/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

Fixed as per John's review.

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
        basedir = safe_unicode(basedir)
 
135
        mutter("opening working tree %r", basedir)
 
136
        self._branch = branch
 
137
        self.basedir = realpath(basedir)
 
138
        # if branch is at our basedir and is a format 6 or less
 
139
        # assume all other formats have their own control files.
 
140
        self._control_files = _control_files
 
141
        self._transport = self._control_files._transport
 
142
        self._dirty = None
 
143
        #-------------
 
144
        # during a read or write lock these objects are set, and are
 
145
        # None the rest of the time.
 
146
        self._dirstate = None
 
147
        self._inventory = None
 
148
        #-------------
 
149
        self._setup_directory_is_tree_reference()
 
150
        self._detect_case_handling()
 
151
 
 
152
    @needs_tree_write_lock
 
153
    def _add(self, files, ids, kinds):
 
154
        """See MutableTree._add."""
 
155
        state = self.current_dirstate()
 
156
        for f, file_id, kind in zip(files, ids, kinds):
 
157
            f = f.strip('/')
 
158
            if self.path2id(f):
 
159
                # special case tree root handling.
 
160
                if f == '' and self.path2id(f) == ROOT_ID:
 
161
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
162
                continue
 
163
            if file_id is None:
 
164
                file_id = generate_ids.gen_file_id(f)
 
165
            # deliberately add the file with no cached stat or sha1
 
166
            # - on the first access it will be gathered, and we can
 
167
            # always change this once tests are all passing.
 
168
            state.add(f, file_id, kind, None, '')
 
169
        self._make_dirty(reset_inventory=True)
 
170
 
 
171
    def _make_dirty(self, reset_inventory):
 
172
        """Make the tree state dirty.
 
173
 
 
174
        :param reset_inventory: True if the cached inventory should be removed
 
175
            (presuming there is one).
 
176
        """
 
177
        self._dirty = True
 
178
        if reset_inventory and self._inventory is not None:
 
179
            self._inventory = None
 
180
 
 
181
    @needs_tree_write_lock
 
182
    def add_reference(self, sub_tree):
 
183
        # use standard implementation, which calls back to self._add
 
184
        # 
 
185
        # So we don't store the reference_revision in the working dirstate,
 
186
        # it's just recorded at the moment of commit. 
 
187
        self._add_reference(sub_tree)
 
188
 
 
189
    def break_lock(self):
 
190
        """Break a lock if one is present from another instance.
 
191
 
 
192
        Uses the ui factory to ask for confirmation if the lock may be from
 
193
        an active process.
 
194
 
 
195
        This will probe the repository for its lock as well.
 
196
        """
 
197
        # if the dirstate is locked by an active process, reject the break lock
 
198
        # call.
 
199
        try:
 
200
            if self._dirstate is None:
 
201
                clear = True
 
202
            else:
 
203
                clear = False
 
204
            state = self._current_dirstate()
 
205
            if state._lock_token is not None:
 
206
                # we already have it locked. sheese, cant break our own lock.
 
207
                raise errors.LockActive(self.basedir)
 
208
            else:
 
209
                try:
 
210
                    # try for a write lock - need permission to get one anyhow
 
211
                    # to break locks.
 
212
                    state.lock_write()
 
213
                except errors.LockContention:
 
214
                    # oslocks fail when a process is still live: fail.
 
215
                    # TODO: get the locked lockdir info and give to the user to
 
216
                    # assist in debugging.
 
217
                    raise errors.LockActive(self.basedir)
 
218
                else:
 
219
                    state.unlock()
 
220
        finally:
 
221
            if clear:
 
222
                self._dirstate = None
 
223
        self._control_files.break_lock()
 
224
        self.branch.break_lock()
 
225
 
 
226
    def _comparison_data(self, entry, path):
 
227
        kind, executable, stat_value = \
 
228
            WorkingTree3._comparison_data(self, entry, path)
 
229
        # it looks like a plain directory, but it's really a reference -- see
 
230
        # also kind()
 
231
        if (self._repo_supports_tree_reference and
 
232
            kind == 'directory' and
 
233
            self._directory_is_tree_reference(path)):
 
234
            kind = 'tree-reference'
 
235
        return kind, executable, stat_value
 
236
 
 
237
    @needs_write_lock
 
238
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
239
        # mark the tree as dirty post commit - commit
 
240
        # can change the current versioned list by doing deletes.
 
241
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
242
        self._make_dirty(reset_inventory=True)
 
243
        return result
 
244
 
 
245
    def current_dirstate(self):
 
246
        """Return the current dirstate object.
 
247
 
 
248
        This is not part of the tree interface and only exposed for ease of
 
249
        testing.
 
250
 
 
251
        :raises errors.NotWriteLocked: when not in a lock.
 
252
        """
 
253
        self._must_be_locked()
 
254
        return self._current_dirstate()
 
255
 
 
256
    def _current_dirstate(self):
 
257
        """Internal function that does not check lock status.
 
258
 
 
259
        This is needed for break_lock which also needs the dirstate.
 
260
        """
 
261
        if self._dirstate is not None:
 
262
            return self._dirstate
 
263
        local_path = self.bzrdir.get_workingtree_transport(None
 
264
            ).local_abspath('dirstate')
 
265
        self._dirstate = dirstate.DirState.on_file(local_path)
 
266
        return self._dirstate
 
267
 
 
268
    def filter_unversioned_files(self, paths):
 
269
        """Filter out paths that are versioned.
 
270
 
 
271
        :return: set of paths.
 
272
        """
 
273
        # TODO: make a generic multi-bisect routine roughly that should list
 
274
        # the paths, then process one half at a time recursively, and feed the
 
275
        # results of each bisect in further still
 
276
        paths = sorted(paths)
 
277
        result = set()
 
278
        state = self.current_dirstate()
 
279
        # TODO we want a paths_to_dirblocks helper I think
 
280
        for path in paths:
 
281
            dirname, basename = os.path.split(path.encode('utf8'))
 
282
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
283
                dirname, basename, 0)
 
284
            if not path_is_versioned:
 
285
                result.add(path)
 
286
        return result
 
287
 
 
288
    def flush(self):
 
289
        """Write all cached data to disk."""
 
290
        if self._control_files._lock_mode != 'w':
 
291
            raise errors.NotWriteLocked(self)
 
292
        self.current_dirstate().save()
 
293
        self._inventory = None
 
294
        self._dirty = False
 
295
 
 
296
    @needs_tree_write_lock
 
297
    def _gather_kinds(self, files, kinds):
 
298
        """See MutableTree._gather_kinds."""
 
299
        for pos, f in enumerate(files):
 
300
            if kinds[pos] is None:
 
301
                kinds[pos] = self._kind(f)
 
302
 
 
303
    def _generate_inventory(self):
 
304
        """Create and set self.inventory from the dirstate object.
 
305
        
 
306
        This is relatively expensive: we have to walk the entire dirstate.
 
307
        Ideally we would not, and can deprecate this function.
 
308
        """
 
309
        #: uncomment to trap on inventory requests.
 
310
        # import pdb;pdb.set_trace()
 
311
        state = self.current_dirstate()
 
312
        state._read_dirblocks_if_needed()
 
313
        root_key, current_entry = self._get_entry(path='')
 
314
        current_id = root_key[2]
 
315
        if not (current_entry[0][0] == 'd'): # directory
 
316
            raise AssertionError(current_entry)
 
317
        inv = Inventory(root_id=current_id)
 
318
        # Turn some things into local variables
 
319
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
320
        factory = entry_factory
 
321
        utf8_decode = cache_utf8._utf8_decode
 
322
        inv_byid = inv._byid
 
323
        # we could do this straight out of the dirstate; it might be fast
 
324
        # and should be profiled - RBC 20070216
 
325
        parent_ies = {'' : inv.root}
 
326
        for block in state._dirblocks[1:]: # skip the root
 
327
            dirname = block[0]
 
328
            try:
 
329
                parent_ie = parent_ies[dirname]
 
330
            except KeyError:
 
331
                # all the paths in this block are not versioned in this tree
 
332
                continue
 
333
            for key, entry in block[1]:
 
334
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
335
                if minikind in ('a', 'r'): # absent, relocated
 
336
                    # a parent tree only entry
 
337
                    continue
 
338
                name = key[1]
 
339
                name_unicode = utf8_decode(name)[0]
 
340
                file_id = key[2]
 
341
                kind = minikind_to_kind[minikind]
 
342
                inv_entry = factory[kind](file_id, name_unicode,
 
343
                                          parent_ie.file_id)
 
344
                if kind == 'file':
 
345
                    # This is only needed on win32, where this is the only way
 
346
                    # we know the executable bit.
 
347
                    inv_entry.executable = executable
 
348
                    # not strictly needed: working tree
 
349
                    #inv_entry.text_size = size
 
350
                    #inv_entry.text_sha1 = sha1
 
351
                elif kind == 'directory':
 
352
                    # add this entry to the parent map.
 
353
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
354
                elif kind == 'tree-reference':
 
355
                    if not self._repo_supports_tree_reference:
 
356
                        raise AssertionError(
 
357
                            "repository of %r "
 
358
                            "doesn't support tree references "
 
359
                            "required by entry %r"
 
360
                            % (self, name))
 
361
                    inv_entry.reference_revision = link_or_sha1 or None
 
362
                elif kind != 'symlink':
 
363
                    raise AssertionError("unknown kind %r" % kind)
 
364
                # These checks cost us around 40ms on a 55k entry tree
 
365
                if file_id in inv_byid:
 
366
                    raise AssertionError('file_id %s already in'
 
367
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                if name_unicode in parent_ie.children:
 
369
                    raise AssertionError('name %r already in parent'
 
370
                        % (name_unicode,))
 
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
    def all_file_ids(self):
 
504
        """See Tree.iter_all_file_ids"""
 
505
        self._must_be_locked()
 
506
        result = set()
 
507
        for key, tree_details in self.current_dirstate()._iter_entries():
 
508
            if tree_details[0][0] in ('a', 'r'): # relocated
 
509
                continue
 
510
            result.add(key[2])
 
511
        return result
 
512
 
 
513
    @needs_read_lock
 
514
    def __iter__(self):
 
515
        """Iterate through file_ids for this tree.
 
516
 
 
517
        file_ids are in a WorkingTree if they are in the working inventory
 
518
        and the working file exists.
 
519
        """
 
520
        result = []
 
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
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
526
            if osutils.lexists(path):
 
527
                result.append(key[2])
 
528
        return iter(result)
 
529
 
 
530
    def iter_references(self):
 
531
        if not self._repo_supports_tree_reference:
 
532
            # When the repo doesn't support references, we will have nothing to
 
533
            # return
 
534
            return
 
535
        for key, tree_details in self.current_dirstate()._iter_entries():
 
536
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
537
                # not relevant to the working tree
 
538
                continue
 
539
            if not key[1]:
 
540
                # the root is not a reference.
 
541
                continue
 
542
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
543
            try:
 
544
                if self._kind(relpath) == 'tree-reference':
 
545
                    yield relpath, key[2]
 
546
            except errors.NoSuchFile:
 
547
                # path is missing on disk.
 
548
                continue
 
549
 
 
550
    def kind(self, file_id):
 
551
        """Return the kind of a file.
 
552
 
 
553
        This is always the actual kind that's on disk, regardless of what it
 
554
        was added as.
 
555
 
 
556
        Note: The caller is expected to take a read-lock before calling this.
 
557
        """
 
558
        relpath = self.id2path(file_id)
 
559
        if relpath is None:
 
560
            raise AssertionError(
 
561
                "path for id {%s} is None!" % file_id)
 
562
        return self._kind(relpath)
 
563
 
 
564
    def _kind(self, relpath):
 
565
        abspath = self.abspath(relpath)
 
566
        kind = file_kind(abspath)
 
567
        if (self._repo_supports_tree_reference and
 
568
            kind == 'directory' and
 
569
            self._directory_is_tree_reference(relpath)):
 
570
            kind = 'tree-reference'
 
571
        return kind
 
572
 
 
573
    @needs_read_lock
 
574
    def _last_revision(self):
 
575
        """See Mutable.last_revision."""
 
