/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Aaron Bentley
  • Date: 2007-12-31 16:10:05 UTC
  • mto: This revision was merged to the branch mainline in revision 3241.
  • Revision ID: aaron.bentley@utoronto.ca-20071231161005-3l7s9521wm9zbjxg
Get apply_inventory_delta kinda working, but not children

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    debug,
 
47
    delta,
 
48
    dirstate,
 
49
    errors,
 
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    trace,
 
59
    transform,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
 
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
 
67
""")
 
68
 
 
69
from bzrlib import symbol_versioning
 
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
 
76
from bzrlib.osutils import (
 
77
    file_kind,
 
78
    isdir,
 
79
    normpath,
 
80
    pathjoin,
 
81
    rand_chars,
 
82
    realpath,
 
83
    safe_unicode,
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
 
87
from bzrlib.transport.local import LocalTransport
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
 
120
    def __init__(self, basedir,
 
121
                 branch,
 
122
                 _control_files=None,
 
123
                 _format=None,
 
124
                 _bzrdir=None):
 
125
        """Construct a WorkingTree for basedir.
 
126
 
 
127
        If the branch is not supplied, it is opened automatically.
 
128
        If the branch is supplied, it must be the branch for this basedir.
 
129
        (branch.base is not cross checked, because for remote branches that
 
130
        would be meaningless).
 
131
        """
 
132
        self._format = _format
 
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
 
136
        basedir = safe_unicode(basedir)
 
137
        mutter("opening working tree %r", basedir)
 
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
 
141
        self.basedir = realpath(basedir)
 
142
        # if branch is at our basedir and is a format 6 or less
 
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
 
146
        self._control_files = _control_files
 
147
        self._dirty = None
 
148
        #-------------
 
149
        # during a read or write lock these objects are set, and are
 
150
        # None the rest of the time.
 
151
        self._dirstate = None
 
152
        self._inventory = None
 
153
        #-------------
 
154
        self._setup_directory_is_tree_reference()
 
155
        self._detect_case_handling()
 
156
 
 
157
    @needs_tree_write_lock
 
158
    def _add(self, files, ids, kinds):
 
159
        """See MutableTree._add."""
 
160
        state = self.current_dirstate()
 
161
        for f, file_id, kind in zip(files, ids, kinds):
 
162
            f = f.strip('/')
 
163
            if self.path2id(f):
 
164
                # special case tree root handling.
 
165
                if f == '' and self.path2id(f) == ROOT_ID:
 
166
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
167
                continue
 
168
            if file_id is None:
 
169
                file_id = generate_ids.gen_file_id(f)
 
170
            # deliberately add the file with no cached stat or sha1
 
171
            # - on the first access it will be gathered, and we can
 
172
            # always change this once tests are all passing.
 
173
            state.add(f, file_id, kind, None, '')
 
174
        self._make_dirty(reset_inventory=True)
 
175
 
 
176
    def _make_dirty(self, reset_inventory):
 
177
        """Make the tree state dirty.
 
178
 
 
179
        :param reset_inventory: True if the cached inventory should be removed
 
180
            (presuming there is one).
 
181
        """
 
182
        self._dirty = True
 
183
        if reset_inventory and self._inventory is not None:
 
184
            self._inventory = None
 
185
 
 
186
    @needs_tree_write_lock
 
187
    def add_reference(self, sub_tree):
 
188
        # use standard implementation, which calls back to self._add
 
189
        # 
 
190
        # So we don't store the reference_revision in the working dirstate,
 
191
        # it's just recorded at the moment of commit. 
 
192
        self._add_reference(sub_tree)
 
193
 
 
194
    def break_lock(self):
 
195
        """Break a lock if one is present from another instance.
 
196
 
 
197
        Uses the ui factory to ask for confirmation if the lock may be from
 
198
        an active process.
 
199
 
 
200
        This will probe the repository for its lock as well.
 
201
        """
 
202
        # if the dirstate is locked by an active process, reject the break lock
 
203
        # call.
 
204
        try:
 
205
            if self._dirstate is None:
 
206
                clear = True
 
207
            else:
 
208
                clear = False
 
209
            state = self._current_dirstate()
 
210
            if state._lock_token is not None:
 
211
                # we already have it locked. sheese, cant break our own lock.
 
212
                raise errors.LockActive(self.basedir)
 
213
            else:
 
214
                try:
 
215
                    # try for a write lock - need permission to get one anyhow
 
216
                    # to break locks.
 
217
                    state.lock_write()
 
218
                except errors.LockContention:
 
219
                    # oslocks fail when a process is still live: fail.
 
220
                    # TODO: get the locked lockdir info and give to the user to
 
221
                    # assist in debugging.
 
222
                    raise errors.LockActive(self.basedir)
 
223
                else:
 
224
                    state.unlock()
 
225
        finally:
 
226
            if clear:
 
227
                self._dirstate = None
 
228
        self._control_files.break_lock()
 
229
        self.branch.break_lock()
 
230
 
 
231
    def _comparison_data(self, entry, path):
 
232
        kind, executable, stat_value = \
 
233
            WorkingTree3._comparison_data(self, entry, path)
 
234
        # it looks like a plain directory, but it's really a reference -- see
 
235
        # also kind()
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
 
239
            kind = 'tree-reference'
 
240
        return kind, executable, stat_value
 
241
 
 
242
    @needs_write_lock
 
243
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
244
        # mark the tree as dirty post commit - commit
 
245
        # can change the current versioned list by doing deletes.
 
