872
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
873
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
874
        # -- make all paths utf8 --
 
875
        paths_utf8 = set()
 
876
        for path in paths:
 
877
            paths_utf8.add(path.encode('utf8'))
 
878
        paths = paths_utf8
 
879
        # -- paths is now a utf8 path set --
 
880
        # -- get the state object and prepare it.
 
881
        state = self.current_dirstate()
 
882
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
883
            and '' not in paths):
 
884
            paths2ids = self._paths2ids_using_bisect
 
885
        else:
 
886
            paths2ids = self._paths2ids_in_memory
 
887
        return paths2ids(paths, search_indexes,
 
888
                         require_versioned=require_versioned)
 
889
 
 
890
    def _paths2ids_in_memory(self, paths, search_indexes,
 
891
                             require_versioned=True):
 
892
        state = self.current_dirstate()
 
893
        state._read_dirblocks_if_needed()
 
894
        def _entries_for_path(path):
 
895
            """Return a list with all the entries that match path for all ids.
 
896
            """
 
897
            dirname, basename = os.path.split(path)
 
898
            key = (dirname, basename, '')
 
899
            block_index, present = state._find_block_index_from_key(key)
 
900
            if not present:
 
901
                # the block which should contain path is absent.
 
902
                return []
 
903
            result = []
 
904
            block = state._dirblocks[block_index][1]
 
905
            entry_index, _ = state._find_entry_index(key, block)
 
906
            # we may need to look at multiple entries at this path: walk while the paths match.
 
907
            while (entry_index < len(block) and
 
908
                block[entry_index][0][0:2] == key[0:2]):
 
909
                result.append(block[entry_index])
 
910
                entry_index += 1
 
911
            return result
 
912
        if require_versioned:
 
913
            # -- check all supplied paths are versioned in a search tree. --
 
914
            all_versioned = True
 
915
            for path in paths:
 
916
                path_entries = _entries_for_path(path)
 
917
                if not path_entries:
 
918
                    # this specified path is not present at all: error
 
919
                    all_versioned = False
 
920
                    break
 
921
                found_versioned = False
 
922
                # for each id at this path
 
923
                for entry in path_entries:
 
924
                    # for each tree.
 
925
                    for index in search_indexes:
 
926
                        if entry[1][index][0] != 'a': # absent
 
927
                            found_versioned = True
 
928
                            # all good: found a versioned cell
 
929
                            break
 
930
                if not found_versioned:
 
931
                    # none of the indexes was not 'absent' at all ids for this
 
932
                    # path.
 
933
                    all_versioned = False
 
934
                    break
 
935
            if not all_versioned:
 
936
                raise errors.PathsNotVersionedError(paths)
 
937
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
938
        search_paths = osutils.minimum_path_selection(paths)
 
939
        # sketch: 
 
940
        # for all search_indexs in each path at or under each element of
 
941
        # search_paths, if the detail is relocated: add the id, and add the
 
942
        # relocated path as one to search if its not searched already. If the
 
943
        # detail is not relocated, add the id.
 
944
        searched_paths = set()
 
945
        found_ids = set()
 
946
        def _process_entry(entry):
 
947
            """Look at search_indexes within entry.
 
948
 
 
949
            If a specific tree's details are relocated, add the relocation
 
950
            target to search_paths if not searched already. If it is absent, do
 
951
            nothing. Otherwise add the id to found_ids.
 
952
            """
 
953
            for index in search_indexes:
 
954
                if entry[1][index][0] == 'r': # relocated
 
955
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
956
                        search_paths.add(entry[1][index][1])
 
957
                elif entry[1][index][0] != 'a': # absent
 
958
                    found_ids.add(entry[0][2])
 
959
        while search_paths:
 
960
            current_root = search_paths.pop()
 
961
            searched_paths.add(current_root)
 
962
            # process the entries for this containing directory: the rest will be
 
963
            # found by their parents recursively.
 
964
            root_entries = _entries_for_path(current_root)
 
965
            if not root_entries:
 
966
                # this specified path is not present at all, skip it.
 
967
                continue
 
968
            for entry in root_entries:
 
969
                _process_entry(entry)
 
970
            initial_key = (current_root, '', '')
 
971
            block_index, _ = state._find_block_index_from_key(initial_key)
 
972
            while (block_index < len(state._dirblocks) and
 
973
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
974
                for entry in state._dirblocks[block_index][1]:
 
975
                    _process_entry(entry)
 
976
                block_index += 1
 
977
        return found_ids
 
978
 
 
979
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
980
                                require_versioned=True):
 
981
        state = self.current_dirstate()
 
982
        found_ids = set()
 
983
 
 
984
        split_paths = sorted(osutils.split(p) for p in paths)
 
985
        found = state._bisect_recursive(split_paths)
 
986
 
 
987
        if require_versioned:
 
988
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
989
            for dir_name in split_paths:
 
990
                if dir_name not in found_dir_names:
 
991
                    raise errors.PathsNotVersionedError(paths)
 
992
 
 
993
        for dir_name_id, trees_info in found.iteritems():
 
994
            for index in search_indexes:
 
995
                if trees_info[index][0] not in ('r', 'a'):
 
996
                    found_ids.add(dir_name_id[2])
 
997
        return found_ids
 
998
 
 
999
    def read_working_inventory(self):
 
1000
        """Read the working inventory.
 
1001
        
 
1002
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1003
        """
 
1004
        return self.inventory
 
1005
 
 
1006
    @needs_read_lock
 
1007
    def revision_tree(self, revision_id):
 
1008
        """See Tree.revision_tree.
 
1009
 
 
1010
        WorkingTree4 supplies revision_trees for any basis tree.
 
1011
        """
 
1012
        dirstate = self.current_dirstate()
 
1013
        parent_ids = dirstate.get_parent_ids()
 
1014
        if revision_id not in parent_ids:
 
1015
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1016
        if revision_id in dirstate.get_ghosts():
 
1017
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1018
        return DirStateRevisionTree(dirstate, revision_id,
 
1019
            self.branch.repository)
 
1020
 
 
1021
    @needs_tree_write_lock
 
1022
    def set_last_revision(self, new_revision):
 
1023
        """Change the last revision in the working tree."""
 
1024
        parents = self.get_parent_ids()
 
1025
        if new_revision in (NULL_REVISION, None):
 
1026
            assert len(parents) < 2, (
 
1027
                "setting the last parent to none with a pending merge is "
 
1028
                "unsupported.")
 
1029
            self.set_parent_ids([])
 
1030
        else:
 
1031
            self.set_parent_ids([new_revision] + parents[1:],
 
1032
                allow_leftmost_as_ghost=True)
 
1033
 
 
1034
    @needs_tree_write_lock
 
1035
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1036
        """Set the parent ids to revision_ids.
 
1037
        
 
1038
        See also set_parent_trees. This api will try to retrieve the tree data
 
1039
        for each element of revision_ids from the trees repository. If you have
 
1040
        tree data already available, it is more efficient to use
 
1041
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1042
        an easier API to use.
 
1043
 
 
1044
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1045
            working tree. Any of these may be ghosts.
 
1046
        """
 
1047
        trees = []
 
1048
        for revision_id in revision_ids:
 
1049
            try:
 
1050
                revtree = self.branch.repository.revision_tree(revision_id)
 
1051
                # TODO: jam 20070213 KnitVersionedFile raises
 
1052
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1053
                #       given revision_id is not present. Should Repository be
 
1054
                #       catching it and re-raising NoSuchRevision?
 
1055
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1056
                revtree = None
 
1057
            trees.append((revision_id, revtree))
 
1058
        self.set_parent_trees(trees,
 
1059
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1060
 
 
1061
    @needs_tree_write_lock
 
1062
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1063
        """Set the parents of the working tree.
 
1064
 
 
1065
        :param parents_list: A list of (revision_id, tree) tuples.
 
1066
            If tree is None, then that element is treated as an unreachable
 
1067
            parent tree - i.e. a ghost.
 
1068
        """
 
1069
        dirstate = self.current_dirstate()
 
1070
        if len(parents_list) > 0:
 
1071
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1072
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1073
        real_trees = []
 
1074
        ghosts = []
 
1075
        # convert absent trees to the null tree, which we convert back to
 
1076
        # missing on access.
 