576
        parent_ids = self.current_dirstate().get_parent_ids()
 
577
        if parent_ids:
 
578
            return parent_ids[0]
 
579
        else:
 
580
            return _mod_revision.NULL_REVISION
 
581
 
 
582
    def lock_read(self):
 
583
        """See Branch.lock_read, and WorkingTree.unlock."""
 
584
        self.branch.lock_read()
 
585
        try:
 
586
            self._control_files.lock_read()
 
587
            try:
 
588
                state = self.current_dirstate()
 
589
                if not state._lock_token:
 
590
                    state.lock_read()
 
591
                # set our support for tree references from the repository in
 
592
                # use.
 
593
                self._repo_supports_tree_reference = getattr(
 
594
                    self.branch.repository._format, "supports_tree_reference",
 
595
                    False)
 
596
            except:
 
597
                self._control_files.unlock()
 
598
                raise
 
599
        except:
 
600
            self.branch.unlock()
 
601
            raise
 
602
 
 
603
    def _lock_self_write(self):
 
604
        """This should be called after the branch is locked."""
 
605
        try:
 
606
            self._control_files.lock_write()
 
607
            try:
 
608
                state = self.current_dirstate()
 
609
                if not state._lock_token:
 
610
                    state.lock_write()
 
611
                # set our support for tree references from the repository in
 
612
                # use.
 
613
                self._repo_supports_tree_reference = getattr(
 
614
                    self.branch.repository._format, "supports_tree_reference",
 
615
                    False)
 
616
            except:
 
617
                self._control_files.unlock()
 
618
                raise
 
619
        except:
 
620
            self.branch.unlock()
 
621
            raise
 
622
 
 
623
    def lock_tree_write(self):
 
624
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
625
        self.branch.lock_read()
 
626
        self._lock_self_write()
 
627
 
 
628
    def lock_write(self):
 
629
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
630
        self.branch.lock_write()
 
631
        self._lock_self_write()
 
632
 
 
633
    @needs_tree_write_lock
 
634
    def move(self, from_paths, to_dir, after=False):
 
635
        """See WorkingTree.move()."""
 
636
        result = []
 
637
        if not from_paths:
 
638
            return result
 
639
        state = self.current_dirstate()
 
640
        if isinstance(from_paths, basestring):
 
641
            raise ValueError()
 
642
        to_dir_utf8 = to_dir.encode('utf8')
 
643
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
644
        id_index = state._get_id_index()
 
645
        # check destination directory
 
646
        # get the details for it
 
647
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
648
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
649
        if not entry_present:
 
650
            raise errors.BzrMoveFailedError('', to_dir,
 
651
                errors.NotVersionedError(to_dir))
 
652
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
653
        # get a handle on the block itself.
 
654
        to_block_index = state._ensure_block(
 
655
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
656
        to_block = state._dirblocks[to_block_index]
 
657
        to_abs = self.abspath(to_dir)
 
658
        if not isdir(to_abs):
 
659
            raise errors.BzrMoveFailedError('',to_dir,
 
660
                errors.NotADirectory(to_abs))
 
661
 
 
662
        if to_entry[1][0][0] != 'd':
 
663
            raise errors.BzrMoveFailedError('',to_dir,
 
664
                errors.NotADirectory(to_abs))
 
665
 
 
666
        if self._inventory is not None:
 
667
            update_inventory = True
 
668
            inv = self.inventory
 
669
            to_dir_id = to_entry[0][2]
 
670
            to_dir_ie = inv[to_dir_id]
 
671
        else:
 
672
            update_inventory = False
 
673
 
 
674
        rollbacks = []
 
675
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
676
                     fingerprint, packed_stat, size,
 
677
                     to_block, to_key, to_path_utf8):
 
678
            state._make_absent(old_entry)
 
679
            from_key = old_entry[0]
 
680
            rollbacks.append(
 
681
                lambda:state.update_minimal(from_key,
 
682
                    minikind,
 
683
                    executable=executable,
 
684
                    fingerprint=fingerprint,
 
685
                    packed_stat=packed_stat,
 
686
                    size=size,
 
687
                    path_utf8=from_path_utf8))
 
688
            state.update_minimal(to_key,
 
689
                    minikind,
 
690
                    executable=executable,
 
691
                    fingerprint=fingerprint,
 
692
                    packed_stat=packed_stat,
 
693
                    size=size,
 
694
                    path_utf8=to_path_utf8)
 
695
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
696
            new_entry = to_block[1][added_entry_index]
 
697
            rollbacks.append(lambda:state._make_absent(new_entry))
 
698
 
 
699
        for from_rel in from_paths:
 
700
            # from_rel is 'pathinroot/foo/bar'
 
701
            from_rel_utf8 = from_rel.encode('utf8')
 
702
            from_dirname, from_tail = osutils.split(from_rel)
 
703
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
704
            from_entry = self._get_entry(path=from_rel)
 
705
            if from_entry == (None, None):
 
706
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
707
                    errors.NotVersionedError(path=str(from_rel)))
 
708
 
 
709
            from_id = from_entry[0][2]
 
710
            to_rel = pathjoin(to_dir, from_tail)
 
711
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
712
            item_to_entry = self._get_entry(path=to_rel)
 
713
            if item_to_entry != (None, None):
 
714
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
715
                    "Target is already versioned.")
 
716
 
 
717
            if from_rel == to_rel:
 
718
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
719
                    "Source and target are identical.")
 
720
 
 
721
            from_missing = not self.has_filename(from_rel)
 
722
            to_missing = not self.has_filename(to_rel)
 
723
            if after:
 
724
                move_file = False
 
725
            else:
 
726
                move_file = True
 
727
            if to_missing:
 
728
                if not move_file:
 
729
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
730
                        errors.NoSuchFile(path=to_rel,
 
731
                        extra="New file has not been created yet"))
 
732
                elif from_missing:
 
733
                    # neither path exists
 
734
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
735
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
736
            else:
 
737
                if from_missing: # implicitly just update our path mapping
 
738
                    move_file = False
 
739
                elif not after:
 
740
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
741
 
 
742
            rollbacks = []
 
743
            def rollback_rename():
 
744
                """A single rename has failed, roll it back."""
 
745
                # roll back everything, even if we encounter trouble doing one
 
746
                # of them.
 
747
                #
 
748
                # TODO: at least log the other exceptions rather than just
 
749
                # losing them mbp 20070307
 
750
                exc_info = None
 
751
                for rollback in reversed(rollbacks):
 
752
                    try:
 
753
                        rollback()
 
754
                    except Exception, e:
 
755
                        exc_info = sys.exc_info()
 
756
                if exc_info:
 
757
                    raise exc_info[0], exc_info[1], exc_info[2]
 
758
 
 
759
            # perform the disk move first - its the most likely failure point.
 
760
            if move_file:
 
761
                from_rel_abs = self.abspath(from_rel)
 
762
                to_rel_abs = self.abspath(to_rel)
 
763
                try:
 
764
                    osutils.rename(from_rel_abs, to_rel_abs)
 
765
                except OSError, e:
 
766
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
767
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
768
            try:
 
769
                # perform the rename in the inventory next if needed: its easy
 
770
                # to rollback
 
771
                if update_inventory:
 
772
                    # rename the entry
 
773
                    from_entry = inv[from_id]
 
774
                    current_parent = from_entry.parent_id
 
775
                    inv.rename(from_id, to_dir_id, from_tail)
 
776
                    rollbacks.append(
 
777
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
778
                # finally do the rename in the dirstate, which is a little
 
779
                # tricky to rollback, but least likely to need it.
 
780
                old_block_index, old_entry_index, dir_present, file_present = \
 
781
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
782
                old_block = state._dirblocks[old_block_index][1]
 
783
                old_entry = old_block[old_entry_index]
 
784
                from_key, old_entry_details = old_entry
 
785
                cur_details = old_entry_details[0]
 
786
                # remove the old row
 
787
                to_key = ((to_block[0],) + from_key[1:3])
 
788
                minikind = cur_details[0]
 
789
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
790
                         minikind=minikind,
 
791
                         executable=cur_details[3],
 
792
                         fingerprint=cur_details[1],
 
793
                         packed_stat=cur_details[4],
 
794
                         size=cur_details[2],
 
795
                         to_block=to_block,
 
796
                         to_key=to_key,
 
797
                         to_path_utf8=to_rel_utf8)
 
798
 
 
799
                if minikind == 'd':
 
800
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
801
                        """Recursively update all entries in this dirblock."""
 
802
                        if from_dir == '':
 
803
                            raise AssertionError("renaming root not supported")
 
804
                        from_key = (from_dir, '')
 
805
                        from_block_idx, present = \
 
806
                            state._find_block_index_from_key(from_key)
 
807
                        if not present:
 
808
                            # This is the old record, if it isn't present, then
 
809
                            # there is theoretically nothing to update.
 
810
                            # (Unless it isn't present because of lazy loading,
 
811
                            # but we don't do that yet)
 
812
                            return
 
813
                        from_block = state._dirblocks[from_block_idx]
 
814
                        to_block_index, to_entry_index, _, _ = \
 
815
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
816
                        to_block_index = state._ensure_block(
 
817
                            to_block_index, to_entry_index, to_dir_utf8)
 
818
                        to_block = state._dirblocks[to_block_index]
 
819
 
 
820
                        # Grab a copy since move_one may update the list.
 
821
                        for entry in from_block[1][:]:
 
822
                            if not (entry[0][0] == from_dir):
 
823
                                raise AssertionError()
 
824
                            cur_details = entry[1][0]
 
825
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
826
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
827
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
828
                            minikind = cur_details[0]
 
829
                            if minikind in 'ar':
 
830
                                # Deleted children of a renamed directory
 
831
                                # Do not need to be updated.
 
832
                                # Children that have been renamed out of this
 
833
                                # directory should also not be updated
 
834
                                continue
 
835
                            move_one(entry, from_path_utf8=from_path_utf8,
 
836
                                     minikind=minikind,
 
837
                                     executable=cur_details[3],
 
838
                                     fingerprint=cur_details[1],
 
839
                                     packed_stat=cur_details[4],
 
840
                                     size=cur_details[2],
 
841
                                     to_block=to_block,
 
842
                                     to_key=to_key,
 
843
                                     to_path_utf8=to_path_utf8)
 
844
                            if minikind == 'd':
 
845
                                # We need to move all the children of this
 
846
                                # entry
 
847
                                update_dirblock(from_path_utf8, to_key,
 
848
                                                to_path_utf8)
 
849
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
850
            except:
 
851
                rollback_rename()
 
852
                raise
 
853
            result.append((from_rel, to_rel))
 
854
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
855
            self._make_dirty(reset_inventory=False)
 
856
 
 
857
        return result
 
858
 
 
859
    def _must_be_locked(self):
 
860
        if not self._control_files._lock_count:
 
861
            raise errors.ObjectNotLocked(self)
 
862
 
 
863
    def _new_tree(self):
 
864
        """Initialize the state in this tree to be a new tree."""
 
865
        self._dirty = True
 
866
 
 
867
    @needs_read_lock
 
868
    def path2id(self, path):
 
869
        """Return the id for path in this tree."""
 
870
        path = path.strip('/')
 
871
        entry = self._get_entry(path=path)
 
872
        if entry == (None, None):
 
873
            return None
 
874
        return entry[0][2]
 
875
 
 
876
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
877
        """See Tree.paths2ids().
 
878
 
 
879
        This specialisation fast-paths the case where all the trees are in the
 
880
        dirstate.
 