246
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
247
        self._make_dirty(reset_inventory=True)
 
248
        return result
 
249
 
 
250
    def current_dirstate(self):
 
251
        """Return the current dirstate object.
 
252
 
 
253
        This is not part of the tree interface and only exposed for ease of
 
254
        testing.
 
255
 
 
256
        :raises errors.NotWriteLocked: when not in a lock.
 
257
        """
 
258
        self._must_be_locked()
 
259
        return self._current_dirstate()
 
260
 
 
261
    def _current_dirstate(self):
 
262
        """Internal function that does not check lock status.
 
263
 
 
264
        This is needed for break_lock which also needs the dirstate.
 
265
        """
 
266
        if self._dirstate is not None:
 
267
            return self._dirstate
 
268
        local_path = self.bzrdir.get_workingtree_transport(None
 
269
            ).local_abspath('dirstate')
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
 
271
        return self._dirstate
 
272
 
 
273
    def filter_unversioned_files(self, paths):
 
274
        """Filter out paths that are versioned.
 
275
 
 
276
        :return: set of paths.
 
277
        """
 
278
        # TODO: make a generic multi-bisect routine roughly that should list
 
279
        # the paths, then process one half at a time recursively, and feed the
 
280
        # results of each bisect in further still
 
281
        paths = sorted(paths)
 
282
        result = set()
 
283
        state = self.current_dirstate()
 
284
        # TODO we want a paths_to_dirblocks helper I think
 
285
        for path in paths:
 
286
            dirname, basename = os.path.split(path.encode('utf8'))
 
287
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
288
                dirname, basename, 0)
 
289
            if not path_is_versioned:
 
290
                result.add(path)
 
291
        return result
 
292
 
 
293
    def flush(self):
 
294
        """Write all cached data to disk."""
 
295
        if self._control_files._lock_mode != 'w':
 
296
            raise errors.NotWriteLocked(self)
 
297
        self.current_dirstate().save()
 
298
        self._inventory = None
 
299
        self._dirty = False
 
300
 
 
301
    @needs_tree_write_lock
 
302
    def _gather_kinds(self, files, kinds):
 
303
        """See MutableTree._gather_kinds."""
 
304
        for pos, f in enumerate(files):
 
305
            if kinds[pos] is None:
 
306
                kinds[pos] = self._kind(f)
 
307
 
 
308
    def _generate_inventory(self):
 
309
        """Create and set self.inventory from the dirstate object.
 
310
        
 
311
        This is relatively expensive: we have to walk the entire dirstate.
 
312
        Ideally we would not, and can deprecate this function.
 
313
        """
 
314
        #: uncomment to trap on inventory requests.
 
315
        # import pdb;pdb.set_trace()
 
316
        state = self.current_dirstate()
 
317
        state._read_dirblocks_if_needed()
 
318
        root_key, current_entry = self._get_entry(path='')
 
319
        current_id = root_key[2]
 
320
        assert current_entry[0][0] == 'd' # directory
 
321
        inv = Inventory(root_id=current_id)
 
322
        # Turn some things into local variables
 
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
324
        factory = entry_factory
 
325
        utf8_decode = cache_utf8._utf8_decode
 
326
        inv_byid = inv._byid
 
327
        # we could do this straight out of the dirstate; it might be fast
 
328
        # and should be profiled - RBC 20070216
 
329
        parent_ies = {'' : inv.root}
 
330
        for block in state._dirblocks[1:]: # skip the root
 
331
            dirname = block[0]
 
332
            try:
 
333
                parent_ie = parent_ies[dirname]
 
334
            except KeyError:
 
335
                # all the paths in this block are not versioned in this tree
 
336
                continue
 
337
            for key, entry in block[1]:
 
338
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
339
                if minikind in ('a', 'r'): # absent, relocated
 
340
                    # a parent tree only entry
 
341
                    continue
 
342
                name = key[1]
 
343
                name_unicode = utf8_decode(name)[0]
 
344
                file_id = key[2]
 
345
                kind = minikind_to_kind[minikind]
 
346
                inv_entry = factory[kind](file_id, name_unicode,
 
347
                                          parent_ie.file_id)
 
348
                if kind == 'file':
 
349
                    # This is only needed on win32, where this is the only way
 
350
                    # we know the executable bit.
 
351
                    inv_entry.executable = executable
 
352
                    # not strictly needed: working tree
 
353
                    #inv_entry.text_size = size
 
354
                    #inv_entry.text_sha1 = sha1
 
355
                elif kind == 'directory':
 
356
                    # add this entry to the parent map.
 
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
358
                elif kind == 'tree-reference':
 
359
                    assert self._repo_supports_tree_reference, \
 
360
                        "repository of %r " \
 
361
                        "doesn't support tree references " \
 
362
                        "required by entry %r" \
 
363
                        % (self, name)
 
364
                    inv_entry.reference_revision = link_or_sha1 or None
 
365
                elif kind != 'symlink':
 
366
                    raise AssertionError("unknown kind %r" % kind)
 
367
                # These checks cost us around 40ms on a 55k entry tree
 
368
                assert file_id not in inv_byid, ('file_id %s already in'
 
369
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
370
                assert name_unicode not in parent_ie.children
 
371
                inv_byid[file_id] = inv_entry
 
372
                parent_ie.children[name_unicode] = inv_entry
 
373
        self._inventory = inv
 
374
 
 
375
    def _get_entry(self, file_id=None, path=None):
 
376
        """Get the dirstate row for file_id or path.
 