1077
        for rev_id, tree in parents_list:
 
1078
            _mod_revision.check_not_reserved_id(rev_id)
 
1079
            if tree is not None:
 
1080
                real_trees.append((rev_id, tree))
 
1081
            else:
 
1082
                real_trees.append((rev_id,
 
1083
                    self.branch.repository.revision_tree(None)))
 
1084
                ghosts.append(rev_id)
 
1085
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1086
        self._make_dirty(reset_inventory=False)
 
1087
 
 
1088
    def _set_root_id(self, file_id):
 
1089
        """See WorkingTree.set_root_id."""
 
1090
        state = self.current_dirstate()
 
1091
        state.set_path_id('', file_id)
 
1092
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1093
            self._make_dirty(reset_inventory=True)
 
1094
 
 
1095
    def _sha_from_stat(self, path, stat_result):
 
1096
        """Get a sha digest from the tree's stat cache.
 
1097
 
 
1098
        The default implementation assumes no stat cache is present.
 
1099
 
 
1100
        :param path: The path.
 
1101
        :param stat_result: The stat result being looked up.
 
1102
        """
 
1103
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1104
 
 
1105
    @needs_read_lock
 
1106
    def supports_tree_reference(self):
 
1107
        return self._repo_supports_tree_reference
 
1108
 
 
1109
    def unlock(self):
 
1110
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1111
        # do non-implementation specific cleanup
 
1112
        self._cleanup()
 
1113
 
 
1114
        if self._control_files._lock_count == 1:
 
1115
            # eventually we should do signature checking during read locks for
 
1116
            # dirstate updates.
 
1117
            if self._control_files._lock_mode == 'w':
 
1118
                if self._dirty:
 
1119
                    self.flush()
 
1120
            if self._dirstate is not None:
 
1121
                # This is a no-op if there are no modifications.
 
1122
                self._dirstate.save()
 
1123
                self._dirstate.unlock()
 
1124
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1125
            #       point. Instead, it could check if the header has been
 
1126
            #       modified when it is locked, and if not, it can hang on to
 
1127
            #       the data it has in memory.
 
1128
            self._dirstate = None
 
1129
            self._inventory = None
 
1130
        # reverse order of locking.
 
1131
        try:
 
1132
            return self._control_files.unlock()
 
1133
        finally:
 
1134
            self.branch.unlock()
 
1135
 
 
1136
    @needs_tree_write_lock
 
1137
    def unversion(self, file_ids):
 
1138
        """Remove the file ids in file_ids from the current versioned set.
 
1139
 
 
1140
        When a file_id is unversioned, all of its children are automatically
 
1141
        unversioned.
 
1142
 
 
1143
        :param file_ids: The file ids to stop versioning.
 
1144
        :raises: NoSuchId if any fileid is not currently versioned.
 
1145
        """
 
1146
        if not file_ids:
 
1147
            return
 
1148
        state = self.current_dirstate()
 
1149
        state._read_dirblocks_if_needed()
 
1150
        ids_to_unversion = set(file_ids)
 
1151
        paths_to_unversion = set()
 
1152
        # sketch:
 
1153
        # check if the root is to be unversioned, if so, assert for now.
 
1154
        # walk the state marking unversioned things as absent.
 
1155
        # if there are any un-unversioned ids at the end, raise
 
1156
        for key, details in state._dirblocks[0][1]:
 
1157
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1158
                key[2] in ids_to_unversion):
 
1159
                # I haven't written the code to unversion / yet - it should be
 
1160
                # supported.
 
1161
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1162
        block_index = 0
 
1163
        while block_index < len(state._dirblocks):
 
1164
            # process one directory at a time.
 
1165
            block = state._dirblocks[block_index]
 
1166
            # first check: is the path one to remove - it or its children
 
1167
            delete_block = False
 
1168
            for path in paths_to_unversion:
 
1169
                if (block[0].startswith(path) and
 
1170
                    (len(block[0]) == len(path) or
 
1171
                     block[0][len(path)] == '/')):
 
1172
                    # this entire block should be deleted - its the block for a
 
1173
                    # path to unversion; or the child of one
 
1174
                    delete_block = True
 
1175
                    break
 
1176
            # TODO: trim paths_to_unversion as we pass by paths
 
1177
            if delete_block:
 
1178
                # this block is to be deleted: process it.
 
1179
                # TODO: we can special case the no-parents case and
 
1180
                # just forget the whole block.
 
1181
                entry_index = 0
 
1182
                while entry_index < len(block[1]):
 
1183
                    # Mark this file id as having been removed
 
1184
                    entry = block[1][entry_index]
 
1185
                    ids_to_unversion.discard(entry[0][2])
 
1186
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1187
                                               # entries
 
1188
                        or not state._make_absent(entry)):
 
1189
                        entry_index += 1
 
1190
                # go to the next block. (At the moment we dont delete empty
 
1191
                # dirblocks)
 
1192
                block_index += 1
 
1193
                continue
 
1194
            entry_index = 0
 
1195
            while entry_index < len(block[1]):
 
1196
                entry = block[1][entry_index]
 
1197
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1198
                    # ^ some parent row.
 
1199
                    entry[0][2] not in ids_to_unversion):
 
1200
                    # ^ not an id to unversion
 
1201
                    entry_index += 1
 
1202
                    continue
 
1203
                if entry[1][0][0] == 'd':
 
1204
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1205
                if not state._make_absent(entry):
 
1206
                    entry_index += 1
 
1207
                # we have unversioned this id
 
1208
                ids_to_unversion.remove(entry[0][2])
 
1209
            block_index += 1
 
1210
        if ids_to_unversion:
 
1211
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1212
        self._make_dirty(reset_inventory=False)
 
1213
        # have to change the legacy inventory too.
 
1214
        if self._inventory is not None:
 
1215
            for file_id in file_ids:
 
1216
                self._inventory.remove_recursive_id(file_id)
 
1217
 
 
1218
    def update_basis_by_delta(self, new_revid, delta):
 
1219
        """See MutableTree.update_basis_by_delta."""
 
1220
        assert self.last_revision() != new_revid
 
1221
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1222
 
 
1223
    @needs_read_lock
 
1224
    def _validate(self):
 
1225
        self._dirstate._validate()
 
1226
 
 
1227
    @needs_tree_write_lock
 
1228
    def _write_inventory(self, inv):
 
1229
        """Write inventory as the current inventory."""
 
1230
        assert not self._dirty, ("attempting to write an inventory when the "
 
1231
            "dirstate is dirty will cause data loss")
 
1232
        self.current_dirstate().set_state_from_inventory(inv)
 
1233
        self._make_dirty(reset_inventory=False)
 
1234
        if self._inventory is not None:
 
1235
            self._inventory = inv
 
1236
        self.flush()
 
1237
 
 
1238
 
 
1239
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1240
    """The first consolidated dirstate working tree format.
 
1241
 
 
1242
    This format:
 
1243
        - exists within a metadir controlling .bzr
 
1244
        - includes an explicit version marker for the workingtree control
 
1245
          files, separate from the BzrDir format
 
1246
        - modifies the hash cache format
 
1247
        - is new in bzr 0.15
 
1248
        - uses a LockDir to guard access to it.
 
1249
    """
 
1250
 
 
1251
    upgrade_recommended = False
 
1252
 
 
1253
    def get_format_string(self):
 
1254
        """See WorkingTreeFormat.get_format_string()."""
 
1255
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1256
 
 
1257
    def get_format_description(self):
 
1258
        """See WorkingTreeFormat.get_format_description()."""
 
1259
        return "Working tree format 4"
 
1260
 
 
1261
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1262
                   accelerator_tree=None, hardlink=False):
 
1263
        """See WorkingTreeFormat.initialize().
 
1264
 
 
1265
        :param revision_id: allows creating a working tree at a different
 
1266
        revision than the branch is at.
 
1267
        :param accelerator_tree: A tree which can be used for retrieving file
 
1268
            contents more quickly than the revision tree, i.e. a workingtree.
 
1269
            The revision tree will be used for cases where accelerator_tree's
 
1270
            content is different.
 
1271
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1272
            where possible.
 
1273
 
 
1274
        These trees get an initial random root id, if their repository supports
 
1275
        rich root data, TREE_ROOT otherwise.
 