881
        """
 
882
        if paths is None:
 
883
            return None
 
884
        parents = self.get_parent_ids()
 
885
        for tree in trees:
 
886
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
887
                parents):
 
888
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
889
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
890
        # -- make all paths utf8 --
 
891
        paths_utf8 = set()
 
892
        for path in paths:
 
893
            paths_utf8.add(path.encode('utf8'))
 
894
        paths = paths_utf8
 
895
        # -- paths is now a utf8 path set --
 
896
        # -- get the state object and prepare it.
 
897
        state = self.current_dirstate()
 
898
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
899
            and '' not in paths):
 
900
            paths2ids = self._paths2ids_using_bisect
 
901
        else:
 
902
            paths2ids = self._paths2ids_in_memory
 
903
        return paths2ids(paths, search_indexes,
 
904
                         require_versioned=require_versioned)
 
905
 
 
906
    def _paths2ids_in_memory(self, paths, search_indexes,
 
907
                             require_versioned=True):
 
908
        state = self.current_dirstate()
 
909
        state._read_dirblocks_if_needed()
 
910
        def _entries_for_path(path):
 
911
            """Return a list with all the entries that match path for all ids.
 
912
            """
 
913
            dirname, basename = os.path.split(path)
 
914
            key = (dirname, basename, '')
 
915
            block_index, present = state._find_block_index_from_key(key)
 
916
            if not present:
 
917
                # the block which should contain path is absent.
 
918
                return []
 
919
            result = []
 
920
            block = state._dirblocks[block_index][1]
 
921
            entry_index, _ = state._find_entry_index(key, block)
 
922
            # we may need to look at multiple entries at this path: walk while the paths match.
 
923
            while (entry_index < len(block) and
 
924
                block[entry_index][0][0:2] == key[0:2]):
 
925
                result.append(block[entry_index])
 
926
                entry_index += 1
 
927
            return result
 
928
        if require_versioned:
 
929
            # -- check all supplied paths are versioned in a search tree. --
 
930
            all_versioned = True
 
931
            for path in paths:
 
932
                path_entries = _entries_for_path(path)
 
933
                if not path_entries:
 
934
                    # this specified path is not present at all: error
 
935
                    all_versioned = False
 
936
                    break
 
937
                found_versioned = False
 
938
                # for each id at this path
 
939
                for entry in path_entries:
 
940
                    # for each tree.
 
941
                    for index in search_indexes:
 
942
                        if entry[1][index][0] != 'a': # absent
 
943
                            found_versioned = True
 
944
                            # all good: found a versioned cell
 
945
                            break
 
946
                if not found_versioned:
 
947
                    # none of the indexes was not 'absent' at all ids for this
 
948
                    # path.
 
949
                    all_versioned = False
 
950
                    break
 
951
            if not all_versioned:
 
952
                raise errors.PathsNotVersionedError(paths)
 
953
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
954
        search_paths = osutils.minimum_path_selection(paths)
 
955
        # sketch: 
 
956
        # for all search_indexs in each path at or under each element of
 
957
        # search_paths, if the detail is relocated: add the id, and add the
 
958
        # relocated path as one to search if its not searched already. If the
 
959
        # detail is not relocated, add the id.
 
960
        searched_paths = set()
 
961
        found_ids = set()
 
962
        def _process_entry(entry):
 
963
            """Look at search_indexes within entry.
 
964
 
 
965
            If a specific tree's details are relocated, add the relocation
 
966
            target to search_paths if not searched already. If it is absent, do
 
967
            nothing. Otherwise add the id to found_ids.
 
968
            """
 
969
            for index in search_indexes:
 
970
                if entry[1][index][0] == 'r': # relocated
 
971
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
972
                        search_paths.add(entry[1][index][1])
 
973
                elif entry[1][index][0] != 'a': # absent
 
974
                    found_ids.add(entry[0][2])
 
975
        while search_paths:
 
976
            current_root = search_paths.pop()
 
977
            searched_paths.add(current_root)
 
978
            # process the entries for this containing directory: the rest will be
 
979
            # found by their parents recursively.
 
980
            root_entries = _entries_for_path(current_root)
 
981
            if not root_entries:
 
982
                # this specified path is not present at all, skip it.
 
983
                continue
 
984
            for entry in root_entries:
 
985
                _process_entry(entry)
 
986
            initial_key = (current_root, '', '')
 
987
            block_index, _ = state._find_block_index_from_key(initial_key)
 
988
            while (block_index < len(state._dirblocks) and
 
989
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
990
                for entry in state._dirblocks[block_index][1]:
 
991
                    _process_entry(entry)
 
992
                block_index += 1
 
993
        return found_ids
 
994
 
 
995
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
996
                                require_versioned=True):
 
997
        state = self.current_dirstate()
 
998
        found_ids = set()
 
999
 
 
1000
        split_paths = sorted(osutils.split(p) for p in paths)
 
1001
        found = state._bisect_recursive(split_paths)
 
1002
 
 
1003
        if require_versioned:
 
1004
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
1005
            for dir_name in split_paths:
 
1006
                if dir_name not in found_dir_names:
 
1007
                    raise errors.PathsNotVersionedError(paths)
 
1008
 
 
1009
        for dir_name_id, trees_info in found.iteritems():
 
1010
            for index in search_indexes:
 
1011
                if trees_info[index][0] not in ('r', 'a'):
 
1012
                    found_ids.add(dir_name_id[2])
 
1013
        return found_ids
 
1014
 
 
1015
    def read_working_inventory(self):
 
1016
        """Read the working inventory.
 
1017
        
 
1018
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1019
        """
 
1020
        return self.inventory
 
1021
 
 
1022
    @needs_read_lock
 
1023
    def revision_tree(self, revision_id):
 
1024
        """See Tree.revision_tree.
 
1025
 
 
1026
        WorkingTree4 supplies revision_trees for any basis tree.
 
1027
        """
 
1028
        dirstate = self.current_dirstate()
 
1029
        parent_ids = dirstate.get_parent_ids()
 
1030
        if revision_id not in parent_ids:
 
1031
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1032
        if revision_id in dirstate.get_ghosts():
 
1033
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1034
        return DirStateRevisionTree(dirstate, revision_id,
 
1035
            self.branch.repository)
 
1036
 
 
1037
    @needs_tree_write_lock
 
1038
    def set_last_revision(self, new_revision):
 
1039
        """Change the last revision in the working tree."""
 
1040
        parents = self.get_parent_ids()
 
1041
        if new_revision in (NULL_REVISION, None):
 
1042
            if len(parents) >= 2:
 
1043
                raise AssertionError(
 
1044
                    "setting the last parent to none with a pending merge is "
 
1045
                    "unsupported.")
 
1046
            self.set_parent_ids([])
 
1047
        else:
 
1048
            self.set_parent_ids([new_revision] + parents[1:],
 
1049
                allow_leftmost_as_ghost=True)
 
1050
 
 
1051
    @needs_tree_write_lock
 
1052
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1053
        """Set the parent ids to revision_ids.
 
1054
        
 
1055
        See also set_parent_trees. This api will try to retrieve the tree data
 
1056
        for each element of revision_ids from the trees repository. If you have
 
1057
        tree data already available, it is more efficient to use
 
1058
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1059
        an easier API to use.
 
1060
 
 
1061
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1062
            working tree. Any of these may be ghosts.
 
1063
        """
 
1064
        trees = []
 
1065
        for revision_id in revision_ids:
 
1066
            try:
 
1067
                revtree = self.branch.repository.revision_tree(revision_id)
 
1068
                # TODO: jam 20070213 KnitVersionedFile raises
 
1069
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1070
                #       given revision_id is not present. Should Repository be
 
1071
                #       catching it and re-raising NoSuchRevision?
 
1072
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1073
                revtree = None
 
1074
            trees.append((revision_id, revtree))
 
1075
        self.set_parent_trees(trees,
 
1076
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1077
 
 
1078
    @needs_tree_write_lock
 
1079
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1080
        """Set the parents of the working tree.
 
1081
 
 
1082
        :param parents_list: A list of (revision_id, tree) tuples.
 
1083
            If tree is None, then that element is treated as an unreachable
 
1084
            parent tree - i.e. a ghost.
 
1085
        """
 
1086
        dirstate = self.current_dirstate()
 
1087
        if len(parents_list) > 0:
 
1088
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1089
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1090
        real_trees = []
 
1091
        ghosts = []
 
1092
 
 
1093
        parent_ids = [rev_id for rev_id, tree in parents_list]
 
1094
        graph = self.branch.repository.get_graph()
 
1095
        heads = graph.heads(parent_ids)
 
1096
        accepted_revisions = set()
 
1097
 
 
1098
        # convert absent trees to the null tree, which we convert back to
 
1099
        # missing on access.
 
1100
        for rev_id, tree in parents_list:
 
1101
            if len(accepted_revisions) > 0:
 
1102
                # we always accept the first tree
 
1103
                if rev_id in accepted_revisions or rev_id not in heads:
 
1104
                    # We have already included either this tree, or its
 
1105
                    # descendent, so we skip it.
 
1106
                    continue
 
1107
            _mod_revision.check_not_reserved_id(rev_id)
 
1108
            if tree is not None:
 
1109
                real_trees.append((rev_id, tree))
 
1110
            else:
 
1111
                real_trees.append((rev_id,
 
1112
                    self.branch.repository.revision_tree(None)))
 
1113
                ghosts.append(rev_id)
 
1114
            accepted_revisions.add(rev_id)
 
1115
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1116
        self._make_dirty(reset_inventory=False)
 
1117
 
 
1118
    def _set_root_id(self, file_id):
 
1119
        """See WorkingTree.set_root_id."""
 
1120
        state = self.current_dirstate()
 
1121
        state.set_path_id('', file_id)
 
1122
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1123
            self._make_dirty(reset_inventory=True)
 
1124
 
 
1125
    def _sha_from_stat(self, path, stat_result):
 
1126
        """Get a sha digest from the tree's stat cache.
 
1127
 
 
1128
        The default implementation assumes no stat cache is present.
 
1129
 
 
1130
        :param path: The path.
 
1131
        :param stat_result: The stat result being looked up.
 
1132
        """
 
1133
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1134
 
 
1135
    @needs_read_lock
 
1136
    def supports_tree_reference(self):
 
1137
        return self._repo_supports_tree_reference
 
1138
 
 
1139
    def unlock(self):
 
1140
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1141
        # do non-implementation specific cleanup
 
1142
        self._cleanup()
 
1143
 
 
1144
        if self._control_files._lock_count == 1:
 
1145
            # eventually we should do signature checking during read locks for
 
1146
            # dirstate updates.
 
1147
            if self._control_files._lock_mode == 'w':
 
1148
                if self._dirty:
 
1149
                    self.flush()
 
1150
            if self._dirstate is not None:
 
1151
                # This is a no-op if there are no modifications.
 
1152
                self._dirstate.save()
 
1153
                self._dirstate.unlock()
 
1154
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1155
            #       point. Instead, it could check if the header has been
 
1156
            #       modified when it is locked, and if not, it can hang on to
 
1157
            #       the data it has in memory.
 
1158
            self._dirstate = None
 
1159
            self._inventory = None
 
1160
        # reverse order of locking.
 
1161
        try:
 
1162
            return self._control_files.unlock()
 
1163
        finally:
 
1164
            self.branch.unlock()
 
1165
 
 
1166
    @needs_tree_write_lock
 
1167
    def unversion(self, file_ids):
 
1168
        """Remove the file ids in file_ids from the current versioned set.
 
1169
 
 
1170
        When a file_id is unversioned, all of its children are automatically
 
1171
        unversioned.
 
1172
 
 
1173
        :param file_ids: The file ids to stop versioning.
 
1174
        :raises: NoSuchId if any fileid is not currently versioned.
 