377
 
 
378
        If either file_id or path is supplied, it is used as the key to lookup.
 
379
        If both are supplied, the fastest lookup is used, and an error is
 
380
        raised if they do not both point at the same row.
 
381
        
 
382
        :param file_id: An optional unicode file_id to be looked up.
 
383
        :param path: An optional unicode path to be looked up.
 
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
385
        """
 
386
        if file_id is None and path is None:
 
387
            raise errors.BzrError('must supply file_id or path')
 
388
        state = self.current_dirstate()
 
389
        if path is not None:
 
390
            path = path.encode('utf8')
 
391
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
392
 
 
393
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
394
        # check file id is valid unconditionally.
 
395
        entry = self._get_entry(file_id=file_id, path=path)
 
396
        if entry[0] is None:
 
397
            raise errors.NoSuchId(self, file_id)
 
398
        if path is None:
 
399
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
400
 
 
401
        file_abspath = self.abspath(path)
 
402
        state = self.current_dirstate()
 
403
        if stat_value is None:
 
404
            try:
 
405
                stat_value = os.lstat(file_abspath)
 
406
            except OSError, e:
 
407
                if e.errno == errno.ENOENT:
 
408
                    return None
 
409
                else:
 
410
                    raise
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
 
413
        if entry[1][0][0] == 'f':
 
414
            return link_or_sha1
 
415
        return None
 
416
 
 
417
    def _get_inventory(self):
 
418
        """Get the inventory for the tree. This is only valid within a lock."""
 
419
        if 'evil' in debug.debug_flags:
 
420
            trace.mutter_callsite(2,
 
421
                "accessing .inventory forces a size of tree translation.")
 
422
        if self._inventory is not None:
 
423
            return self._inventory
 
424
        self._must_be_locked()
 
425
        self._generate_inventory()
 
426
        return self._inventory
 
427
 
 
428
    inventory = property(_get_inventory,
 
429
                         doc="Inventory of this Tree")
 
430
 
 
431
    @needs_read_lock
 
432
    def get_parent_ids(self):
 
433
        """See Tree.get_parent_ids.
 
434
        
 
435
        This implementation requests the ids list from the dirstate file.
 
436
        """
 
437
        return self.current_dirstate().get_parent_ids()
 
438
 
 
439
    def get_reference_revision(self, file_id, path=None):
 
440
        # referenced tree's revision is whatever's currently there
 
441
        return self.get_nested_tree(file_id, path).last_revision()
 
442
 
 
443
    def get_nested_tree(self, file_id, path=None):
 
444
        if path is None:
 
445
            path = self.id2path(file_id)
 
446
        # else: check file_id is at path?
 
447
        return WorkingTree.open(self.abspath(path))
 
448
 
 
449
    @needs_read_lock
 
450
    def get_root_id(self):
 
451
        """Return the id of this trees root"""
 
452
        return self._get_entry(path='')[0][2]
 
453
 
 
454
    def has_id(self, file_id):
 
455
        state = self.current_dirstate()
 
456
        row, parents = self._get_entry(file_id=file_id)
 
457
        if row is None:
 
458
            return False
 
459
        return osutils.lexists(pathjoin(
 
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
461
 
 
462
    @needs_read_lock
 
463
    def id2path(self, file_id):
 
464
        "Convert a file-id to a path."
 
465
        state = self.current_dirstate()
 
466
        entry = self._get_entry(file_id=file_id)
 
467
        if entry == (None, None):
 
468
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
469
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
470
        return path_utf8.decode('utf8')
 
471
 
 
472
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
473
        entry = self._get_entry(path=path)
 
474
        if entry == (None, None):
 
475
            return False # Missing entries are not executable
 
476
        return entry[1][0][3] # Executable?
 
477
 
 
478
    if not osutils.supports_executable():
 
479
        def is_executable(self, file_id, path=None):
 
480
            """Test if a file is executable or not.
 
481
 
 
482
            Note: The caller is expected to take a read-lock before calling this.
 
483
            """
 
484
            entry = self._get_entry(file_id=file_id, path=path)
 
485
            if entry == (None, None):
 
486
                return False
 
487
            return entry[1][0][3]
 
488
 
 
489
        _is_executable_from_path_and_stat = \
 
490
            _is_executable_from_path_and_stat_from_basis
 
491
    else:
 
492
        def is_executable(self, file_id, path=None):
 
493
            """Test if a file is executable or not.
 
494
 
 
495
            Note: The caller is expected to take a read-lock before calling this.
 
496
            """
 
497
            if not path:
 
498
                path = self.id2path(file_id)
 
499
            mode = os.lstat(self.abspath(path)).st_mode
 
500
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
501
 
 
502
    def iter_all_file_ids(self):
 
503
        """See Tree.iter_all_file_ids"""
 
504
        self._must_be_locked()
 
505
        for key, tree_details in self.current_dirstate()._iter_entries():
 
506
            if tree_details[0][0] in ('a', 'r'): # relocated
 
507
                continue
 
508
            yield key[2]
 
509
 
 
510
    @needs_read_lock
 
511
    def __iter__(self):
 
512
        """Iterate through file_ids for this tree.
 