1276
        """
 
1277
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1278
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1279
        transport = a_bzrdir.get_workingtree_transport(self)
 
1280
        control_files = self._open_control_files(a_bzrdir)
 
1281
        control_files.create_lock()
 
1282
        control_files.lock_write()
 
1283
        control_files.put_utf8('format', self.get_format_string())
 
1284
        if from_branch is not None:
 
1285
            branch = from_branch
 
1286
        else:
 
1287
            branch = a_bzrdir.open_branch()
 
1288
        if revision_id is None:
 
1289
            revision_id = branch.last_revision()
 
1290
        local_path = transport.local_abspath('dirstate')
 
1291
        # write out new dirstate (must exist when we create the tree)
 
1292
        state = dirstate.DirState.initialize(local_path)
 
1293
        state.unlock()
 
1294
        del state
 
1295
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1296
                         branch,
 
1297
                         _format=self,
 
1298
                         _bzrdir=a_bzrdir,
 
1299
                         _control_files=control_files)
 
1300
        wt._new_tree()
 
1301
        wt.lock_tree_write()
 
1302
        try:
 
1303
            if revision_id in (None, NULL_REVISION):
 
1304
                if branch.repository.supports_rich_root():
 
1305
                    wt._set_root_id(generate_ids.gen_root_id())
 
1306
                else:
 
1307
                    wt._set_root_id(ROOT_ID)
 
1308
                wt.flush()
 
1309
            basis = None
 
1310
            # frequently, we will get here due to branching.  The accelerator
 
1311
            # tree will be the tree from the branch, so the desired basis
 
1312
            # tree will often be a parent of the accelerator tree.
 
1313
            if accelerator_tree is not None:
 
1314
                try:
 
1315
                    basis = accelerator_tree.revision_tree(revision_id)
 
1316
                except errors.NoSuchRevision:
 
1317
                    pass
 
1318
            if basis is None:
 
1319
                basis = branch.repository.revision_tree(revision_id)
 
1320
            if revision_id == NULL_REVISION:
 
1321
                parents_list = []
 
1322
            else:
 
1323
                parents_list = [(revision_id, basis)]
 
1324
            basis.lock_read()
 
1325
            wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1326
            wt.flush()
 
1327
            # if the basis has a root id we have to use that; otherwise we use
 
1328
            # a new random one
 
1329
            basis_root_id = basis.get_root_id()
 
1330
            if basis_root_id is not None:
 
1331
                wt._set_root_id(basis_root_id)
 
1332
                wt.flush()
 
1333
            transform.build_tree(basis, wt, accelerator_tree,
 
1334
                                 hardlink=hardlink)
 
1335
            basis.unlock()
 
1336
        finally:
 
1337
            control_files.unlock()
 
1338
            wt.unlock()
 
1339
        return wt
 
1340
 
 
1341
    def _open(self, a_bzrdir, control_files):
 
1342
        """Open the tree itself.
 
1343
 
 
1344
        :param a_bzrdir: the dir for the tree.
 
1345
        :param control_files: the control files for the tree.
 
1346
        """
 
1347
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1348
                           branch=a_bzrdir.open_branch(),
 
1349
                           _format=self,
 
1350
                           _bzrdir=a_bzrdir,
 
1351
                           _control_files=control_files)
 
1352
 
 
1353
    def __get_matchingbzrdir(self):
 
1354
        # please test against something that will let us do tree references
 
1355
        return bzrdir.format_registry.make_bzrdir(
 
1356
            'dirstate-with-subtree')
 
1357
 
 
1358
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1359
 
 
1360
 
 
1361
class DirStateRevisionTree(Tree):
 
1362
    """A revision tree pulling the inventory from a dirstate."""
 
1363
 
 
1364
    def __init__(self, dirstate, revision_id, repository):
 
1365
        self._dirstate = dirstate
 
1366
        self._revision_id = revision_id
 
1367
        self._repository = repository
 
1368
        self._inventory = None
 
1369
        self._locked = 0
 
1370
        self._dirstate_locked = False
 
1371
 
 
1372
    def __repr__(self):
 
1373
        return "<%s of %s in %s>" % \
 
1374
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1375
 
 
1376
    def annotate_iter(self, file_id,
 
1377
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1378
        """See Tree.annotate_iter"""
 
1379
        w = self._get_weave(file_id)
 
1380
        return w.annotate_iter(self.inventory[file_id].revision)
 
1381
 
 
1382
    def _get_ancestors(self, default_revision):
 
1383
        return set(self._repository.get_ancestry(self._revision_id,
 
1384
                                                 topo_sorted=False))
 
1385
    def _comparison_data(self, entry, path):
 
1386
        """See Tree._comparison_data."""
 
1387
        if entry is None:
 
1388
            return None, False, None
 
1389
        # trust the entry as RevisionTree does, but this may not be
 
1390
        # sensible: the entry might not have come from us?
 
1391
        return entry.kind, entry.executable, None
 
1392
 
 
1393
    def _file_size(self, entry, stat_value):
 
1394
        return entry.text_size
 
1395
 
 
1396
    def filter_unversioned_files(self, paths):
 
1397
        """Filter out paths that are not versioned.
 
1398
 
 
1399
        :return: set of paths.
 
1400
        """
 
1401
        pred = self.has_filename
 
1402
        return set((p for p in paths if not pred(p)))
 
1403
 
 
1404
    def get_root_id(self):
 
1405
        return self.path2id('')
 
1406
 
 
1407
    def id2path(self, file_id):
 
1408
        "Convert a file-id to a path."
 
1409
        entry = self._get_entry(file_id=file_id)
 
1410
        if entry == (None, None):
 
1411
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1412
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1413
        return path_utf8.decode('utf8')
 
1414
 
 
1415
    def _get_parent_index(self):
 
1416
        """Return the index in the dirstate referenced by this tree."""
 
1417
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1418
 
 
1419
    def _get_entry(self, file_id=None, path=None):
 
1420
        """Get the dirstate row for file_id or path.
 
1421
 
 
1422
        If either file_id or path is supplied, it is used as the key to lookup.
 
1423
        If both are supplied, the fastest lookup is used, and an error is
 
1424
        raised if they do not both point at the same row.
 
1425
        
 
1426
        :param file_id: An optional unicode file_id to be looked up.
 
1427
        :param path: An optional unicode path to be looked up.
 
1428
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1429
        """
 
1430
        if file_id is None and path is None:
 
1431
            raise errors.BzrError('must supply file_id or path')
 
1432
        if path is not None:
 
1433
            path = path.encode('utf8')
 
1434
        parent_index = self._get_parent_index()
 
1435
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1436
 
 
1437
    def _generate_inventory(self):
 
1438
        """Create and set self.inventory from the dirstate object.
 
1439
 
 
1440
        (So this is only called the first time the inventory is requested for
 
1441
        this tree; it then remains in memory until it's out of date.)
 
1442
 
 
1443
        This is relatively expensive: we have to walk the entire dirstate.
 
1444
        """
 
1445
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1446
            'dirstate revision tree'
 
1447
        # separate call for profiling - makes it clear where the costs are.
 
1448
        self._dirstate._read_dirblocks_if_needed()
 
1449
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1450
            'parent %s has disappeared from %s' % (
 
1451
            self._revision_id, self._dirstate.get_parent_ids())
 
1452
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1453
        # This is identical now to the WorkingTree _generate_inventory except
 
1454
        # for the tree index use.
 
1455
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1456
        current_id = root_key[2]
 
1457
        assert current_entry[parent_index][0] == 'd'
 
1458
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1459
        inv.root.revision = current_entry[parent_index][4]
 
1460
        # Turn some things into local variables
 
1461
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1462
        factory = entry_factory
 
1463
        utf8_decode = cache_utf8._utf8_decode
 
1464
        inv_byid = inv._byid
 
1465
        # we could do this straight out of the dirstate; it might be fast
 
1466
        # and should be profiled - RBC 20070216
 
1467
        parent_ies = {'' : inv.root}
 
1468
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1469
            dirname = block[0]
 
1470
            try:
 
1471
                parent_ie = parent_ies[dirname]
 
1472
            except KeyError:
 
1473
                # all the paths in this block are not versioned in this tree
 
1474
                continue
 
1475
            for key, entry in block[1]:
 
1476
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1477
                if minikind in ('a', 'r'): # absent, relocated
 
1478
                    # not this tree
 
1479
                    continue
 
1480
                name = key[1]
 
1481
                name_unicode = utf8_decode(name)[0]
 
1482
                file_id = key[2]
 
1483
                kind = minikind_to_kind[minikind]
 
1484
                inv_entry = factory[kind](file_id, name_unicode,
 
1485
                                          parent_ie.file_id)
 
1486
                inv_entry.revision = revid
 
1487
                if kind == 'file':
 
1488
                    inv_entry.executable = executable
 
1489
                    inv_entry.text_size = size
 
1490
                    inv_entry.text_sha1 = fingerprint
 
1491
                elif kind == 'directory':
 
1492
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1493
                elif kind == 'symlink':
 
1494
                    inv_entry.executable = False
 
1495
                    inv_entry.text_size = None
 
1496
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1497
                elif kind == 'tree-reference':
 
1498
                    inv_entry.reference_revision = fingerprint or None
 
1499
                else:
 
1500
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1501
                            % entry)
 
1502
                # These checks cost us around 40ms on a 55k entry tree
 
1503
                assert file_id not in inv_byid
 
1504
                assert name_unicode not in parent_ie.children
 
1505
                inv_byid[file_id] = inv_entry
 
1506
                parent_ie.children[name_unicode] = inv_entry
 
1507
        self._inventory = inv
 
1508
 
 
1509
    def get_file_mtime(self, file_id, path=None):
 
1510
        """Return the modification time for this record.
 