1175
        """
 
1176
        if not file_ids:
 
1177
            return
 
1178
        state = self.current_dirstate()
 
1179
        state._read_dirblocks_if_needed()
 
1180
        ids_to_unversion = set(file_ids)
 
1181
        paths_to_unversion = set()
 
1182
        # sketch:
 
1183
        # check if the root is to be unversioned, if so, assert for now.
 
1184
        # walk the state marking unversioned things as absent.
 
1185
        # if there are any un-unversioned ids at the end, raise
 
1186
        for key, details in state._dirblocks[0][1]:
 
1187
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1188
                key[2] in ids_to_unversion):
 
1189
                # I haven't written the code to unversion / yet - it should be
 
1190
                # supported.
 
1191
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1192
        block_index = 0
 
1193
        while block_index < len(state._dirblocks):
 
1194
            # process one directory at a time.
 
1195
            block = state._dirblocks[block_index]
 
1196
            # first check: is the path one to remove - it or its children
 
1197
            delete_block = False
 
1198
            for path in paths_to_unversion:
 
1199
                if (block[0].startswith(path) and
 
1200
                    (len(block[0]) == len(path) or
 
1201
                     block[0][len(path)] == '/')):
 
1202
                    # this entire block should be deleted - its the block for a
 
1203
                    # path to unversion; or the child of one
 
1204
                    delete_block = True
 
1205
                    break
 
1206
            # TODO: trim paths_to_unversion as we pass by paths
 
1207
            if delete_block:
 
1208
                # this block is to be deleted: process it.
 
1209
                # TODO: we can special case the no-parents case and
 
1210
                # just forget the whole block.
 
1211
                entry_index = 0
 
1212
                while entry_index < len(block[1]):
 
1213
                    # Mark this file id as having been removed
 
1214
                    entry = block[1][entry_index]
 
1215
                    ids_to_unversion.discard(entry[0][2])
 
1216
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1217
                                               # entries
 
1218
                        or not state._make_absent(entry)):
 
1219
                        entry_index += 1
 
1220
                # go to the next block. (At the moment we dont delete empty
 
1221
                # dirblocks)
 
1222
                block_index += 1
 
1223
                continue
 
1224
            entry_index = 0
 
1225
            while entry_index < len(block[1]):
 
1226
                entry = block[1][entry_index]
 
1227
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1228
                    # ^ some parent row.
 
1229
                    entry[0][2] not in ids_to_unversion):
 
1230
                    # ^ not an id to unversion
 
1231
                    entry_index += 1
 
1232
                    continue
 
1233
                if entry[1][0][0] == 'd':
 
1234
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1235
                if not state._make_absent(entry):
 
1236
                    entry_index += 1
 
1237
                # we have unversioned this id
 
1238
                ids_to_unversion.remove(entry[0][2])
 
1239
            block_index += 1
 
1240
        if ids_to_unversion:
 
1241
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1242
        self._make_dirty(reset_inventory=False)
 
1243
        # have to change the legacy inventory too.
 
1244
        if self._inventory is not None:
 
1245
            for file_id in file_ids:
 
1246
                self._inventory.remove_recursive_id(file_id)
 
1247
 
 
1248
    @needs_tree_write_lock
 
1249
    def rename_one(self, from_rel, to_rel, after=False):
 
1250
        """See WorkingTree.rename_one"""
 
1251
        self.flush()
 
1252
        WorkingTree.rename_one(self, from_rel, to_rel, after)
 
1253
 
 
1254
    @needs_tree_write_lock
 
1255
    def apply_inventory_delta(self, changes):
 
1256
        """See MutableTree.apply_inventory_delta"""
 
1257
        state = self.current_dirstate()
 
1258
        state.update_by_delta(changes)
 
1259
        self._make_dirty(reset_inventory=True)
 
1260
 
 
1261
    def update_basis_by_delta(self, new_revid, delta):
 
1262
        """See MutableTree.update_basis_by_delta."""
 
1263
        if self.last_revision() == new_revid:
 
1264
            raise AssertionError()
 
1265
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1266
 
 
1267
    @needs_read_lock
 
1268
    def _validate(self):
 
1269
        self._dirstate._validate()
 
1270
 
 
1271
    @needs_tree_write_lock
 
1272
    def _write_inventory(self, inv):
 
1273
        """Write inventory as the current inventory."""
 
1274
        if self._dirty:
 
1275
            raise AssertionError("attempting to write an inventory when the "
 
1276
                "dirstate is dirty will lose pending changes")
 
1277
        self.current_dirstate().set_state_from_inventory(inv)
 
1278
        self._make_dirty(reset_inventory=False)
 
1279
        if self._inventory is not None:
 
1280
            self._inventory = inv
 
1281
        self.flush()
 
1282
 
 
1283
 
 
1284
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1285
    """The first consolidated dirstate working tree format.
 
1286
 
 
1287
    This format:
 
1288
        - exists within a metadir controlling .bzr
 
1289
        - includes an explicit version marker for the workingtree control
 
1290
          files, separate from the BzrDir format
 
1291
        - modifies the hash cache format
 
1292
        - is new in bzr 0.15
 
1293
        - uses a LockDir to guard access to it.
 
1294
    """
 
1295
 
 
1296
    upgrade_recommended = False
 
1297
 
 
1298
    def get_format_string(self):
 
1299
        """See WorkingTreeFormat.get_format_string()."""
 
1300
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1301
 
 
1302
    def get_format_description(self):
 
1303
        """See WorkingTreeFormat.get_format_description()."""
 
1304
        return "Working tree format 4"
 
1305
 
 
1306
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1307
                   accelerator_tree=None, hardlink=False):
 
1308
        """See WorkingTreeFormat.initialize().
 
1309
 
 
1310
        :param revision_id: allows creating a working tree at a different
 
1311
        revision than the branch is at.
 
1312
        :param accelerator_tree: A tree which can be used for retrieving file
 
1313
            contents more quickly than the revision tree, i.e. a workingtree.
 
1314
            The revision tree will be used for cases where accelerator_tree's
 
1315
            content is different.
 
1316
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1317
            where possible.
 
1318
 
 
1319
        These trees get an initial random root id, if their repository supports
 
1320
        rich root data, TREE_ROOT otherwise.
 
1321
        """
 
1322
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1323
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1324
        transport = a_bzrdir.get_workingtree_transport(self)
 
1325
        control_files = self._open_control_files(a_bzrdir)
 
1326
        control_files.create_lock()
 
1327
        control_files.lock_write()
 
1328
        transport.put_bytes('format', self.get_format_string(),
 
1329
            mode=a_bzrdir._get_file_mode())
 
1330
        if from_branch is not None:
 
1331
            branch = from_branch
 
1332
        else:
 
1333
            branch = a_bzrdir.open_branch()
 
1334
        if revision_id is None:
 
1335
            revision_id = branch.last_revision()
 
1336
        local_path = transport.local_abspath('dirstate')
 
1337
        # write out new dirstate (must exist when we create the tree)
 
1338
        state = dirstate.DirState.initialize(local_path)
 
1339
        state.unlock()
 
1340
        del state
 
1341
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1342
                         branch,
 
1343
                         _format=self,
 
1344
                         _bzrdir=a_bzrdir,
 
1345
                         _control_files=control_files)
 
1346
        wt._new_tree()
 
1347
        wt.lock_tree_write()
 
1348
        try:
 
1349
            if revision_id in (None, NULL_REVISION):
 
1350
                if branch.repository.supports_rich_root():
 
1351
                    wt._set_root_id(generate_ids.gen_root_id())
 
1352
                else:
 
1353
                    wt._set_root_id(ROOT_ID)
 
1354
                wt.flush()
 
1355
            basis = None
 
1356
            # frequently, we will get here due to branching.  The accelerator
 
1357
            # tree will be the tree from the branch, so the desired basis
 
1358
            # tree will often be a parent of the accelerator tree.
 
1359
            if accelerator_tree is not None:
 
1360
                try:
 
1361
                    basis = accelerator_tree.revision_tree(revision_id)
 
1362
                except errors.NoSuchRevision:
 
1363
                    pass
 
1364
            if basis is None:
 
1365
                basis = branch.repository.revision_tree(revision_id)
 
1366
            if revision_id == NULL_REVISION:
 
1367
                parents_list = []
 
1368
            else:
 
1369
                parents_list = [(revision_id, basis)]
 
1370
            basis.lock_read()
 
1371
            try:
 
1372
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1373
                wt.flush()
 
1374
                # if the basis has a root id we have to use that; otherwise we
 
1375
                # use a new random one
 
1376
                basis_root_id = basis.get_root_id()
 
1377
                if basis_root_id is not None:
 
1378
                    wt._set_root_id(basis_root_id)
 
1379
                    wt.flush()
 
1380
                # delta_from_tree is safe even for DirStateRevisionTrees,
 
1381
                # because wt4.apply_inventory_delta does not mutate the input
 
1382
                # inventory entries.
 
1383
                transform.build_tree(basis, wt, accelerator_tree,
 
1384
                                     hardlink=hardlink, delta_from_tree=True)
 
1385
            finally:
 
1386
                basis.unlock()
 
1387
        finally:
 
1388
            control_files.unlock()
 
1389
            wt.unlock()
 
1390
        return wt
 
1391
 
 
1392
    def _open(self, a_bzrdir, control_files):
 
1393
        """Open the tree itself.
 
1394
 
 
1395
        :param a_bzrdir: the dir for the tree.
 
1396
        :param control_files: the control files for the tree.
 
1397
        """
 
1398
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1399
                           branch=a_bzrdir.open_branch(),
 
1400
                           _format=self,
 
1401
                           _bzrdir=a_bzrdir,
 
1402
                           _control_files=control_files)
 
1403
 
 
1404
    def __get_matchingbzrdir(self):
 
1405
        # please test against something that will let us do tree references
 
1406
        return bzrdir.format_registry.make_bzrdir(
 
1407
            'dirstate-with-subtree')
 
1408
 
 
1409
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1410
 
 
1411
 
 
1412
class DirStateRevisionTree(Tree):
 
1413
    """A revision tree pulling the inventory from a dirstate."""
 
1414
 
 
1415
    def __init__(self, dirstate, revision_id, repository):
 
1416
        self._dirstate = dirstate
 
1417
        self._revision_id = revision_id
 
1418
        self._repository = repository
 
1419
        self._inventory = None
 
1420
        self._locked = 0
 
1421
        self._dirstate_locked = False
 
1422
        self._repo_supports_tree_reference = getattr(
 
1423
            repository._format, "supports_tree_reference",
 
1424
            False)
 
1425
 
 
1426
    def __repr__(self):
 
1427
        return "<%s of %s in %s>" % \
 
1428
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1429
 
 
1430
    def annotate_iter(self, file_id,
 
1431
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1432
        """See Tree.annotate_iter"""
 
1433
        w = self._get_weave(file_id)
 
1434
        return w.annotate(self.inventory[file_id].revision)
 
1435
 
 
1436
    def _get_ancestors(self, default_revision):
 
1437
        return set(self._repository.get_ancestry(self._revision_id,
 
1438
                                                 topo_sorted=False))
 
1439
    def _comparison_data(self, entry, path):
 
1440
        """See Tree._comparison_data."""
 
1441
        if entry is None:
 
1442
            return None, False, None
 
1443
        # trust the entry as RevisionTree does, but this may not be
 
1444
        # sensible: the entry might not have come from us?
 
1445
        return entry.kind, entry.executable, None
 
1446
 
 
1447
    def _file_size(self, entry, stat_value):
 
1448
        return entry.text_size
 
1449
 
 
1450
    def filter_unversioned_files(self, paths):
 
1451
        """Filter out paths that are not versioned.
 
1452
 
 
1453
        :return: set of paths.
 