513
 
 
514
        file_ids are in a WorkingTree if they are in the working inventory
 
515
        and the working file exists.
 
516
        """
 
517
        result = []
 
518
        for key, tree_details in self.current_dirstate()._iter_entries():
 
519
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
520
                # not relevant to the working tree
 
521
                continue
 
522
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
523
            if osutils.lexists(path):
 
524
                result.append(key[2])
 
525
        return iter(result)
 
526
 
 
527
    def iter_references(self):
 
528
        for key, tree_details in self.current_dirstate()._iter_entries():
 
529
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
530
                # not relevant to the working tree
 
531
                continue
 
532
            if not key[1]:
 
533
                # the root is not a reference.
 
534
                continue
 
535
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
536
            try:
 
537
                if self._kind(path) == 'tree-reference':
 
538
                    yield path, key[2]
 
539
            except errors.NoSuchFile:
 
540
                # path is missing on disk.
 
541
                continue
 
542
 
 
543
    def kind(self, file_id):
 
544
        """Return the kind of a file.
 
545
 
 
546
        This is always the actual kind that's on disk, regardless of what it
 
547
        was added as.
 
548
 
 
549
        Note: The caller is expected to take a read-lock before calling this.
 
550
        """
 
551
        relpath = self.id2path(file_id)
 
552
        assert relpath != None, \
 
553
            "path for id {%s} is None!" % file_id
 
554
        return self._kind(relpath)
 
555
 
 
556
    def _kind(self, relpath):
 
557
        abspath = self.abspath(relpath)
 
558
        kind = file_kind(abspath)
 
559
        if (self._repo_supports_tree_reference and
 
560
            kind == 'directory' and
 
561
            self._directory_is_tree_reference(relpath)):
 
562
            kind = 'tree-reference'
 
563
        return kind
 
564
 
 
565
    @needs_read_lock
 
566
    def _last_revision(self):
 
567
        """See Mutable.last_revision."""
 
568
        parent_ids = self.current_dirstate().get_parent_ids()
 
569
        if parent_ids:
 
570
            return parent_ids[0]
 
571
        else:
 
572
            return _mod_revision.NULL_REVISION
 
573
 
 
574
    def lock_read(self):
 
575
        """See Branch.lock_read, and WorkingTree.unlock."""
 
576
        self.branch.lock_read()
 
577
        try:
 
578
            self._control_files.lock_read()
 
579
            try:
 
580
                state = self.current_dirstate()
 
581
                if not state._lock_token:
 
582
                    state.lock_read()
 
583
                # set our support for tree references from the repository in
 
584
                # use.
 
585
                self._repo_supports_tree_reference = getattr(
 
586
                    self.branch.repository._format, "supports_tree_reference",
 
587
                    False)
 
588
            except:
 
589
                self._control_files.unlock()
 
590
                raise
 
591
        except:
 
592
            self.branch.unlock()
 
593
            raise
 
594
 
 
595
    def _lock_self_write(self):
 
596
        """This should be called after the branch is locked."""
 
597
        try:
 
598
            self._control_files.lock_write()
 
599
            try:
 
600
                state = self.current_dirstate()
 
601
                if not state._lock_token:
 
602
                    state.lock_write()
 
603
                # set our support for tree references from the repository in
 
604
                # use.
 
605
                self._repo_supports_tree_reference = getattr(
 
606
                    self.branch.repository._format, "supports_tree_reference",
 
607
                    False)
 
608
            except:
 
609
                self._control_files.unlock()
 
610
                raise
 
611
        except:
 
612
            self.branch.unlock()
 
613
            raise
 
614
 
 
615
    def lock_tree_write(self):
 
616
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
617
        self.branch.lock_read()
 
618
        self._lock_self_write()
 
619
 
 
620
    def lock_write(self):
 
621
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
622
        self.branch.lock_write()
 
623
        self._lock_self_write()
 
624
 
 
625
    @needs_tree_write_lock
 
626
    def move(self, from_paths, to_dir, after=False):
 
627
        """See WorkingTree.move()."""
 
628
        result = []
 
629
        if not from_paths:
 
630
            return result
 
631
 
 
632
        state = self.current_dirstate()
 
633
 
 
634
        assert not isinstance(from_paths, basestring)
 
635
        to_dir_utf8 = to_dir.encode('utf8')
 
636
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
637
        id_index = state._get_id_index()
 
638
        # check destination directory
 
639
        # get the details for it
 
640
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
641
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
642
        if not entry_present:
 
643
            raise errors.BzrMoveFailedError('', to_dir,
 
644
                errors.NotVersionedError(to_dir))
 
645
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
646
        # get a handle on the block itself.
 