1511
 
 
1512
        We return the timestamp of the last-changed revision.
 
1513
        """
 
1514
        # Make sure the file exists
 
1515
        entry = self._get_entry(file_id, path=path)
 
1516
        if entry == (None, None): # do we raise?
 
1517
            return None
 
1518
        parent_index = self._get_parent_index()
 
1519
        last_changed_revision = entry[1][parent_index][4]
 
1520
        return self._repository.get_revision(last_changed_revision).timestamp
 
1521
 
 
1522
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1523
        entry = self._get_entry(file_id=file_id, path=path)
 
1524
        parent_index = self._get_parent_index()
 
1525
        parent_details = entry[1][parent_index]
 
1526
        if parent_details[0] == 'f':
 
1527
            return parent_details[1]
 
1528
        return None
 
1529
 
 
1530
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1531
    def get_weave(self, file_id):
 
1532
        return self._get_weave(file_id)
 
1533
 
 
1534
    def _get_weave(self, file_id):
 
1535
        return self._repository.weave_store.get_weave(file_id,
 
1536
                self._repository.get_transaction())
 
1537
 
 
1538
    def get_file(self, file_id, path=None):
 
1539
        return StringIO(self.get_file_text(file_id))
 
1540
 
 
1541
    def get_file_lines(self, file_id):
 
1542
        entry = self._get_entry(file_id=file_id)[1]
 
1543
        if entry == None:
 
1544
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1545
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1546
 
 
1547
    def get_file_size(self, file_id):
 
1548
        return self.inventory[file_id].text_size
 
1549
 
 
1550
    def get_file_text(self, file_id):
 
1551
        return ''.join(self.get_file_lines(file_id))
 
1552
 
 
1553
    def get_reference_revision(self, file_id, path=None):
 
1554
        return self.inventory[file_id].reference_revision
 
1555
 
 
1556
    def iter_files_bytes(self, desired_files):
 
1557
        """See Tree.iter_files_bytes.
 
1558
 
 
1559
        This version is implemented on top of Repository.iter_files_bytes"""
 
1560
        parent_index = self._get_parent_index()
 
1561
        repo_desired_files = []
 
1562
        for file_id, identifier in desired_files:
 
1563
            entry = self._get_entry(file_id)
 
1564
            if entry == (None, None):
 
1565
                raise errors.NoSuchId(self, file_id)
 
1566
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1567
                                       identifier))
 
1568
        return self._repository.iter_files_bytes(repo_desired_files)
 
1569
 
 
1570
    def get_symlink_target(self, file_id):
 
1571
        entry = self._get_entry(file_id=file_id)
 
1572
        parent_index = self._get_parent_index()
 
1573
        if entry[1][parent_index][0] != 'l':
 
1574
            return None
 
1575
        else:
 
1576
            # At present, none of the tree implementations supports non-ascii
 
1577
            # symlink targets. So we will just assume that the dirstate path is
 
1578
            # correct.
 
1579
            return entry[1][parent_index][1]
 
1580
 
 
1581
    def get_revision_id(self):
 
1582
        """Return the revision id for this tree."""
 
1583
        return self._revision_id
 
1584
 
 
1585
    def _get_inventory(self):
 
1586
        if self._inventory is not None:
 
1587
            return self._inventory
 
1588
        self._must_be_locked()
 
1589
        self._generate_inventory()
 
1590
        return self._inventory
 
1591
 
 
1592
    inventory = property(_get_inventory,
 
1593
                         doc="Inventory of this Tree")
 
1594
 
 
1595
    def get_parent_ids(self):
 
1596
        """The parents of a tree in the dirstate are not cached."""
 
1597
        return self._repository.get_revision(self._revision_id).parent_ids
 
1598
 
 
1599
    def has_filename(self, filename):
 
1600
        return bool(self.path2id(filename))
 
1601
 
 
1602
    def kind(self, file_id):
 
1603
        entry = self._get_entry(file_id=file_id)[1]
 
1604
        if entry == None:
 
1605
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1606
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1607
 
 
1608
    def path_content_summary(self, path):
 
1609
        """See Tree.path_content_summary."""
 
1610
        id = self.inventory.path2id(path)
 
1611
        if id is None:
 
1612
            return ('missing', None, None, None)
 
1613
        entry = self._inventory[id]
 
1614
        kind = entry.kind
 
1615
        if kind == 'file':
 
1616
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1617
        elif kind == 'symlink':
 
1618
            return (kind, None, None, entry.symlink_target)
 
1619
        else:
 
1620
            return (kind, None, None, None)
 
1621
 
 
1622
    def is_executable(self, file_id, path=None):
 
1623
        ie = self.inventory[file_id]
 
1624
        if ie.kind != "file":
 
1625
            return None
 
1626
        return ie.executable
 
1627
 
 
1628
    def list_files(self, include_root=False):
 
1629
        # We use a standard implementation, because DirStateRevisionTree is
 
1630
        # dealing with one of the parents of the current state
 
1631
        inv = self._get_inventory()
 
1632
        entries = inv.iter_entries()
 
1633
        if self.inventory.root is not None and not include_root:
 
1634
            entries.next()
 
1635
        for path, entry in entries:
 
1636
            yield path, 'V', entry.kind, entry.file_id, entry
 
1637
 
 
1638
    def lock_read(self):
 
1639
        """Lock the tree for a set of operations."""
 
1640
        if not self._locked:
 
1641
            self._repository.lock_read()
 
1642
            if self._dirstate._lock_token is None:
 
1643
                self._dirstate.lock_read()
 
1644
                self._dirstate_locked = True
 
1645
        self._locked += 1
 
1646
 
 
1647
    def _must_be_locked(self):
 
1648
        if not self._locked:
 
1649
            raise errors.ObjectNotLocked(self)
 
1650
 
 
1651
    @needs_read_lock
 
1652
    def path2id(self, path):
 
1653
        """Return the id for path in this tree."""
 
1654
        # lookup by path: faster than splitting and walking the ivnentory.
 
1655
        entry = self._get_entry(path=path)
 
1656
        if entry == (None, None):
 
1657
            return None
 
1658
        return entry[0][2]
 
1659
 
 
1660
    def unlock(self):
 
1661
        """Unlock, freeing any cache memory used during the lock."""
 
1662
        # outside of a lock, the inventory is suspect: release it.
 
1663
        self._locked -=1
 
1664
        if not self._locked:
 
1665
            self._inventory = None
 
1666
            self._locked = 0
 
1667
            if self._dirstate_locked:
 
1668
                self._dirstate.unlock()
 
1669
                self._dirstate_locked = False
 
1670
            self._repository.unlock()
 
1671
 
 
1672
    def walkdirs(self, prefix=""):
 
1673
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1674
        # implementation based on an inventory.  
 
1675
        # This should be cleaned up to use the much faster Dirstate code
 
1676
        # So for now, we just build up the parent inventory, and extract
 
1677
        # it the same way RevisionTree does.
 
1678
        _directory = 'directory'
 
1679
        inv = self._get_inventory()
 
1680
        top_id = inv.path2id(prefix)
 
1681
        if top_id is None:
 
1682
            pending = []
 
1683
        else:
 
1684
            pending = [(prefix, top_id)]
 
1685
        while pending:
 
1686
            dirblock = []
 
1687
            relpath, file_id = pending.pop()
 
1688
            # 0 - relpath, 1- file-id
 
1689
            if relpath:
 
1690
                relroot = relpath + '/'
 
1691
            else:
 
1692
                relroot = ""
 
1693
            # FIXME: stash the node in pending
 
1694
            entry = inv[file_id]
 
1695
            for name, child in entry.sorted_children():
 
1696
                toppath = relroot + name
 
1697
                dirblock.append((toppath, name, child.kind, None,
 
1698
                    child.file_id, child.kind
 
1699
                    ))
 
1700
            yield (relpath, entry.file_id), dirblock
 
1701
            # push the user specified dirs from dirblock
 
1702
            for dir in reversed(dirblock):
 
1703
                if dir[2] == _directory:
 
1704
                    pending.append((dir[0], dir[4]))
 
1705
 
 
1706
 
 
1707
class InterDirStateTree(InterTree):
 
1708
    """Fast path optimiser for changes_from with dirstate trees.
 