1454
        """
 
1455
        pred = self.has_filename
 
1456
        return set((p for p in paths if not pred(p)))
 
1457
 
 
1458
    def get_root_id(self):
 
1459
        return self.path2id('')
 
1460
 
 
1461
    def id2path(self, file_id):
 
1462
        "Convert a file-id to a path."
 
1463
        entry = self._get_entry(file_id=file_id)
 
1464
        if entry == (None, None):
 
1465
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1466
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1467
        return path_utf8.decode('utf8')
 
1468
 
 
1469
    def iter_references(self):
 
1470
        if not self._repo_supports_tree_reference:
 
1471
            # When the repo doesn't support references, we will have nothing to
 
1472
            # return
 
1473
            return iter([])
 
1474
        # Otherwise, fall back to the default implementation
 
1475
        return super(DirStateRevisionTree, self).iter_references()
 
1476
 
 
1477
    def _get_parent_index(self):
 
1478
        """Return the index in the dirstate referenced by this tree."""
 
1479
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1480
 
 
1481
    def _get_entry(self, file_id=None, path=None):
 
1482
        """Get the dirstate row for file_id or path.
 
1483
 
 
1484
        If either file_id or path is supplied, it is used as the key to lookup.
 
1485
        If both are supplied, the fastest lookup is used, and an error is
 
1486
        raised if they do not both point at the same row.
 
1487
        
 
1488
        :param file_id: An optional unicode file_id to be looked up.
 
1489
        :param path: An optional unicode path to be looked up.
 
1490
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1491
        """
 
1492
        if file_id is None and path is None:
 
1493
            raise errors.BzrError('must supply file_id or path')
 
1494
        if path is not None:
 
1495
            path = path.encode('utf8')
 
1496
        parent_index = self._get_parent_index()
 
1497
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1498
 
 
1499
    def _generate_inventory(self):
 
1500
        """Create and set self.inventory from the dirstate object.
 
1501
 
 
1502
        (So this is only called the first time the inventory is requested for
 
1503
        this tree; it then remains in memory until it's out of date.)
 
1504
 
 
1505
        This is relatively expensive: we have to walk the entire dirstate.
 
1506
        """
 
1507
        if not self._locked:
 
1508
            raise AssertionError(
 
1509
                'cannot generate inventory of an unlocked '
 
1510
                'dirstate revision tree')
 
1511
        # separate call for profiling - makes it clear where the costs are.
 
1512
        self._dirstate._read_dirblocks_if_needed()
 
1513
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1514
            raise AssertionError(
 
1515
                'parent %s has disappeared from %s' % (
 
1516
                self._revision_id, self._dirstate.get_parent_ids()))
 
1517
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1518
        # This is identical now to the WorkingTree _generate_inventory except
 
1519
        # for the tree index use.
 
1520
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1521
        current_id = root_key[2]
 
1522
        if current_entry[parent_index][0] != 'd':
 
1523
            raise AssertionError()
 
1524
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1525
        inv.root.revision = current_entry[parent_index][4]
 
1526
        # Turn some things into local variables
 
1527
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1528
        factory = entry_factory
 
1529
        utf8_decode = cache_utf8._utf8_decode
 
1530
        inv_byid = inv._byid
 
1531
        # we could do this straight out of the dirstate; it might be fast
 
1532
        # and should be profiled - RBC 20070216
 
1533
        parent_ies = {'' : inv.root}
 
1534
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1535
            dirname = block[0]
 
1536
            try:
 
1537
                parent_ie = parent_ies[dirname]
 
1538
            except KeyError:
 
1539
                # all the paths in this block are not versioned in this tree
 
1540
                continue
 
1541
            for key, entry in block[1]:
 
1542
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1543
                if minikind in ('a', 'r'): # absent, relocated
 
1544
                    # not this tree
 
1545
                    continue
 
1546
                name = key[1]
 
1547
                name_unicode = utf8_decode(name)[0]
 
1548
                file_id = key[2]
 
1549
                kind = minikind_to_kind[minikind]
 
1550
                inv_entry = factory[kind](file_id, name_unicode,
 
1551
                                          parent_ie.file_id)
 
1552
                inv_entry.revision = revid
 
1553
                if kind == 'file':
 
1554
                    inv_entry.executable = executable
 
1555
                    inv_entry.text_size = size
 
1556
                    inv_entry.text_sha1 = fingerprint
 
1557
                elif kind == 'directory':
 
1558
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1559
                elif kind == 'symlink':
 
1560
                    inv_entry.executable = False
 
1561
                    inv_entry.text_size = None
 
1562
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1563
                elif kind == 'tree-reference':
 
1564
                    inv_entry.reference_revision = fingerprint or None
 
1565
                else:
 
1566
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1567
                            % entry)
 
1568
                # These checks cost us around 40ms on a 55k entry tree
 
1569
                if file_id in inv_byid:
 
1570
                    raise AssertionError('file_id %s already in'
 
1571
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1572
                if name_unicode in parent_ie.children:
 
1573
                    raise AssertionError('name %r already in parent'
 
1574
                        % (name_unicode,))
 
1575
                inv_byid[file_id] = inv_entry
 
1576
                parent_ie.children[name_unicode] = inv_entry
 
1577
        self._inventory = inv
 
1578
 
 
1579
    def get_file_mtime(self, file_id, path=None):
 
1580
        """Return the modification time for this record.
 
1581
 
 
1582
        We return the timestamp of the last-changed revision.
 
1583
        """
 
1584
        # Make sure the file exists
 
1585
        entry = self._get_entry(file_id, path=path)
 
1586
        if entry == (None, None): # do we raise?
 
1587
            return None
 
1588
        parent_index = self._get_parent_index()
 
1589
        last_changed_revision = entry[1][parent_index][4]
 
1590
        return self._repository.get_revision(last_changed_revision).timestamp
 
1591
 
 
1592
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1593
        entry = self._get_entry(file_id=file_id, path=path)
 
1594
        parent_index = self._get_parent_index()
 
1595
        parent_details = entry[1][parent_index]
 
1596
        if parent_details[0] == 'f':
 
1597
            return parent_details[1]
 
1598
        return None
 
1599
 
 
1600
    def _get_weave(self, file_id):
 
1601
        return self._repository.weave_store.get_weave(file_id,
 
1602
                self._repository.get_transaction())
 
1603
 
 
1604
    def get_file(self, file_id, path=None):
 
1605
        return StringIO(self.get_file_text(file_id))
 
1606
 
 
1607
    def get_file_lines(self, file_id):
 
1608
        entry = self._get_entry(file_id=file_id)[1]
 
1609
        if entry is None:
 
1610
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1611
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1612
 
 
1613
    def get_file_size(self, file_id):
 
1614
        """See Tree.get_file_size"""
 
1615
        return self.inventory[file_id].text_size
 
1616
 
 
1617
    def get_file_text(self, file_id):
 
1618
        return ''.join(self.get_file_lines(file_id))
 
1619
 
 
1620
    def get_reference_revision(self, file_id, path=None):
 
1621
        return self.inventory[file_id].reference_revision
 
1622
 
 
1623
    def iter_files_bytes(self, desired_files):
 
1624
        """See Tree.iter_files_bytes.
 
1625
 
 
1626
        This version is implemented on top of Repository.iter_files_bytes"""
 
1627
        parent_index = self._get_parent_index()
 
1628
        repo_desired_files = []
 
1629
        for file_id, identifier in desired_files:
 
1630
            entry = self._get_entry(file_id)
 
1631
            if entry == (None, None):
 
1632
                raise errors.NoSuchId(self, file_id)
 
1633
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1634
                                       identifier))
 
1635
        return self._repository.iter_files_bytes(repo_desired_files)
 
1636
 
 
1637
    def get_symlink_target(self, file_id):
 
1638
        entry = self._get_entry(file_id=file_id)
 
1639
        parent_index = self._get_parent_index()
 
1640
        if entry[1][parent_index][0] != 'l':
 
1641
            return None
 
1642
        else:
 
1643
            # At present, none of the tree implementations supports non-ascii
 
1644
            # symlink targets. So we will just assume that the dirstate path is
 
1645
            # correct.
 
1646
            return entry[1][parent_index][1]
 
1647
 
 
1648
    def get_revision_id(self):
 
1649
        """Return the revision id for this tree."""
 
1650
        return self._revision_id
 
1651
 
 
1652
    def _get_inventory(self):
 
1653
        if self._inventory is not None:
 
1654
            return self._inventory
 
1655
        self._must_be_locked()
 
1656
        self._generate_inventory()
 
1657
        return self._inventory
 
1658
 
 
1659
    inventory = property(_get_inventory,
 
1660
                         doc="Inventory of this Tree")
 
1661
 
 
1662
    def get_parent_ids(self):
 
1663
        """The parents of a tree in the dirstate are not cached."""
 
1664
        return self._repository.get_revision(self._revision_id).parent_ids
 
1665
 
 
1666
    def has_filename(self, filename):
 
1667
        return bool(self.path2id(filename))
 
1668
 
 
1669
    def kind(self, file_id):
 
1670
        entry = self._get_entry(file_id=file_id)[1]
 
1671
        if entry is None:
 
1672
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1673
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1674
 
 
1675
    def stored_kind(self, file_id):
 
1676
        """See Tree.stored_kind"""
 
1677
        return self.kind(file_id)
 
1678
 
 
1679
    def path_content_summary(self, path):
 
1680
        """See Tree.path_content_summary."""
 
1681
        id = self.inventory.path2id(path)
 
1682
        if id is None:
 
1683
            return ('missing', None, None, None)
 
1684
        entry = self._inventory[id]
 
1685
        kind = entry.kind
 
1686
        if kind == 'file':
 
1687
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1688
        elif kind == 'symlink':
 
1689
            return (kind, None, None, entry.symlink_target)
 
1690
        else:
 
1691
            return (kind, None, None, None)
 
1692
 
 
1693
    def is_executable(self, file_id, path=None):
 
1694
        ie = self.inventory[file_id]
 
1695
        if ie.kind != "file":
 
1696
            return None
 
1697
        return ie.executable
 
1698
 
 
1699
    def list_files(self, include_root=False):
 
1700
        # We use a standard implementation, because DirStateRevisionTree is
 
1701
        # dealing with one of the parents of the current state
 
1702
        inv = self._get_inventory()
 
1703
        entries = inv.iter_entries()
 
1704
        if self.inventory.root is not None and not include_root:
 
1705
            entries.next()
 
1706
        for path, entry in entries:
 
1707
            yield path, 'V', entry.kind, entry.file_id, entry
 
1708
 
 
1709
    def lock_read(self):
 
1710
        """Lock the tree for a set of operations."""
 
1711
        if not self._locked:
 
1712
            self._repository.lock_read()
 
1713
            if self._dirstate._lock_token is None:
 
1714
                self._dirstate.lock_read()
 
1715
                self._dirstate_locked = True
 
1716
        self._locked += 1
 
1717
 
 
1718
    def _must_be_locked(self):
 
1719
        if not self._locked:
 
1720
            raise errors.ObjectNotLocked(self)
 
1721
 
 
1722
    @needs_read_lock
 
1723
    def path2id(self, path):
 
1724
        """Return the id for path in this tree."""
 
1725
        # lookup by path: faster than splitting and walking the ivnentory.
 
1726
        entry = self._get_entry(path=path)
 
1727
        if entry == (None, None):
 
1728
            return None
 
1729
        return entry[0][2]
 
1730
 
 
1731
    def unlock(self):
 
1732
        """Unlock, freeing any cache memory used during the lock."""
 
1733
        # outside of a lock, the inventory is suspect: release it.
 