647
        to_block_index = state._ensure_block(
 
648
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
649
        to_block = state._dirblocks[to_block_index]
 
650
        to_abs = self.abspath(to_dir)
 
651
        if not isdir(to_abs):
 
652
            raise errors.BzrMoveFailedError('',to_dir,
 
653
                errors.NotADirectory(to_abs))
 
654
 
 
655
        if to_entry[1][0][0] != 'd':
 
656
            raise errors.BzrMoveFailedError('',to_dir,
 
657
                errors.NotADirectory(to_abs))
 
658
 
 
659
        if self._inventory is not None:
 
660
            update_inventory = True
 
661
            inv = self.inventory
 
662
            to_dir_ie = inv[to_dir_id]
 
663
            to_dir_id = to_entry[0][2]
 
664
        else:
 
665
            update_inventory = False
 
666
 
 
667
        rollbacks = []
 
668
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
669
                     fingerprint, packed_stat, size,
 
670
                     to_block, to_key, to_path_utf8):
 
671
            state._make_absent(old_entry)
 
672
            from_key = old_entry[0]
 
673
            rollbacks.append(
 
674
                lambda:state.update_minimal(from_key,
 
675
                    minikind,
 
676
                    executable=executable,
 
677
                    fingerprint=fingerprint,
 
678
                    packed_stat=packed_stat,
 
679
                    size=size,
 
680
                    path_utf8=from_path_utf8))
 
681
            state.update_minimal(to_key,
 
682
                    minikind,
 
683
                    executable=executable,
 
684
                    fingerprint=fingerprint,
 
685
                    packed_stat=packed_stat,
 
686
                    size=size,
 
687
                    path_utf8=to_path_utf8)
 
688
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
689
            new_entry = to_block[1][added_entry_index]
 
690
            rollbacks.append(lambda:state._make_absent(new_entry))
 
691
 
 
692
        for from_rel in from_paths:
 
693
            # from_rel is 'pathinroot/foo/bar'
 
694
            from_rel_utf8 = from_rel.encode('utf8')
 
695
            from_dirname, from_tail = osutils.split(from_rel)
 
696
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
697
            from_entry = self._get_entry(path=from_rel)
 
698
            if from_entry == (None, None):
 
699
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
700
                    errors.NotVersionedError(path=str(from_rel)))
 
701
 
 
702
            from_id = from_entry[0][2]
 
703
            to_rel = pathjoin(to_dir, from_tail)
 
704
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
705
            item_to_entry = self._get_entry(path=to_rel)
 
706
            if item_to_entry != (None, None):
 
707
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
708
                    "Target is already versioned.")
 
709
 
 
710
            if from_rel == to_rel:
 
711
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
712
                    "Source and target are identical.")
 
713
 
 
714
            from_missing = not self.has_filename(from_rel)
 
715
            to_missing = not self.has_filename(to_rel)
 
716
            if after:
 
717
                move_file = False
 
718
            else:
 
719
                move_file = True
 
720
            if to_missing:
 
721
                if not move_file:
 
722
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
723
                        errors.NoSuchFile(path=to_rel,
 
724
                        extra="New file has not been created yet"))
 
725
                elif from_missing:
 
726
                    # neither path exists
 
727
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
728
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
729
            else:
 
730
                if from_missing: # implicitly just update our path mapping
 
731
                    move_file = False
 
732
                elif not after:
 
733
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
734
 
 
735
            rollbacks = []
 
736
            def rollback_rename():
 
737
                """A single rename has failed, roll it back."""
 
738
                # roll back everything, even if we encounter trouble doing one
 
739
                # of them.
 
740
                #
 
741
                # TODO: at least log the other exceptions rather than just
 
742
                # losing them mbp 20070307
 
743
                exc_info = None
 
744
                for rollback in reversed(rollbacks):
 
745
                    try:
 
746
                        rollback()
 
747
                    except Exception, e:
 
748
                        exc_info = sys.exc_info()
 
749
                if exc_info:
 
750
                    raise exc_info[0], exc_info[1], exc_info[2]
 
751
 
 
752
            # perform the disk move first - its the most likely failure point.
 
753
            if move_file:
 
754
                from_rel_abs = self.abspath(from_rel)
 
755
                to_rel_abs = self.abspath(to_rel)
 
756
                try:
 
757
                    osutils.rename(from_rel_abs, to_rel_abs)
 
758
                except OSError, e:
 
759
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
760
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
761
            try:
 
762
                # perform the rename in the inventory next if needed: its easy
 
763
                # to rollback
 
764
                if update_inventory:
 
765
                    # rename the entry
 
766
                    from_entry = inv[from_id]
 
767
                    current_parent = from_entry.parent_id
 
768
                    inv.rename(from_id, to_dir_id, from_tail)
 
769
                    rollbacks.append(
 
770
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
771
                # finally do the rename in the dirstate, which is a little
 
772
                # tricky to rollback, but least likely to need it.
 
773
                old_block_index, old_entry_index, dir_present, file_present = \
 
774
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
775
                old_block = state._dirblocks[old_block_index][1]
 
776
                old_entry = old_block[old_entry_index]
 
777
                from_key, old_entry_details = old_entry
 
778
                cur_details = old_entry_details[0]
 
779
                # remove the old row
 
780
                to_key = ((to_block[0],) + from_key[1:3])
 
781
                minikind = cur_details[0]
 
782
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
783
                         minikind=minikind,
 
784
                         executable=cur_details[3],
 
785
                         fingerprint=cur_details[1],
 
786
                         packed_stat=cur_details[4],
 
787
                         size=cur_details[2],
 
788
                         to_block=to_block,
 
789
                         to_key=to_key,
 
790
                         to_path_utf8=to_rel_utf8)
 
791
 
 
792
                if minikind == 'd':
 
793
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
794
                        """Recursively update all entries in this dirblock."""
 
795
                        assert from_dir != '', "renaming root not supported"
 
796
                        from_key = (from_dir, '')
 
797
                        from_block_idx, present = \
 
798
                            state._find_block_index_from_key(from_key)
 
799
                        if not present:
 
800
                            # This is the old record, if it isn't present, then
 
801
                            # there is theoretically nothing to update.
 