1709
    
 
1710
    This is used only when both trees are in the dirstate working file, and 
 
1711
    the source is any parent within the dirstate, and the destination is 
 
1712
    the current working tree of the same dirstate.
 
1713
    """
 
1714
    # this could be generalized to allow comparisons between any trees in the
 
1715
    # dirstate, and possibly between trees stored in different dirstates.
 
1716
 
 
1717
    def __init__(self, source, target):
 
1718
        super(InterDirStateTree, self).__init__(source, target)
 
1719
        if not InterDirStateTree.is_compatible(source, target):
 
1720
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1721
 
 
1722
    @staticmethod
 
1723
    def make_source_parent_tree(source, target):
 
1724
        """Change the source tree into a parent of the target."""
 
1725
        revid = source.commit('record tree')
 
1726
        target.branch.repository.fetch(source.branch.repository, revid)
 
1727
        target.set_parent_ids([revid])
 
1728
        return target.basis_tree(), target
 
1729
 
 
1730
    _matching_from_tree_format = WorkingTreeFormat4()
 
1731
    _matching_to_tree_format = WorkingTreeFormat4()
 
1732
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1733
 
 
1734
    def _iter_changes(self, include_unchanged=False,
 
1735
                      specific_files=None, pb=None, extra_trees=[],
 
1736
                      require_versioned=True, want_unversioned=False):
 
1737
        """Return the changes from source to target.
 
1738
 
 
1739
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1740
            for details.
 
1741
        :param specific_files: An optional list of file paths to restrict the
 
1742
            comparison to. When mapping filenames to ids, all matches in all
 
1743
            trees (including optional extra_trees) are used, and all children of
 
1744
            matched directories are included.
 
1745
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1746
            unchanged entries in the result.
 
1747
        :param extra_trees: An optional list of additional trees to use when
 
1748
            mapping the contents of specific_files (paths) to file_ids.
 
1749
        :param require_versioned: If True, all files in specific_files must be
 
1750
            versioned in one of source, target, extra_trees or
 
1751
            PathsNotVersionedError is raised.
 
1752
        :param want_unversioned: Should unversioned files be returned in the
 
1753
            output. An unversioned file is defined as one with (False, False)
 
1754
            for the versioned pair.
 
1755
        """
 
1756
        utf8_decode = cache_utf8._utf8_decode
 
1757
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1758
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1759
        # NB: show_status depends on being able to pass in non-versioned files
 
1760
        # and report them as unknown
 
1761
        # TODO: handle extra trees in the dirstate.
 
1762
        if (extra_trees or specific_files == []):
 
1763
            # we can't fast-path these cases (yet)
 
1764
            for f in super(InterDirStateTree, self)._iter_changes(
 
1765
                include_unchanged, specific_files, pb, extra_trees,
 
1766
                require_versioned, want_unversioned=want_unversioned):
 
1767
                yield f
 
1768
            return
 
1769
        parent_ids = self.target.get_parent_ids()
 
1770
        assert (self.source._revision_id in parent_ids
 
1771
                or self.source._revision_id == NULL_REVISION), \
 
1772
                "revision {%s} is not stored in {%s}, but %s " \
 
1773
                "can only be used for trees stored in the dirstate" \
 
1774
                % (self.source._revision_id, self.target, self._iter_changes)
 
1775
        target_index = 0
 
1776
        if self.source._revision_id == NULL_REVISION:
 
1777
            source_index = None
 
1778
            indices = (target_index,)
 
1779
        else:
 
1780
            assert (self.source._revision_id in parent_ids), \
 
1781
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1782
                self.source._revision_id, parent_ids)
 
1783
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1784
            indices = (source_index, target_index)
 
1785
        # -- make all specific_files utf8 --
 
1786
        if specific_files:
 
1787
            specific_files_utf8 = set()
 
1788
            for path in specific_files:
 
1789
                specific_files_utf8.add(path.encode('utf8'))
 
1790
            specific_files = specific_files_utf8
 
1791
        else:
 
1792
            specific_files = set([''])
 
1793
        # -- specific_files is now a utf8 path set --
 
1794
        # -- get the state object and prepare it.
 
1795
        state = self.target.current_dirstate()
 
1796
        state._read_dirblocks_if_needed()
 
1797
        def _entries_for_path(path):
 
1798
            """Return a list with all the entries that match path for all ids.
 
1799
            """
 
1800
            dirname, basename = os.path.split(path)
 
1801
            key = (dirname, basename, '')
 
1802
            block_index, present = state._find_block_index_from_key(key)
 
1803
            if not present:
 
1804
                # the block which should contain path is absent.
 
1805
                return []
 
1806
            result = []
 
1807
            block = state._dirblocks[block_index][1]
 
1808
            entry_index, _ = state._find_entry_index(key, block)
 
1809
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1810
            while (entry_index < len(block) and
 
1811
                block[entry_index][0][0:2] == key[0:2]):
 
1812
                result.append(block[entry_index])
 
1813
                entry_index += 1
 
1814
            return result
 
1815
        if require_versioned:
 
1816
            # -- check all supplied paths are versioned in a search tree. --
 
1817
            all_versioned = True
 
1818
            for path in specific_files:
 
1819
                path_entries = _entries_for_path(path)
 
1820
                if not path_entries:
 
1821
                    # this specified path is not present at all: error
 
1822
                    all_versioned = False
 
1823
                    break
 
1824
                found_versioned = False
 
1825
                # for each id at this path
 
1826
                for entry in path_entries:
 
1827
                    # for each tree.
 
1828
                    for index in indices:
 
1829
                        if entry[1][index][0] != 'a': # absent
 
1830
                            found_versioned = True
 
1831
                            # all good: found a versioned cell
 
1832
                            break
 
1833
                if not found_versioned:
 
1834
                    # none of the indexes was not 'absent' at all ids for this
 
1835
                    # path.
 
1836
                    all_versioned = False
 
1837
                    break
 
1838
            if not all_versioned:
 
1839
                raise errors.PathsNotVersionedError(specific_files)
 
1840
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1841
        search_specific_files = set()
 
1842
        for path in specific_files:
 
1843
            other_specific_files = specific_files.difference(set([path]))
 
1844
            if not osutils.is_inside_any(other_specific_files, path):
 
1845
                # this is a top level path, we must check it.
 
1846
                search_specific_files.add(path)
 
1847
        # sketch: 
 
1848
        # compare source_index and target_index at or under each element of search_specific_files.
 
1849
        # follow the following comparison table. Note that we only want to do diff operations when
 
1850
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1851
        # for the target.
 
1852
        # cases:
 
1853
        # 
 
1854
        # Source | Target | disk | action
 
1855
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1856
        #        |        |      | diff check on source-target
 
1857
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1858
        #        |        |      | ???
 
1859
        #   r    |  a     |      | add source to search
 
1860
        #   r    |  a     |  a   | 
 
1861
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1862
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1863
        #   a    | fdlt   |      | add new id
 
1864
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1865
        #   a    |  a     |      | not present in either tree, skip
 
1866
        #   a    |  a     |  a   | not present in any tree, skip
 
1867
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1868
        #        |        |      | may not be selected by the users list of paths.
 
1869
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1870
        #        |        |      | may not be selected by the users list of paths.
 
1871
        #  fdlt  | fdlt   |      | content in both: diff them
 
1872
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1873
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1874
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1875
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1876
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1877
        #        |        |      | this id at the other path.
 
1878
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1879
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1880
        #        |        |      | this id at the other path.
 
1881
 
 
1882
        # for all search_indexs in each path at or under each element of
 
1883
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1884
        # relocated path as one to search if its not searched already. If the
 
1885
        # detail is not relocated, add the id.
 