1734
        self._locked -=1
 
1735
        if not self._locked:
 
1736
            self._inventory = None
 
1737
            self._locked = 0
 
1738
            if self._dirstate_locked:
 
1739
                self._dirstate.unlock()
 
1740
                self._dirstate_locked = False
 
1741
            self._repository.unlock()
 
1742
 
 
1743
    @needs_read_lock
 
1744
    def supports_tree_reference(self):
 
1745
        return self._repo_supports_tree_reference
 
1746
 
 
1747
    def walkdirs(self, prefix=""):
 
1748
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1749
        # implementation based on an inventory.  
 
1750
        # This should be cleaned up to use the much faster Dirstate code
 
1751
        # So for now, we just build up the parent inventory, and extract
 
1752
        # it the same way RevisionTree does.
 
1753
        _directory = 'directory'
 
1754
        inv = self._get_inventory()
 
1755
        top_id = inv.path2id(prefix)
 
1756
        if top_id is None:
 
1757
            pending = []
 
1758
        else:
 
1759
            pending = [(prefix, top_id)]
 
1760
        while pending:
 
1761
            dirblock = []
 
1762
            relpath, file_id = pending.pop()
 
1763
            # 0 - relpath, 1- file-id
 
1764
            if relpath:
 
1765
                relroot = relpath + '/'
 
1766
            else:
 
1767
                relroot = ""
 
1768
            # FIXME: stash the node in pending
 
1769
            entry = inv[file_id]
 
1770
            for name, child in entry.sorted_children():
 
1771
                toppath = relroot + name
 
1772
                dirblock.append((toppath, name, child.kind, None,
 
1773
                    child.file_id, child.kind
 
1774
                    ))
 
1775
            yield (relpath, entry.file_id), dirblock
 
1776
            # push the user specified dirs from dirblock
 
1777
            for dir in reversed(dirblock):
 
1778
                if dir[2] == _directory:
 
1779
                    pending.append((dir[0], dir[4]))
 
1780
 
 
1781
 
 
1782
class InterDirStateTree(InterTree):
 
1783
    """Fast path optimiser for changes_from with dirstate trees.
 
1784
    
 
1785
    This is used only when both trees are in the dirstate working file, and 
 
1786
    the source is any parent within the dirstate, and the destination is 
 
1787
    the current working tree of the same dirstate.
 
1788
    """
 
1789
    # this could be generalized to allow comparisons between any trees in the
 
1790
    # dirstate, and possibly between trees stored in different dirstates.
 
1791
 
 
1792
    def __init__(self, source, target):
 
1793
        super(InterDirStateTree, self).__init__(source, target)
 
1794
        if not InterDirStateTree.is_compatible(source, target):
 
1795
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1796
 
 
1797
    @staticmethod
 
1798
    def make_source_parent_tree(source, target):
 
1799
        """Change the source tree into a parent of the target."""
 
1800
        revid = source.commit('record tree')
 
1801
        target.branch.repository.fetch(source.branch.repository, revid)
 
1802
        target.set_parent_ids([revid])
 
1803
        return target.basis_tree(), target
 
1804
 
 
1805
    _matching_from_tree_format = WorkingTreeFormat4()
 
1806
    _matching_to_tree_format = WorkingTreeFormat4()
 
1807
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1808
 
 
1809
    def iter_changes(self, include_unchanged=False,
 
1810
                      specific_files=None, pb=None, extra_trees=[],
 
1811
                      require_versioned=True, want_unversioned=False):
 
1812
        """Return the changes from source to target.
 
1813
 
 
1814
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1815
            for details.
 
1816
        :param specific_files: An optional list of file paths to restrict the
 
1817
            comparison to. When mapping filenames to ids, all matches in all
 
1818
            trees (including optional extra_trees) are used, and all children of
 
1819
            matched directories are included.
 
1820
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1821
            unchanged entries in the result.
 
1822
        :param extra_trees: An optional list of additional trees to use when
 
1823
            mapping the contents of specific_files (paths) to file_ids.
 
1824
        :param require_versioned: If True, all files in specific_files must be
 
1825
            versioned in one of source, target, extra_trees or
 
1826
            PathsNotVersionedError is raised.
 
1827
        :param want_unversioned: Should unversioned files be returned in the
 
1828
            output. An unversioned file is defined as one with (False, False)
 
1829
            for the versioned pair.
 
1830
        """
 
1831
        utf8_decode = cache_utf8._utf8_decode
 
1832
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1833
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1834
        # NB: show_status depends on being able to pass in non-versioned files
 
1835
        # and report them as unknown
 
1836
        # TODO: handle extra trees in the dirstate.
 
1837
        if (extra_trees or specific_files == []):
 
1838
            # we can't fast-path these cases (yet)
 
1839
            for f in super(InterDirStateTree, self).iter_changes(
 
1840
                include_unchanged, specific_files, pb, extra_trees,
 
1841
                require_versioned, want_unversioned=want_unversioned):
 
1842
                yield f
 
1843
            return
 
1844
        parent_ids = self.target.get_parent_ids()
 
1845
        if not (self.source._revision_id in parent_ids
 
1846
                or self.source._revision_id == NULL_REVISION):
 
1847
            raise AssertionError(
 
1848
                "revision {%s} is not stored in {%s}, but %s "
 
1849
                "can only be used for trees stored in the dirstate"
 
1850
                % (self.source._revision_id, self.target, self.iter_changes))
 
1851
        target_index = 0
 
1852
        if self.source._revision_id == NULL_REVISION:
 
1853
            source_index = None
 
1854
            indices = (target_index,)
 
1855
        else:
 
1856
            if not (self.source._revision_id in parent_ids):
 
1857
                raise AssertionError(
 
1858
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1859
                    self.source._revision_id, parent_ids))
 
1860
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1861
            indices = (source_index, target_index)
 
1862
        # -- make all specific_files utf8 --
 
1863
        if specific_files:
 
1864
            specific_files_utf8 = set()
 
1865
            for path in specific_files:
 
1866
                specific_files_utf8.add(path.encode('utf8'))
 
1867
            specific_files = specific_files_utf8
 
1868
        else:
 
1869
            specific_files = set([''])
 
1870
        # -- specific_files is now a utf8 path set --
 
1871
        # -- get the state object and prepare it.
 
1872
        state = self.target.current_dirstate()
 
1873
        state._read_dirblocks_if_needed()
 
1874
        def _entries_for_path(path):
 
1875
            """Return a list with all the entries that match path for all ids.
 
1876
            """
 
1877
            dirname, basename = os.path.split(path)
 
1878
            key = (dirname, basename, '')
 
1879
            block_index, present = state._find_block_index_from_key(key)
 
1880
            if not present:
 
1881
                # the block which should contain path is absent.
 
1882
                return []
 
1883
            result = []
 
1884
            block = state._dirblocks[block_index][1]
 
1885
            entry_index, _ = state._find_entry_index(key, block)
 
1886
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1887
            while (entry_index < len(block) and
 
1888
                block[entry_index][0][0:2] == key[0:2]):
 
1889
                result.append(block[entry_index])
 
1890
                entry_index += 1
 
1891
            return result
 
1892
        if require_versioned:
 
1893
            # -- check all supplied paths are versioned in a search tree. --
 
1894
            all_versioned = True
 
1895
            for path in specific_files:
 
1896
                path_entries = _entries_for_path(path)
 
1897
                if not path_entries:
 
1898
                    # this specified path is not present at all: error
 
1899
                    all_versioned = False
 
1900
                    break
 
1901
                found_versioned = False
 
1902
                # for each id at this path
 
1903
                for entry in path_entries:
 
1904
                    # for each tree.
 
1905
                    for index in indices:
 
1906
                        if entry[1][index][0] != 'a': # absent
 
1907
                            found_versioned = True
 
1908
                            # all good: found a versioned cell
 
1909
                            break
 
1910
                if not found_versioned:
 
1911
                    # none of the indexes was not 'absent' at all ids for this
 
1912
                    # path.
 
1913
                    all_versioned = False
 
1914
                    break
 
1915
            if not all_versioned:
 
1916
                raise errors.PathsNotVersionedError(specific_files)
 
1917
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1918
        search_specific_files = set()
 
1919
        for path in specific_files:
 
1920
            other_specific_files = specific_files.difference(set([path]))
 
1921
            if not osutils.is_inside_any(other_specific_files, path):
 
1922
                # this is a top level path, we must check it.
 
1923
                search_specific_files.add(path)
 
1924
        # sketch: 
 
1925
        # compare source_index and target_index at or under each element of search_specific_files.
 
1926
        # follow the following comparison table. Note that we only want to do diff operations when
 
1927
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1928
        # for the target.
 
1929
        # cases:
 
1930
        # 
 
1931
        # Source | Target | disk | action
 
1932
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1933
        #        |        |      | diff check on source-target
 
1934
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1935
        #        |        |      | ???
 
1936
        #   r    |  a     |      | add source to search
 
1937
        #   r    |  a     |  a   | 
 
1938
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1939
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1940
        #   a    | fdlt   |      | add new id
 
1941
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1942
        #   a    |  a     |      | not present in either tree, skip
 
1943
        #   a    |  a     |  a   | not present in any tree, skip
 
1944
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1945
        #        |        |      | may not be selected by the users list of paths.
 
1946
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1947
        #        |        |      | may not be selected by the users list of paths.
 
1948
        #  fdlt  | fdlt   |      | content in both: diff them
 
1949
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1950
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1951
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1952
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1953
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1954
        #        |        |      | this id at the other path.
 
1955
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1956
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1957
        #        |        |      | this id at the other path.
 
1958
 
 
1959
        # for all search_indexs in each path at or under each element of
 
1960
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1961
        # relocated path as one to search if its not searched already. If the
 
1962
        # detail is not relocated, add the id.
 
1963
        searched_specific_files = set()
 
1964
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1965
        # Using a list so that we can access the values and change them in
 
1966
        # nested scope. Each one is [path, file_id, entry]
 
1967
        last_source_parent = [None, None]
 
1968
        last_target_parent = [None, None]
 
1969
 
 
1970
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1971
 
 
1972
        # Just a sentry, so that _process_entry can say that this
 
1973
        # record is handled, but isn't interesting to process (unchanged)
 
1974
        uninteresting = object()
 
1975
 
 
1976
 
 
1977
        old_dirname_to_file_id = {}
 
1978
        new_dirname_to_file_id = {}
 
1979
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1980
        #       keeping a cache of directories that we have seen.
 
1981
 
 
1982
        def _process_entry(entry, path_info):
 
1983
            """Compare an entry and real disk to generate delta information.
 
1984
 
 
1985
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1986
                the path of entry. If None, then the path is considered absent.
 
1987
                (Perhaps we should pass in a concrete entry for this ?)
 
1988
                Basename is returned as a utf8 string because we expect this
 
1989
                tuple will be ignored, and don't want to take the time to
 
1990
                decode.
 
1991
            :return: None if these don't match
 
1992
                     A tuple of information about the change, or
 
1993
                     the object 'uninteresting' if these match, but are
 
1994
                     basically identical.
 
1995
            """
 
1996
            if source_index is None:
 
1997
                source_details = NULL_PARENT_DETAILS
 
1998
            else:
 
1999
                source_details = entry[1][source_index]
 
2000
            target_details = entry[1][target_index]
 
2001
            target_minikind = target_details[0]
 
2002
            if path_info is not None and target_minikind in 'fdlt':
 
2003
                if not (target_index == 0):
 
2004
                    raise AssertionError()
 
2005
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
2006
                                                  stat_value=path_info[3])
 
2007
                # The entry may have been modified by update_entry
 
2008
                target_details = entry[1][target_index]
 
2009
                target_minikind = target_details[0]
 
2010
            else:
 
2011
                link_or_sha1 = None
 
2012
            file_id = entry[0][2]
 
2013
            source_minikind = source_details[0]
 
2014
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
2015
                # claimed content in both: diff
 
2016
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
2017
                #        |        |      | diff check on source-target
 
2018
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
2019
                #        |        |      | ???
 