802
                            # (Unless it isn't present because of lazy loading,
 
803
                            # but we don't do that yet)
 
804
                            return
 
805
                        from_block = state._dirblocks[from_block_idx]
 
806
                        to_block_index, to_entry_index, _, _ = \
 
807
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
808
                        to_block_index = state._ensure_block(
 
809
                            to_block_index, to_entry_index, to_dir_utf8)
 
810
                        to_block = state._dirblocks[to_block_index]
 
811
 
 
812
                        # Grab a copy since move_one may update the list.
 
813
                        for entry in from_block[1][:]:
 
814
                            assert entry[0][0] == from_dir
 
815
                            cur_details = entry[1][0]
 
816
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
817
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
818
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
819
                            minikind = cur_details[0]
 
820
                            if minikind in 'ar':
 
821
                                # Deleted children of a renamed directory
 
822
                                # Do not need to be updated.
 
823
                                # Children that have been renamed out of this
 
824
                                # directory should also not be updated
 
825
                                continue
 
826
                            move_one(entry, from_path_utf8=from_path_utf8,
 
827
                                     minikind=minikind,
 
828
                                     executable=cur_details[3],
 
829
                                     fingerprint=cur_details[1],
 
830
                                     packed_stat=cur_details[4],
 
831
                                     size=cur_details[2],
 
832
                                     to_block=to_block,
 
833
                                     to_key=to_key,
 
834
                                     to_path_utf8=to_path_utf8)
 
835
                            if minikind == 'd':
 
836
                                # We need to move all the children of this
 
837
                                # entry
 
838
                                update_dirblock(from_path_utf8, to_key,
 
839
                                                to_path_utf8)
 
840
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
841
            except:
 
842
                rollback_rename()
 
843
                raise
 
844
            result.append((from_rel, to_rel))
 
845
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
846
            self._make_dirty(reset_inventory=False)
 
847
 
 
848
        return result
 
849
 
 
850
    def _must_be_locked(self):
 
851
        if not self._control_files._lock_count:
 
852
            raise errors.ObjectNotLocked(self)
 
853
 
 
854
    def _new_tree(self):
 
855
        """Initialize the state in this tree to be a new tree."""
 
856
        self._dirty = True
 
857
 
 
858
    @needs_read_lock
 
859
    def path2id(self, path):
 
860
        """Return the id for path in this tree."""
 
861
        path = path.strip('/')
 
862
        entry = self._get_entry(path=path)
 
863
        if entry == (None, None):
 
864
            return None
 
865
        return entry[0][2]
 
866
 
 
867
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
868
        """See Tree.paths2ids().
 
869
 
 
870
        This specialisation fast-paths the case where all the trees are in the
 
871
        dirstate.
 
872
        """
 
873
        if paths is None:
 
874
            return None
 
875
        parents = self.get_parent_ids()
 
876
        for tree in trees:
 
877
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
878
                parents):
 
879
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
880
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
881
        # -- make all paths utf8 --
 
882
        paths_utf8 = set()
 
883
        for path in paths:
 
884
            paths_utf8.add(path.encode('utf8'))
 
885
        paths = paths_utf8
 
886
        # -- paths is now a utf8 path set --
 
887
        # -- get the state object and prepare it.
 
888
        state = self.current_dirstate()
 
889
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
890
            and '' not in paths):
 
891
            paths2ids = self._paths2ids_using_bisect
 
892
        else:
 
893
            paths2ids = self._paths2ids_in_memory
 
894
        return paths2ids(paths, search_indexes,
 
895
                         require_versioned=require_versioned)
 
896
 
 
897
    def _paths2ids_in_memory(self, paths, search_indexes,
 
898
                             require_versioned=True):
 
899
        state = self.current_dirstate()
 
900
        state._read_dirblocks_if_needed()
 
901
        def _entries_for_path(path):
 
902
            """Return a list with all the entries that match path for all ids.
 
903
            """
 
904
            dirname, basename = os.path.split(path)
 
905
            key = (dirname, basename, '')
 
906
            block_index, present = state._find_block_index_from_key(key)
 
907
            if not present:
 
908
                # the block which should contain path is absent.
 
909
                return []
 
910
            result = []
 
911
            block = state._dirblocks[block_index][1]
 
912
            entry_index, _ = state._find_entry_index(key, block)
 
913
            # we may need to look at multiple entries at this path: walk while the paths match.
 
914
            while (entry_index < len(block) and
 
915
                block[entry_index][0][0:2] == key[0:2]):
 
916
                result.append(block[entry_index])
 
917
                entry_index += 1
 
918
            return result
 
919
        if require_versioned:
 
920
            # -- check all supplied paths are versioned in a search tree. --
 
921
            all_versioned = True
 
922
            for path in paths:
 
923
                path_entries = _entries_for_path(path)
 
924
                if not path_entries:
 
925
                    # this specified path is not present at all: error
 
926
                    all_versioned = False
 
927
                    break
 
928
                found_versioned = False
 
929
                # for each id at this path
 
930
                for entry in path_entries:
 
931
                    # for each tree.
 
932
                    for index in search_indexes:
 
933
                        if entry[1][index][0] != 'a': # absent
 
934
                            found_versioned = True
 
935
                            # all good: found a versioned cell
 
936
                            break
 
937
                if not found_versioned:
 
938
                    # none of the indexes was not 'absent' at all ids for this
 
939
                    # path.
 