1886
        searched_specific_files = set()
 
1887
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1888
        # Using a list so that we can access the values and change them in
 
1889
        # nested scope. Each one is [path, file_id, entry]
 
1890
        last_source_parent = [None, None]
 
1891
        last_target_parent = [None, None]
 
1892
 
 
1893
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1894
 
 
1895
        # Just a sentry, so that _process_entry can say that this
 
1896
        # record is handled, but isn't interesting to process (unchanged)
 
1897
        uninteresting = object()
 
1898
 
 
1899
 
 
1900
        old_dirname_to_file_id = {}
 
1901
        new_dirname_to_file_id = {}
 
1902
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1903
        #       keeping a cache of directories that we have seen.
 
1904
 
 
1905
        def _process_entry(entry, path_info):
 
1906
            """Compare an entry and real disk to generate delta information.
 
1907
 
 
1908
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1909
                the path of entry. If None, then the path is considered absent.
 
1910
                (Perhaps we should pass in a concrete entry for this ?)
 
1911
                Basename is returned as a utf8 string because we expect this
 
1912
                tuple will be ignored, and don't want to take the time to
 
1913
                decode.
 
1914
            :return: None if these don't match
 
1915
                     A tuple of information about the change, or
 
1916
                     the object 'uninteresting' if these match, but are
 
1917
                     basically identical.
 
1918
            """
 
1919
            if source_index is None:
 
1920
                source_details = NULL_PARENT_DETAILS
 
1921
            else:
 
1922
                source_details = entry[1][source_index]
 
1923
            target_details = entry[1][target_index]
 
1924
            target_minikind = target_details[0]
 
1925
            if path_info is not None and target_minikind in 'fdlt':
 
1926
                assert target_index == 0
 
1927
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1928
                                                  stat_value=path_info[3])
 
1929
                # The entry may have been modified by update_entry
 
1930
                target_details = entry[1][target_index]
 
1931
                target_minikind = target_details[0]
 
1932
            else:
 
1933
                link_or_sha1 = None
 
1934
            file_id = entry[0][2]
 
1935
            source_minikind = source_details[0]
 
1936
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1937
                # claimed content in both: diff
 
1938
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1939
                #        |        |      | diff check on source-target
 
1940
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1941
                #        |        |      | ???
 
1942
                if source_minikind in 'r':
 
1943
                    # add the source to the search path to find any children it
 
1944
                    # has.  TODO ? : only add if it is a container ?
 
1945
                    if not osutils.is_inside_any(searched_specific_files,
 
1946
                                                 source_details[1]):
 
1947
                        search_specific_files.add(source_details[1])
 
1948
                    # generate the old path; this is needed for stating later
 
1949
                    # as well.
 
1950
                    old_path = source_details[1]
 
1951
                    old_dirname, old_basename = os.path.split(old_path)
 
1952
                    path = pathjoin(entry[0][0], entry[0][1])
 
1953
                    old_entry = state._get_entry(source_index,
 
1954
                                                 path_utf8=old_path)
 
1955
                    # update the source details variable to be the real
 
1956
                    # location.
 
1957
                    if old_entry == (None, None):
 
1958
                        raise errors.CorruptDirstate(state._filename,
 
1959
                            "entry '%s/%s' is considered renamed from %r"
 
1960
                            " but source does not exist\n"
 
1961
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
1962
                    source_details = old_entry[1][source_index]
 
1963
                    source_minikind = source_details[0]
 
1964
                else:
 
1965
                    old_dirname = entry[0][0]
 
1966
                    old_basename = entry[0][1]
 
1967
                    old_path = path = None
 
1968
                if path_info is None:
 
1969
                    # the file is missing on disk, show as removed.
 
1970
                    content_change = True
 
1971
                    target_kind = None
 
1972
                    target_exec = False
 
1973
                else:
 
1974
                    # source and target are both versioned and disk file is present.
 
1975
                    target_kind = path_info[2]
 
1976
                    if target_kind == 'directory':
 
1977
                        if path is None:
 
1978
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1979
                        new_dirname_to_file_id[path] = file_id
 
1980
                        if source_minikind != 'd':
 
1981
                            content_change = True
 
1982
                        else:
 
1983
                            # directories have no fingerprint
 
1984
                            content_change = False
 
1985
                        target_exec = False
 
1986
                    elif target_kind == 'file':
 
1987
                        if source_minikind != 'f':
 
1988
                            content_change = True
 
1989
                        else:
 
1990
                            # We could check the size, but we already have the
 
1991
                            # sha1 hash.
 
1992
                            content_change = (link_or_sha1 != source_details[1])
 
1993
                        # Target details is updated at update_entry time
 
1994
                        if use_filesystem_for_exec:
 
1995
                            # We don't need S_ISREG here, because we are sure
 
1996
                            # we are dealing with a file.
 
1997
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1998
                        else:
 
1999
                            target_exec = target_details[3]
 
2000
                    elif target_kind == 'symlink':
 
2001
                        if source_minikind != 'l':
 
2002
                            content_change = True
 
2003
                        else:
 
2004
                            content_change = (link_or_sha1 != source_details[1])
 
2005
                        target_exec = False
 
2006
                    elif target_kind == 'tree-reference':
 
2007
                        if source_minikind != 't':
 
2008
                            content_change = True
 
2009
                        else:
 
2010
                            content_change = False
 
2011
                        target_exec = False
 
2012
                    else:
 
2013
                        raise Exception, "unknown kind %s" % path_info[2]
 
2014
                if source_minikind == 'd':
 
2015
                    if path is None:
 
2016
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2017
                    old_dirname_to_file_id[old_path] = file_id
 
2018
                # parent id is the entry for the path in the target tree
 
2019
                if old_dirname == last_source_parent[0]:
 
2020
                    source_parent_id = last_source_parent[1]
 
2021
                else:
 
2022
                    try:
 
2023
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2024
                    except KeyError:
 
2025
                        source_parent_entry = state._get_entry(source_index,
 
2026
                                                               path_utf8=old_dirname)
 
2027
                        source_parent_id = source_parent_entry[0][2]
 
2028
                    if source_parent_id == entry[0][2]:
 
2029
                        # This is the root, so the parent is None
 
2030
                        source_parent_id = None
 
2031
                    else:
 
2032
                        last_source_parent[0] = old_dirname
 
2033
                        last_source_parent[1] = source_parent_id
 
2034
                new_dirname = entry[0][0]
 
2035
                if new_dirname == last_target_parent[0]:
 
2036
                    target_parent_id = last_target_parent[1]
 
2037
                else:
 
2038
                    try:
 
2039
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2040
                    except KeyError:
 
2041
                        # TODO: We don't always need to do the lookup, because the
 
2042
                        #       parent entry will be the same as the source entry.
 
2043
                        target_parent_entry = state._get_entry(target_index,
 
2044
                                                               path_utf8=new_dirname)
 
2045
                        assert target_parent_entry != (None, None), (
 
2046
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2047
                            % (new_dirname, entry))
 
2048
                        target_parent_id = target_parent_entry[0][2]
 
2049
                    if target_parent_id == entry[0][2]:
 
2050
                        # This is the root, so the parent is None
 
2051
                        target_parent_id = None
 
2052
                    else:
 
2053
                        last_target_parent[0] = new_dirname
 
2054
                        last_target_parent[1] = target_parent_id
 
2055
 
 
2056
                source_exec = source_details[3]
 
2057
                if (include_unchanged
 
2058
                    or content_change
 
2059
                    or source_parent_id != target_parent_id
 
2060
                    or old_basename != entry[0][1]
 
2061
                    or source_exec != target_exec
 
2062
                    ):
 
2063
                    if old_path is None:
 
2064
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2065
                        old_path_u = utf8_decode(old_path)[0]
 
2066
                        path_u = old_path_u
 
2067
                    else:
 
2068
                        old_path_u = utf8_decode(old_path)[0]
 
2069
                        if old_path == path:
 
2070
                            path_u = old_path_u
 
2071
                        else:
 
2072
                            path_u = utf8_decode(path)[0]
 
2073
                    source_kind = _minikind_to_kind[source_minikind]
 
2074
                    return (entry[0][2],
 
2075
                           (old_path_u, path_u),
 
2076
                           content_change,
 
2077
                           (True, True),
 
2078
                           (source_parent_id, target_parent_id),
 
2079
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2080
                           (source_kind, target_kind),
 
2081
                           (source_exec, target_exec))
 
2082
                else:
 
2083
                    return uninteresting
 
2084
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2085
                # looks like a new file
 
2086
                if path_info is not None:
 
2087
                    path = pathjoin(entry[0][0], entry[0][1])
 
2088
                    # parent id is the entry for the path in the target tree
 
2089
                    # TODO: these are the same for an entire directory: cache em.
 