2020
                if source_minikind in 'r':
 
2021
                    # add the source to the search path to find any children it
 
2022
                    # has.  TODO ? : only add if it is a container ?
 
2023
                    if not osutils.is_inside_any(searched_specific_files,
 
2024
                                                 source_details[1]):
 
2025
                        search_specific_files.add(source_details[1])
 
2026
                    # generate the old path; this is needed for stating later
 
2027
                    # as well.
 
2028
                    old_path = source_details[1]
 
2029
                    old_dirname, old_basename = os.path.split(old_path)
 
2030
                    path = pathjoin(entry[0][0], entry[0][1])
 
2031
                    old_entry = state._get_entry(source_index,
 
2032
                                                 path_utf8=old_path)
 
2033
                    # update the source details variable to be the real
 
2034
                    # location.
 
2035
                    if old_entry == (None, None):
 
2036
                        raise errors.CorruptDirstate(state._filename,
 
2037
                            "entry '%s/%s' is considered renamed from %r"
 
2038
                            " but source does not exist\n"
 
2039
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
2040
                    source_details = old_entry[1][source_index]
 
2041
                    source_minikind = source_details[0]
 
2042
                else:
 
2043
                    old_dirname = entry[0][0]
 
2044
                    old_basename = entry[0][1]
 
2045
                    old_path = path = None
 
2046
                if path_info is None:
 
2047
                    # the file is missing on disk, show as removed.
 
2048
                    content_change = True
 
2049
                    target_kind = None
 
2050
                    target_exec = False
 
2051
                else:
 
2052
                    # source and target are both versioned and disk file is present.
 
2053
                    target_kind = path_info[2]
 
2054
                    if target_kind == 'directory':
 
2055
                        if path is None:
 
2056
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2057
                        new_dirname_to_file_id[path] = file_id
 
2058
                        if source_minikind != 'd':
 
2059
                            content_change = True
 
2060
                        else:
 
2061
                            # directories have no fingerprint
 
2062
                            content_change = False
 
2063
                        target_exec = False
 
2064
                    elif target_kind == 'file':
 
2065
                        if source_minikind != 'f':
 
2066
                            content_change = True
 
2067
                        else:
 
2068
                            # We could check the size, but we already have the
 
2069
                            # sha1 hash.
 
2070
                            content_change = (link_or_sha1 != source_details[1])
 
2071
                        # Target details is updated at update_entry time
 
2072
                        if use_filesystem_for_exec:
 
2073
                            # We don't need S_ISREG here, because we are sure
 
2074
                            # we are dealing with a file.
 
2075
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2076
                        else:
 
2077
                            target_exec = target_details[3]
 
2078
                    elif target_kind == 'symlink':
 
2079
                        if source_minikind != 'l':
 
2080
                            content_change = True
 
2081
                        else:
 
2082
                            content_change = (link_or_sha1 != source_details[1])
 
2083
                        target_exec = False
 
2084
                    elif target_kind == 'tree-reference':
 
2085
                        if source_minikind != 't':
 
2086
                            content_change = True
 
2087
                        else:
 
2088
                            content_change = False
 
2089
                        target_exec = False
 
2090
                    else:
 
2091
                        raise Exception, "unknown kind %s" % path_info[2]
 
2092
                if source_minikind == 'd':
 
2093
                    if path is None:
 
2094
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2095
                    old_dirname_to_file_id[old_path] = file_id
 
2096
                # parent id is the entry for the path in the target tree
 
2097
                if old_dirname == last_source_parent[0]:
 
2098
                    source_parent_id = last_source_parent[1]
 
2099
                else:
 
2100
                    try:
 
2101
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2102
                    except KeyError:
 
2103
                        source_parent_entry = state._get_entry(source_index,
 
2104
                                                               path_utf8=old_dirname)
 
2105
                        source_parent_id = source_parent_entry[0][2]
 
2106
                    if source_parent_id == entry[0][2]:
 
2107
                        # This is the root, so the parent is None
 
2108
                        source_parent_id = None
 
2109
                    else:
 
2110
                        last_source_parent[0] = old_dirname
 
2111
                        last_source_parent[1] = source_parent_id
 
2112
                new_dirname = entry[0][0]
 
2113
                if new_dirname == last_target_parent[0]:
 
2114
                    target_parent_id = last_target_parent[1]
 
2115
                else:
 
2116
                    try:
 
2117
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2118
                    except KeyError:
 
2119
                        # TODO: We don't always need to do the lookup, because the
 
2120
                        #       parent entry will be the same as the source entry.
 
2121
                        target_parent_entry = state._get_entry(target_index,
 
2122
                                                               path_utf8=new_dirname)
 
2123
                        if target_parent_entry == (None, None):
 
2124
                            raise AssertionError(
 
2125
                                "Could not find target parent in wt: %s\nparent of: %s"
 
2126
                                % (new_dirname, entry))
 
2127
                        target_parent_id = target_parent_entry[0][2]
 
2128
                    if target_parent_id == entry[0][2]:
 
2129
                        # This is the root, so the parent is None
 
2130
                        target_parent_id = None
 
2131
                    else:
 
2132
                        last_target_parent[0] = new_dirname
 
2133
                        last_target_parent[1] = target_parent_id
 
2134
 
 
2135
                source_exec = source_details[3]
 
2136
                if (include_unchanged
 
2137
                    or content_change
 
2138
                    or source_parent_id != target_parent_id
 
2139
                    or old_basename != entry[0][1]
 
2140
                    or source_exec != target_exec
 
2141
                    ):
 
2142
                    if old_path is None:
 
2143
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2144
                        old_path_u = utf8_decode(old_path)[0]
 
2145
                        path_u = old_path_u
 
2146
                    else:
 
2147
                        old_path_u = utf8_decode(old_path)[0]
 
2148
                        if old_path == path:
 
2149
                            path_u = old_path_u
 
2150
                        else:
 
2151
                            path_u = utf8_decode(path)[0]
 
2152
                    source_kind = _minikind_to_kind[source_minikind]
 
2153
                    return (entry[0][2],
 
2154
                           (old_path_u, path_u),
 
2155
                           content_change,
 
2156
                           (True, True),
 
2157
                           (source_parent_id, target_parent_id),
 
2158
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2159
                           (source_kind, target_kind),
 
2160
                           (source_exec, target_exec))
 
2161
                else:
 
2162
                    return uninteresting
 
2163
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2164
                # looks like a new file
 
2165
                if path_info is not None:
 
2166
                    path = pathjoin(entry[0][0], entry[0][1])
 
2167
                    # parent id is the entry for the path in the target tree
 
2168
                    # TODO: these are the same for an entire directory: cache em.
 
2169
                    parent_id = state._get_entry(target_index,
 
2170
                                                 path_utf8=entry[0][0])[0][2]
 
2171
                    if parent_id == entry[0][2]:
 
2172
                        parent_id = None
 
2173
                    if use_filesystem_for_exec:
 
2174
                        # We need S_ISREG here, because we aren't sure if this
 
2175
                        # is a file or not.
 
2176
                        target_exec = bool(
 
2177
                            stat.S_ISREG(path_info[3].st_mode)
 
2178
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2179
                    else:
 
2180
                        target_exec = target_details[3]
 
2181
                    return (entry[0][2],
 
2182
                           (None, utf8_decode(path)[0]),
 
2183
                           True,
 
2184
                           (False, True),
 
2185
                           (None, parent_id),
 
2186
                           (None, utf8_decode(entry[0][1])[0]),
 
2187
                           (None, path_info[2]),
 
2188
                           (None, target_exec))
 
2189
                else:
 
2190
                    # but its not on disk: we deliberately treat this as just
 
2191
                    # never-present. (Why ?! - RBC 20070224)
 
2192
                    pass
 
2193
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2194
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2195
                # if its still on disk, *and* theres no other entry at this
 
2196
                # path [we dont know this in this routine at the moment -
 
2197
                # perhaps we should change this - then it would be an unknown.
 
2198
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2199
                # parent id is the entry for the path in the target tree
 
2200
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2201
                if parent_id == entry[0][2]:
 
2202
                    parent_id = None
 
2203
                return (entry[0][2],
 
2204
                       (utf8_decode(old_path)[0], None),
 
2205
                       True,
 
2206
                       (True, False),
 
2207
                       (parent_id, None),
 
2208
                       (utf8_decode(entry[0][1])[0], None),
 
2209
                       (_minikind_to_kind[source_minikind], None),
 
2210
                       (source_details[3], None))
 
2211
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2212
                # a rename; could be a true rename, or a rename inherited from
 
2213
                # a renamed parent. TODO: handle this efficiently. Its not
 
2214
                # common case to rename dirs though, so a correct but slow
 
2215
                # implementation will do.
 
2216
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2217
                    search_specific_files.add(target_details[1])
 
2218
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2219
                # neither of the selected trees contain this file,
 
2220
                # so skip over it. This is not currently directly tested, but
 
2221
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2222
                pass
 
2223
            else:
 
2224
                raise AssertionError("don't know how to compare "
 
2225
                    "source_minikind=%r, target_minikind=%r"
 
2226
                    % (source_minikind, target_minikind))
 
2227
                ## import pdb;pdb.set_trace()
 
2228
            return None
 
2229
 
 
2230
        while search_specific_files:
 
2231
            # TODO: the pending list should be lexically sorted?  the
 
2232
            # interface doesn't require it.
 
2233
            current_root = search_specific_files.pop()
 
2234
            current_root_unicode = current_root.decode('utf8')
 
2235
            searched_specific_files.add(current_root)
 
2236
            # process the entries for this containing directory: the rest will be
 
2237
            # found by their parents recursively.
 
2238
            root_entries = _entries_for_path(current_root)
 
2239
            root_abspath = self.target.abspath(current_root_unicode)
 
2240
            try:
 
2241
                root_stat = os.lstat(root_abspath)
 
2242
            except OSError, e:
 
2243
                if e.errno == errno.ENOENT:
 
2244
                    # the path does not exist: let _process_entry know that.
 
2245
                    root_dir_info = None
 
2246
                else:
 
2247
                    # some other random error: hand it up.
 
2248
                    raise
 
2249
            else:
 
2250
                root_dir_info = ('', current_root,
 
2251
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2252
                    root_abspath)
 
2253
                if root_dir_info[2] == 'directory':
 
2254
                    if self.target._directory_is_tree_reference(
 
2255
                        current_root.decode('utf8')):
 
2256
                        root_dir_info = root_dir_info[:2] + \
 
2257
                            ('tree-reference',) + root_dir_info[3:]
 
2258
 
 
2259
            if not root_entries and not root_dir_info:
 
2260
                # this specified path is not present at all, skip it.
 
2261
                continue
 
2262
            path_handled = False
 
2263
            for entry in root_entries:
 
2264
                result = _process_entry(entry, root_dir_info)
 
2265
                if result is not None:
 
2266
                    path_handled = True
 
2267
                    if result is not uninteresting:
 
2268
                        yield result
 
2269
            if want_unversioned and not path_handled and root_dir_info:
 
2270
                new_executable = bool(
 
2271
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2272
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2273
                yield (None,
 
2274
                       (None, current_root_unicode),
 
2275
                       True,
 
2276
                       (False, False),
 
2277
                       (None, None),
 
2278
                       (None, splitpath(current_root_unicode)[-1]),
 
2279
                       (None, root_dir_info[2]),
 
2280
                       (None, new_executable)
 
2281
                      )
 
2282
            initial_key = (current_root, '', '')
 
2283
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2284
            if block_index == 0:
 
2285
                # we have processed the total root already, but because the
 
2286
                # initial key matched it we should skip it here.
 