940
                    all_versioned = False
 
941
                    break
 
942
            if not all_versioned:
 
943
                raise errors.PathsNotVersionedError(paths)
 
944
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
945
        search_paths = osutils.minimum_path_selection(paths)
 
946
        # sketch: 
 
947
        # for all search_indexs in each path at or under each element of
 
948
        # search_paths, if the detail is relocated: add the id, and add the
 
949
        # relocated path as one to search if its not searched already. If the
 
950
        # detail is not relocated, add the id.
 
951
        searched_paths = set()
 
952
        found_ids = set()
 
953
        def _process_entry(entry):
 
954
            """Look at search_indexes within entry.
 
955
 
 
956
            If a specific tree's details are relocated, add the relocation
 
957
            target to search_paths if not searched already. If it is absent, do
 
958
            nothing. Otherwise add the id to found_ids.
 
959
            """
 
960
            for index in search_indexes:
 
961
                if entry[1][index][0] == 'r': # relocated
 
962
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
963
                        search_paths.add(entry[1][index][1])
 
964
                elif entry[1][index][0] != 'a': # absent
 
965
                    found_ids.add(entry[0][2])
 
966
        while search_paths:
 
967
            current_root = search_paths.pop()
 
968
            searched_paths.add(current_root)
 
969
            # process the entries for this containing directory: the rest will be
 
970
            # found by their parents recursively.
 
971
            root_entries = _entries_for_path(current_root)
 
972
            if not root_entries:
 
973
                # this specified path is not present at all, skip it.
 
974
                continue
 
975
            for entry in root_entries:
 
976
                _process_entry(entry)
 
977
            initial_key = (current_root, '', '')
 
978
            block_index, _ = state._find_block_index_from_key(initial_key)
 
979
            while (block_index < len(state._dirblocks) and
 
980
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
981
                for entry in state._dirblocks[block_index][1]:
 
982
                    _process_entry(entry)
 
983
                block_index += 1
 
984
        return found_ids
 
985
 
 
986
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
987
                                require_versioned=True):
 
988
        state = self.current_dirstate()
 
989
        found_ids = set()
 
990
 
 
991
        split_paths = sorted(osutils.split(p) for p in paths)
 
992
        found = state._bisect_recursive(split_paths)
 
993
 
 
994
        if require_versioned:
 
995
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
996
            for dir_name in split_paths:
 
997
                if dir_name not in found_dir_names:
 
998
                    raise errors.PathsNotVersionedError(paths)
 
999
 
 
1000
        for dir_name_id, trees_info in found.iteritems():
 
1001
            for index in search_indexes:
 
1002
                if trees_info[index][0] not in ('r', 'a'):
 
1003
                    found_ids.add(dir_name_id[2])
 
1004
        return found_ids
 
1005
 
 
1006
    def read_working_inventory(self):
 
1007
        """Read the working inventory.
 
1008
        
 
1009
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1010
        """
 
1011
        return self.inventory
 
1012
 
 
1013
    @needs_read_lock
 
1014
    def revision_tree(self, revision_id):
 
1015
        """See Tree.revision_tree.
 
1016
 
 
1017
        WorkingTree4 supplies revision_trees for any basis tree.
 
1018
        """
 
1019
        dirstate = self.current_dirstate()
 
1020
        parent_ids = dirstate.get_parent_ids()
 
1021
        if revision_id not in parent_ids:
 
1022
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1023
        if revision_id in dirstate.get_ghosts():
 
1024
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1025
        return DirStateRevisionTree(dirstate, revision_id,
 
1026
            self.branch.repository)
 
1027
 
 
1028
    @needs_tree_write_lock
 
1029
    def set_last_revision(self, new_revision):
 
1030
        """Change the last revision in the working tree."""
 
1031
        parents = self.get_parent_ids()
 
1032
        if new_revision in (NULL_REVISION, None):
 
1033
            assert len(parents) < 2, (
 
1034
                "setting the last parent to none with a pending merge is "
 
1035
                "unsupported.")
 
1036
            self.set_parent_ids([])
 
1037
        else:
 
1038
            self.set_parent_ids([new_revision] + parents[1:],
 
1039
                allow_leftmost_as_ghost=True)
 
1040
 
 
1041
    @needs_tree_write_lock
 
1042
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1043
        """Set the parent ids to revision_ids.
 
1044
        
 
1045
        See also set_parent_trees. This api will try to retrieve the tree data
 
1046
        for each element of revision_ids from the trees repository. If you have
 
1047
        tree data already available, it is more efficient to use
 
1048
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1049
        an easier API to use.
 
1050
 
 
1051
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1052
            working tree. Any of these may be ghosts.
 
1053
        """
 
1054
        trees = []
 
1055
        for revision_id in revision_ids:
 
1056
            try:
 
1057
                revtree = self.branch.repository.revision_tree(revision_id)
 
1058
                # TODO: jam 20070213 KnitVersionedFile raises
 
1059
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1060
                #       given revision_id is not present. Should Repository be
 
1061
                #       catching it and re-raising NoSuchRevision?
 
1062
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1063
                revtree = None
 
1064
            trees.append((revision_id, revtree))
 
1065
        self.set_parent_trees(trees,
 
1066
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1067
 
 
1068
    @needs_tree_write_lock
 
1069
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1070
        """Set the parents of the working tree.
 