2090
                    parent_id = state._get_entry(target_index,
 
2091
                                                 path_utf8=entry[0][0])[0][2]
 
2092
                    if parent_id == entry[0][2]:
 
2093
                        parent_id = None
 
2094
                    if use_filesystem_for_exec:
 
2095
                        # We need S_ISREG here, because we aren't sure if this
 
2096
                        # is a file or not.
 
2097
                        target_exec = bool(
 
2098
                            stat.S_ISREG(path_info[3].st_mode)
 
2099
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2100
                    else:
 
2101
                        target_exec = target_details[3]
 
2102
                    return (entry[0][2],
 
2103
                           (None, utf8_decode(path)[0]),
 
2104
                           True,
 
2105
                           (False, True),
 
2106
                           (None, parent_id),
 
2107
                           (None, utf8_decode(entry[0][1])[0]),
 
2108
                           (None, path_info[2]),
 
2109
                           (None, target_exec))
 
2110
                else:
 
2111
                    # but its not on disk: we deliberately treat this as just
 
2112
                    # never-present. (Why ?! - RBC 20070224)
 
2113
                    pass
 
2114
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2115
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2116
                # if its still on disk, *and* theres no other entry at this
 
2117
                # path [we dont know this in this routine at the moment -
 
2118
                # perhaps we should change this - then it would be an unknown.
 
2119
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2120
                # parent id is the entry for the path in the target tree
 
2121
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2122
                if parent_id == entry[0][2]:
 
2123
                    parent_id = None
 
2124
                return (entry[0][2],
 
2125
                       (utf8_decode(old_path)[0], None),
 
2126
                       True,
 
2127
                       (True, False),
 
2128
                       (parent_id, None),
 
2129
                       (utf8_decode(entry[0][1])[0], None),
 
2130
                       (_minikind_to_kind[source_minikind], None),
 
2131
                       (source_details[3], None))
 
2132
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2133
                # a rename; could be a true rename, or a rename inherited from
 
2134
                # a renamed parent. TODO: handle this efficiently. Its not
 
2135
                # common case to rename dirs though, so a correct but slow
 
2136
                # implementation will do.
 
2137
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2138
                    search_specific_files.add(target_details[1])
 
2139
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2140
                # neither of the selected trees contain this file,
 
2141
                # so skip over it. This is not currently directly tested, but
 
2142
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2143
                pass
 
2144
            else:
 
2145
                raise AssertionError("don't know how to compare "
 
2146
                    "source_minikind=%r, target_minikind=%r"
 
2147
                    % (source_minikind, target_minikind))
 
2148
                ## import pdb;pdb.set_trace()
 
2149
            return None
 
2150
 
 
2151
        while search_specific_files:
 
2152
            # TODO: the pending list should be lexically sorted?  the
 
2153
            # interface doesn't require it.
 
2154
            current_root = search_specific_files.pop()
 
2155
            current_root_unicode = current_root.decode('utf8')
 
2156
            searched_specific_files.add(current_root)
 
2157
            # process the entries for this containing directory: the rest will be
 
2158
            # found by their parents recursively.
 
2159
            root_entries = _entries_for_path(current_root)
 
2160
            root_abspath = self.target.abspath(current_root_unicode)
 
2161
            try:
 
2162
                root_stat = os.lstat(root_abspath)
 
2163
            except OSError, e:
 
2164
                if e.errno == errno.ENOENT:
 
2165
                    # the path does not exist: let _process_entry know that.
 
2166
                    root_dir_info = None
 
2167
                else:
 
2168
                    # some other random error: hand it up.
 
2169
                    raise
 
2170
            else:
 
2171
                root_dir_info = ('', current_root,
 
2172
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2173
                    root_abspath)
 
2174
                if root_dir_info[2] == 'directory':
 
2175
                    if self.target._directory_is_tree_reference(
 
2176
                        current_root.decode('utf8')):
 
2177
                        root_dir_info = root_dir_info[:2] + \
 
2178
                            ('tree-reference',) + root_dir_info[3:]
 
2179
 
 
2180
            if not root_entries and not root_dir_info:
 
2181
                # this specified path is not present at all, skip it.
 
2182
                continue
 
2183
            path_handled = False
 
2184
            for entry in root_entries:
 
2185
                result = _process_entry(entry, root_dir_info)
 
2186
                if result is not None:
 
2187
                    path_handled = True
 
2188
                    if result is not uninteresting:
 
2189
                        yield result
 
2190
            if want_unversioned and not path_handled and root_dir_info:
 
2191
                new_executable = bool(
 
2192
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2193
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2194
                yield (None,
 
2195
                       (None, current_root_unicode),
 
2196
                       True,
 
2197
                       (False, False),
 
2198
                       (None, None),
 
2199
                       (None, splitpath(current_root_unicode)[-1]),
 
2200
                       (None, root_dir_info[2]),
 
2201
                       (None, new_executable)
 
2202
                      )
 
2203
            initial_key = (current_root, '', '')
 
2204
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2205
            if block_index == 0:
 
2206
                # we have processed the total root already, but because the
 
2207
                # initial key matched it we should skip it here.
 
2208
                block_index +=1
 
2209
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2210
                current_dir_info = None
 
2211
            else:
 
2212
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2213
                try:
 
2214
                    current_dir_info = dir_iterator.next()
 
2215
                except OSError, e:
 
2216
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2217
                    # python 2.5 has e.errno == EINVAL,
 
2218
                    #            and e.winerror == ERROR_DIRECTORY
 
2219
                    e_winerror = getattr(e, 'winerror', None)
 
2220
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2221
                    # there may be directories in the inventory even though
 
2222
                    # this path is not a file on disk: so mark it as end of
 
2223
                    # iterator
 
2224
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2225
                        current_dir_info = None
 
2226
                    elif (sys.platform == 'win32'
 
2227
                          and (e.errno in win_errors
 
2228
                               or e_winerror in win_errors)):
 
2229
                        current_dir_info = None
 
2230
                    else:
 
2231
                        raise
 
2232
                else:
 
2233
                    if current_dir_info[0][0] == '':
 
2234
                        # remove .bzr from iteration
 
2235
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2236
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2237
                        del current_dir_info[1][bzr_index]
 
2238
            # walk until both the directory listing and the versioned metadata
 
2239
            # are exhausted. 
 
2240
            if (block_index < len(state._dirblocks) and
 
2241
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2242
                current_block = state._dirblocks[block_index]
 
2243
            else:
 
2244
                current_block = None
 
2245
            while (current_dir_info is not None or
 
2246
                   current_block is not None):
 
2247
                if (current_dir_info and current_block
 
2248
                    and current_dir_info[0][0] != current_block[0]):
 
2249
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2250
                        # filesystem data refers to paths not covered by the dirblock.
 
2251
                        # this has two possibilities:
 
2252
                        # A) it is versioned but empty, so there is no block for it
 
2253
                        # B) it is not versioned.
 
2254
 
 
2255
                        # if (A) then we need to recurse into it to check for
 
2256
                        # new unknown files or directories.
 
2257
                        # if (B) then we should ignore it, because we don't
 
2258
                        # recurse into unknown directories.
 
2259
                        path_index = 0
 
2260
                        while path_index < len(current_dir_info[1]):
 
2261
                                current_path_info = current_dir_info[1][path_index]
 
2262
                                if want_unversioned:
 
2263
                                    if current_path_info[2] == 'directory':
 
2264
                                        if self.target._directory_is_tree_reference(
 
2265
                                            current_path_info[0].decode('utf8')):
 
2266
                                            current_path_info = current_path_info[:2] + \
 
2267
                                                ('tree-reference',) + current_path_info[3:]
 
2268
                                    new_executable = bool(
 
2269
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2270
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2271
                                    yield (None,
 
2272
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2273
                                        True,
 
2274
                                        (False, False),
 
2275
                                        (None, None),
 
2276
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2277
                                        (None, current_path_info[2]),
 
2278
                                        (None, new_executable))
 
2279
                                # dont descend into this unversioned path if it is
 
2280
                                # a dir
 
2281
                                if current_path_info[2] in ('directory',
 
2282
                                                            'tree-reference'):
 
2283
                                    del current_dir_info[1][path_index]
 
2284
                                    path_index -= 1
 
2285
                                path_index += 1
 
2286
 
 
2287
                        # This dir info has been handled, go to the next
 
2288
                        try:
 
2289
                            current_dir_info = dir_iterator.next()
 
2290
                        except StopIteration:
 
2291
                            current_dir_info = None
 
2292
                    else:
 
2293
                        # We have a dirblock entry for this location, but there
 
2294
                        # is no filesystem path for this. This is most likely
 
2295
                        # because a directory was removed from the disk.
 