2287
                block_index +=1
 
2288
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2289
                current_dir_info = None
 
2290
            else:
 
2291
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2292
                try:
 
2293
                    current_dir_info = dir_iterator.next()
 
2294
                except OSError, e:
 
2295
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2296
                    # python 2.5 has e.errno == EINVAL,
 
2297
                    #            and e.winerror == ERROR_DIRECTORY
 
2298
                    e_winerror = getattr(e, 'winerror', None)
 
2299
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2300
                    # there may be directories in the inventory even though
 
2301
                    # this path is not a file on disk: so mark it as end of
 
2302
                    # iterator
 
2303
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2304
                        current_dir_info = None
 
2305
                    elif (sys.platform == 'win32'
 
2306
                          and (e.errno in win_errors
 
2307
                               or e_winerror in win_errors)):
 
2308
                        current_dir_info = None
 
2309
                    else:
 
2310
                        raise
 
2311
                else:
 
2312
                    if current_dir_info[0][0] == '':
 
2313
                        # remove .bzr from iteration
 
2314
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2315
                        if current_dir_info[1][bzr_index][0] != '.bzr':
 
2316
                            raise AssertionError()
 
2317
                        del current_dir_info[1][bzr_index]
 
2318
            # walk until both the directory listing and the versioned metadata
 
2319
            # are exhausted. 
 
2320
            if (block_index < len(state._dirblocks) and
 
2321
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2322
                current_block = state._dirblocks[block_index]
 
2323
            else:
 
2324
                current_block = None
 
2325
            while (current_dir_info is not None or
 
2326
                   current_block is not None):
 
2327
                if (current_dir_info and current_block
 
2328
                    and current_dir_info[0][0] != current_block[0]):
 
2329
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2330
                        # filesystem data refers to paths not covered by the dirblock.
 
2331
                        # this has two possibilities:
 
2332
                        # A) it is versioned but empty, so there is no block for it
 
2333
                        # B) it is not versioned.
 
2334
 
 
2335
                        # if (A) then we need to recurse into it to check for
 
2336
                        # new unknown files or directories.
 
2337
                        # if (B) then we should ignore it, because we don't
 
2338
                        # recurse into unknown directories.
 
2339
                        path_index = 0
 
2340
                        while path_index < len(current_dir_info[1]):
 
2341
                                current_path_info = current_dir_info[1][path_index]
 
2342
                                if want_unversioned:
 
2343
                                    if current_path_info[2] == 'directory':
 
2344
                                        if self.target._directory_is_tree_reference(
 
2345
                                            current_path_info[0].decode('utf8')):
 
2346
                                            current_path_info = current_path_info[:2] + \
 
2347
                                                ('tree-reference',) + current_path_info[3:]
 
2348
                                    new_executable = bool(
 
2349
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2350
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2351
                                    yield (None,
 
2352
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2353
                                        True,
 
2354
                                        (False, False),
 
2355
                                        (None, None),
 
2356
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2357
                                        (None, current_path_info[2]),
 
2358
                                        (None, new_executable))
 
2359
                                # dont descend into this unversioned path if it is
 
2360
                                # a dir
 
2361
                                if current_path_info[2] in ('directory',
 
2362
                                                            'tree-reference'):
 
2363
                                    del current_dir_info[1][path_index]
 
2364
                                    path_index -= 1
 
2365
                                path_index += 1
 
2366
 
 
2367
                        # This dir info has been handled, go to the next
 
2368
                        try:
 
2369
                            current_dir_info = dir_iterator.next()
 
2370
                        except StopIteration:
 
2371
                            current_dir_info = None
 
2372
                    else:
 
2373
                        # We have a dirblock entry for this location, but there
 
2374
                        # is no filesystem path for this. This is most likely
 
2375
                        # because a directory was removed from the disk.
 
2376
                        # We don't have to report the missing directory,
 
2377
                        # because that should have already been handled, but we
 
2378
                        # need to handle all of the files that are contained
 
2379
                        # within.
 
2380
                        for current_entry in current_block[1]:
 
2381
                            # entry referring to file not present on disk.
 
2382
                            # advance the entry only, after processing.
 
2383
                            result = _process_entry(current_entry, None)
 
2384
                            if result is not None:
 
2385
                                if result is not uninteresting:
 
2386
                                    yield result
 
2387
                        block_index +=1
 
2388
                        if (block_index < len(state._dirblocks) and
 
2389
                            osutils.is_inside(current_root,
 
2390
                                              state._dirblocks[block_index][0])):
 
2391
                            current_block = state._dirblocks[block_index]
 
2392
                        else:
 
2393
                            current_block = None
 
2394
                    continue
 
2395
                entry_index = 0
 
2396
                if current_block and entry_index < len(current_block[1]):
 
2397
                    current_entry = current_block[1][entry_index]
 
2398
                else:
 
2399
                    current_entry = None
 
2400
                advance_entry = True
 
2401
                path_index = 0
 
2402
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2403
                    current_path_info = current_dir_info[1][path_index]
 
2404
                    if current_path_info[2] == 'directory':
 
2405
                        if self.target._directory_is_tree_reference(
 
2406
                            current_path_info[0].decode('utf8')):
 
2407
                            current_path_info = current_path_info[:2] + \
 
2408
                                ('tree-reference',) + current_path_info[3:]
 
2409
                else:
 
2410
                    current_path_info = None
 
2411
                advance_path = True
 
2412
                path_handled = False
 
2413
                while (current_entry is not None or
 
2414
                    current_path_info is not None):
 
2415
                    if current_entry is None:
 
2416
                        # the check for path_handled when the path is adnvaced
 
2417
                        # will yield this path if needed.
 
2418
                        pass
 
2419
                    elif current_path_info is None:
 
2420
                        # no path is fine: the per entry code will handle it.
 
2421
                        result = _process_entry(current_entry, current_path_info)
 
2422
                        if result is not None:
 
2423
                            if result is not uninteresting:
 
2424
                                yield result
 
2425
                    elif (current_entry[0][1] != current_path_info[1]
 
2426
                          or current_entry[1][target_index][0] in 'ar'):
 
2427
                        # The current path on disk doesn't match the dirblock
 
2428
                        # record. Either the dirblock is marked as absent, or
 
2429
                        # the file on disk is not present at all in the
 
2430
                        # dirblock. Either way, report about the dirblock
 
2431
                        # entry, and let other code handle the filesystem one.
 
2432
 
 
2433
                        # Compare the basename for these files to determine
 
2434
                        # which comes first
 
2435
                        if current_path_info[1] < current_entry[0][1]:
 
2436
                            # extra file on disk: pass for now, but only
 
2437
                            # increment the path, not the entry
 
2438
                            advance_entry = False
 
2439
                        else:
 
2440
                            # entry referring to file not present on disk.
 
2441
                            # advance the entry only, after processing.
 
2442
                            result = _process_entry(current_entry, None)
 
2443
                            if result is not None:
 
2444
                                if result is not uninteresting:
 
2445
                                    yield result
 
2446
                            advance_path = False
 
2447
                    else:
 
2448
                        result = _process_entry(current_entry, current_path_info)
 
2449
                        if result is not None:
 
2450
                            path_handled = True
 
2451
                            if result is not uninteresting:
 
2452
                                yield result
 
2453
                    if advance_entry and current_entry is not None:
 
2454
                        entry_index += 1
 
2455
                        if entry_index < len(current_block[1]):
 
2456
                            current_entry = current_block[1][entry_index]
 
2457
                        else:
 
2458
                            current_entry = None
 
2459
                    else:
 
2460
                        advance_entry = True # reset the advance flaga
 
2461
                    if advance_path and current_path_info is not None:
 
2462
                        if not path_handled:
 
2463
                            # unversioned in all regards
 
2464
                            if want_unversioned:
 
2465
                                new_executable = bool(
 
2466
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2467
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2468
                                yield (None,
 
2469
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2470
                                    True,
 
2471
                                    (False, False),
 
2472
                                    (None, None),
 
2473
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2474
                                    (None, current_path_info[2]),
 
2475
                                    (None, new_executable))
 
2476
                            # dont descend into this unversioned path if it is
 
2477
                            # a dir
 
2478
                            if current_path_info[2] in ('directory'):
 
2479
                                del current_dir_info[1][path_index]
 
2480
                                path_index -= 1
 
2481
                        # dont descend the disk iterator into any tree 
 
2482
                        # paths.
 
2483
                        if current_path_info[2] == 'tree-reference':
 
2484
                            del current_dir_info[1][path_index]
 
2485
                            path_index -= 1
 
2486
                        path_index += 1
 
2487
                        if path_index < len(current_dir_info[1]):
 
2488
                            current_path_info = current_dir_info[1][path_index]
 
2489
                            if current_path_info[2] == 'directory':
 
2490
                                if self.target._directory_is_tree_reference(
 
2491
                                    current_path_info[0].decode('utf8')):
 
2492
                                    current_path_info = current_path_info[:2] + \
 
2493
                                        ('tree-reference',) + current_path_info[3:]
 
2494
                        else:
 
2495
                            current_path_info = None
 
2496
                        path_handled = False
 
2497
                    else:
 
2498
                        advance_path = True # reset the advance flagg.
 
2499
                if current_block is not None:
 
2500
                    block_index += 1
 
2501
                    if (block_index < len(state._dirblocks) and
 
2502
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2503
                        current_block = state._dirblocks[block_index]
 
2504
                    else:
 
2505
                        current_block = None
 
2506
                if current_dir_info is not None:
 
2507
                    try:
 
2508
                        current_dir_info = dir_iterator.next()
 
2509
                    except StopIteration:
 
2510
                        current_dir_info = None
 
2511
 
 
2512
 
 
2513
    @staticmethod
 
2514
    def is_compatible(source, target):
 
2515
        # the target must be a dirstate working tree
 
2516
        if not isinstance(target, WorkingTree4):
 
2517
            return False
 
2518
        # the source must be a revtreee or dirstate rev tree.
 
2519
        if not isinstance(source,
 
2520
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2521
            return False
 
2522
        # the source revid must be in the target dirstate
 
2523
        if not (source._revision_id == NULL_REVISION or
 
2524
            source._revision_id in target.get_parent_ids()):
 
2525
            # TODO: what about ghosts? it may well need to 
 
2526
            # check for them explicitly.
 
2527
            return False
 
2528
        return True
 
2529
 
 
2530
InterTree.register_optimiser(InterDirStateTree)
 
2531
 
 
2532
 
 
2533
class Converter3to4(object):
 
2534
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2535
 
 
2536
    def __init__(self):
 
2537
        self.target_format = WorkingTreeFormat4()
 
2538
 
 
2539
    def convert(self, tree):
 
2540
        # lock the control files not the tree, so that we dont get tree
 
2541
        # on-unlock behaviours, and so that noone else diddles with the 
 
2542
        # tree during upgrade.
 
2543
        tree._control_files.lock_write()
 
2544
        try:
 
2545
            tree.read_working_inventory()
 
2546
            self.create_dirstate_data(tree)
 
2547
            self.update_format(tree)
 
2548
            self.remove_xml_files(tree)
 
2549
        finally:
 
2550
            tree._control_files.unlock()
 
2551
 
 
2552
    def create_dirstate_data(self, tree):
 
2553
        """Create the dirstate based data for tree."""
 
2554
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2555
            ).local_abspath('dirstate')
 
2556
        state = dirstate.DirState.from_tree(tree, local_path)
 
2557
        state.save()
 
2558
        state.unlock()
 
2559
 
 
2560
    def remove_xml_files(self, tree):
 
2561
        """Remove the oldformat 3 data."""
 
2562
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2563
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2564
            'pending-merges', 'stat-cache']:
 
2565
            try:
 
2566
                transport.delete(path)
 
2567
            except errors.NoSuchFile:
 
2568
                # some files are optional - just deal.
 
2569
                pass
 
2570
 
 
2571
    def update_format(self, tree):
 
2572
        """Change the format marker."""
 
2573
        tree._transport.put_bytes('format',
 
2574
            self.target_format.get_format_string(),
 
2575
            mode=tree.bzrdir._get_file_mode())