2296
                        # We don't have to report the missing directory,
 
2297
                        # because that should have already been handled, but we
 
2298
                        # need to handle all of the files that are contained
 
2299
                        # within.
 
2300
                        for current_entry in current_block[1]:
 
2301
                            # entry referring to file not present on disk.
 
2302
                            # advance the entry only, after processing.
 
2303
                            result = _process_entry(current_entry, None)
 
2304
                            if result is not None:
 
2305
                                if result is not uninteresting:
 
2306
                                    yield result
 
2307
                        block_index +=1
 
2308
                        if (block_index < len(state._dirblocks) and
 
2309
                            osutils.is_inside(current_root,
 
2310
                                              state._dirblocks[block_index][0])):
 
2311
                            current_block = state._dirblocks[block_index]
 
2312
                        else:
 
2313
                            current_block = None
 
2314
                    continue
 
2315
                entry_index = 0
 
2316
                if current_block and entry_index < len(current_block[1]):
 
2317
                    current_entry = current_block[1][entry_index]
 
2318
                else:
 
2319
                    current_entry = None
 
2320
                advance_entry = True
 
2321
                path_index = 0
 
2322
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2323
                    current_path_info = current_dir_info[1][path_index]
 
2324
                    if current_path_info[2] == 'directory':
 
2325
                        if self.target._directory_is_tree_reference(
 
2326
                            current_path_info[0].decode('utf8')):
 
2327
                            current_path_info = current_path_info[:2] + \
 
2328
                                ('tree-reference',) + current_path_info[3:]
 
2329
                else:
 
2330
                    current_path_info = None
 
2331
                advance_path = True
 
2332
                path_handled = False
 
2333
                while (current_entry is not None or
 
2334
                    current_path_info is not None):
 
2335
                    if current_entry is None:
 
2336
                        # the check for path_handled when the path is adnvaced
 
2337
                        # will yield this path if needed.
 
2338
                        pass
 
2339
                    elif current_path_info is None:
 
2340
                        # no path is fine: the per entry code will handle it.
 
2341
                        result = _process_entry(current_entry, current_path_info)
 
2342
                        if result is not None:
 
2343
                            if result is not uninteresting:
 
2344
                                yield result
 
2345
                    elif (current_entry[0][1] != current_path_info[1]
 
2346
                          or current_entry[1][target_index][0] in 'ar'):
 
2347
                        # The current path on disk doesn't match the dirblock
 
2348
                        # record. Either the dirblock is marked as absent, or
 
2349
                        # the file on disk is not present at all in the
 
2350
                        # dirblock. Either way, report about the dirblock
 
2351
                        # entry, and let other code handle the filesystem one.
 
2352
 
 
2353
                        # Compare the basename for these files to determine
 
2354
                        # which comes first
 
2355
                        if current_path_info[1] < current_entry[0][1]:
 
2356
                            # extra file on disk: pass for now, but only
 
2357
                            # increment the path, not the entry
 
2358
                            advance_entry = False
 
2359
                        else:
 
2360
                            # entry referring to file not present on disk.
 
2361
                            # advance the entry only, after processing.
 
2362
                            result = _process_entry(current_entry, None)
 
2363
                            if result is not None:
 
2364
                                if result is not uninteresting:
 
2365
                                    yield result
 
2366
                            advance_path = False
 
2367
                    else:
 
2368
                        result = _process_entry(current_entry, current_path_info)
 
2369
                        if result is not None:
 
2370
                            path_handled = True
 
2371
                            if result is not uninteresting:
 
2372
                                yield result
 
2373
                    if advance_entry and current_entry is not None:
 
2374
                        entry_index += 1
 
2375
                        if entry_index < len(current_block[1]):
 
2376
                            current_entry = current_block[1][entry_index]
 
2377
                        else:
 
2378
                            current_entry = None
 
2379
                    else:
 
2380
                        advance_entry = True # reset the advance flaga
 
2381
                    if advance_path and current_path_info is not None:
 
2382
                        if not path_handled:
 
2383
                            # unversioned in all regards
 
2384
                            if want_unversioned:
 
2385
                                new_executable = bool(
 
2386
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2387
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2388
                                yield (None,
 
2389
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2390
                                    True,
 
2391
                                    (False, False),
 
2392
                                    (None, None),
 
2393
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2394
                                    (None, current_path_info[2]),
 
2395
                                    (None, new_executable))
 
2396
                            # dont descend into this unversioned path if it is
 
2397
                            # a dir
 
2398
                            if current_path_info[2] in ('directory'):
 
2399
                                del current_dir_info[1][path_index]
 
2400
                                path_index -= 1
 
2401
                        # dont descend the disk iterator into any tree 
 
2402
                        # paths.
 
2403
                        if current_path_info[2] == 'tree-reference':
 
2404
                            del current_dir_info[1][path_index]
 
2405
                            path_index -= 1
 
2406
                        path_index += 1
 
2407
                        if path_index < len(current_dir_info[1]):
 
2408
                            current_path_info = current_dir_info[1][path_index]
 
2409
                            if current_path_info[2] == 'directory':
 
2410
                                if self.target._directory_is_tree_reference(
 
2411
                                    current_path_info[0].decode('utf8')):
 
2412
                                    current_path_info = current_path_info[:2] + \
 
2413
                                        ('tree-reference',) + current_path_info[3:]
 
2414
                        else:
 
2415
                            current_path_info = None
 
2416
                        path_handled = False
 
2417
                    else:
 
2418
                        advance_path = True # reset the advance flagg.
 
2419
                if current_block is not None:
 
2420
                    block_index += 1
 
2421
                    if (block_index < len(state._dirblocks) and
 
2422
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2423
                        current_block = state._dirblocks[block_index]
 
2424
                    else:
 
2425
                        current_block = None
 
2426
                if current_dir_info is not None:
 
2427
                    try:
 
2428
                        current_dir_info = dir_iterator.next()
 
2429
                    except StopIteration:
 
2430
                        current_dir_info = None
 
2431
 
 
2432
 
 
2433
    @staticmethod
 
2434
    def is_compatible(source, target):
 
2435
        # the target must be a dirstate working tree
 
2436
        if not isinstance(target, WorkingTree4):
 
2437
            return False
 
2438
        # the source must be a revtreee or dirstate rev tree.
 
2439
        if not isinstance(source,
 
2440
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2441
            return False
 
2442
        # the source revid must be in the target dirstate
 
2443
        if not (source._revision_id == NULL_REVISION or
 
2444
            source._revision_id in target.get_parent_ids()):
 
2445
            # TODO: what about ghosts? it may well need to 
 
2446
            # check for them explicitly.
 
2447
            return False
 
2448
        return True
 
2449
 
 
2450
InterTree.register_optimiser(InterDirStateTree)
 
2451
 
 
2452
 
 
2453
class Converter3to4(object):
 
2454
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2455
 
 
2456
    def __init__(self):
 
2457
        self.target_format = WorkingTreeFormat4()
 
2458
 
 
2459
    def convert(self, tree):
 
2460
        # lock the control files not the tree, so that we dont get tree
 
2461
        # on-unlock behaviours, and so that noone else diddles with the 
 
2462
        # tree during upgrade.
 
2463
        tree._control_files.lock_write()
 
2464
        try:
 
2465
            tree.read_working_inventory()
 
2466
            self.create_dirstate_data(tree)
 
2467
            self.update_format(tree)
 
2468
            self.remove_xml_files(tree)
 
2469
        finally:
 
2470
            tree._control_files.unlock()
 
2471
 
 
2472
    def create_dirstate_data(self, tree):
 
2473
        """Create the dirstate based data for tree."""
 
2474
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2475
            ).local_abspath('dirstate')
 
2476
        state = dirstate.DirState.from_tree(tree, local_path)
 
2477
        state.save()
 
2478
        state.unlock()
 
2479
 
 
2480
    def remove_xml_files(self, tree):
 
2481
        """Remove the oldformat 3 data."""
 
2482
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2483
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2484
            'pending-merges', 'stat-cache']:
 
2485
            try:
 
2486
                transport.delete(path)
 
2487
            except errors.NoSuchFile:
 
2488
                # some files are optional - just deal.
 
2489
                pass
 
2490
 
 
2491
    def update_format(self, tree):
 
2492
        """Change the format marker."""
 
2493
        tree._control_files.put_utf8('format',
 
2494
            self.target_format.get_format_string())