1071
 
 
1072
        :param parents_list: A list of (revision_id, tree) tuples.
 
1073
            If tree is None, then that element is treated as an unreachable
 
1074
            parent tree - i.e. a ghost.
 
1075
        """
 
1076
        dirstate = self.current_dirstate()
 
1077
        if len(parents_list) > 0:
 
1078
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1079
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1080
        real_trees = []
 
1081
        ghosts = []
 
1082
        # convert absent trees to the null tree, which we convert back to
 
1083
        # missing on access.
 
1084
        for rev_id, tree in parents_list:
 
1085
            _mod_revision.check_not_reserved_id(rev_id)
 
1086
            if tree is not None:
 
1087
                real_trees.append((rev_id, tree))
 
1088
            else:
 
1089
                real_trees.append((rev_id,
 
1090
                    self.branch.repository.revision_tree(None)))
 
1091
                ghosts.append(rev_id)
 
1092
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1093
        self._make_dirty(reset_inventory=False)
 
1094
 
 
1095
    def _set_root_id(self, file_id):
 
1096
        """See WorkingTree.set_root_id."""
 
1097
        state = self.current_dirstate()
 
1098
        state.set_path_id('', file_id)
 
1099
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1100
            self._make_dirty(reset_inventory=True)
 
1101
 
 
1102
    def _sha_from_stat(self, path, stat_result):
 
1103
        """Get a sha digest from the tree's stat cache.
 
1104
 
 
1105
        The default implementation assumes no stat cache is present.
 
1106
 
 
1107
        :param path: The path.
 
1108
        :param stat_result: The stat result being looked up.
 
1109
        """
 
1110
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1111
 
 
1112
    @needs_read_lock
 
1113
    def supports_tree_reference(self):
 
1114
        return self._repo_supports_tree_reference
 
1115
 
 
1116
    def unlock(self):
 
1117
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1118
        # do non-implementation specific cleanup
 
1119
        self._cleanup()
 
1120
 
 
1121
        if self._control_files._lock_count == 1:
 
1122
            # eventually we should do signature checking during read locks for
 
1123
            # dirstate updates.
 
1124
            if self._control_files._lock_mode == 'w':
 
1125
                if self._dirty:
 
1126
                    self.flush()
 
1127
            if self._dirstate is not None:
 
1128
                # This is a no-op if there are no modifications.
 
1129
                self._dirstate.save()
 
1130
                self._dirstate.unlock()
 
1131
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1132
            #       point. Instead, it could check if the header has been
 
1133
            #       modified when it is locked, and if not, it can hang on to
 
1134
            #       the data it has in memory.
 
1135
            self._dirstate = None
 
1136
            self._inventory = None
 
1137
        # reverse order of locking.
 
1138
        try:
 
1139
            return self._control_files.unlock()
 
1140
        finally:
 
1141
            self.branch.unlock()
 
1142
 
 
1143
    @needs_tree_write_lock
 
1144
    def unversion(self, file_ids):
 
1145
        """Remove the file ids in file_ids from the current versioned set.
 
1146
 
 
1147
        When a file_id is unversioned, all of its children are automatically
 
1148
        unversioned.
 
1149
 
 
1150
        :param file_ids: The file ids to stop versioning.
 
1151
        :raises: NoSuchId if any fileid is not currently versioned.
 
1152
        """
 
1153
        if not file_ids:
 
1154
            return
 
1155
        state = self.current_dirstate()
 
1156
        state._read_dirblocks_if_needed()
 
1157
        ids_to_unversion = set(file_ids)
 
1158
        paths_to_unversion = set()
 
1159
        # sketch:
 
1160
        # check if the root is to be unversioned, if so, assert for now.
 
1161
        # walk the state marking unversioned things as absent.
 
1162
        # if there are any un-unversioned ids at the end, raise
 
1163
        for key, details in state._dirblocks[0][1]:
 
1164
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1165
                key[2] in ids_to_unversion):
 
1166
                # I haven't written the code to unversion / yet - it should be
 
1167
                # supported.
 
1168
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1169
        block_index = 0
 
1170
        while block_index < len(state._dirblocks):
 
1171
            # process one directory at a time.
 
1172
            block = state._dirblocks[block_index]
 
1173
            # first check: is the path one to remove - it or its children
 
1174
            delete_block = False
 
1175
            for path in paths_to_unversion:
 
1176
                if (block[0].startswith(path) and
 
1177
                    (len(block[0]) == len(path) or
 
1178
                     block[0][len(path)] == '/')):
 
1179
                    # this entire block should be deleted - its the block for a
 
1180
                    # path to unversion; or the child of one
 
1181
                    delete_block = True
 
1182
                    break
 
1183
            # TODO: trim paths_to_unversion as we pass by paths
 
1184
            if delete_block:
 
1185
                # this block is to be deleted: process it.
 
1186
                # TODO: we can special case the no-parents case and
 
1187
                # just forget the whole block.
 
1188
                entry_index = 0
 
1189
                while entry_index < len(block[1]):
 
1190
                    # Mark this file id as having been removed
 
1191
                    entry = block[1][entry_index]
 
1192
                    ids_to_unversion.discard(entry[0][2])
 
1193
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1194
                                               # entries
 
1195
                        or not state._make_absent(entry)):
 
1196
                        entry_index += 1
 
1197
                # go to the next block. (At the moment we dont delete empty
 
1198
                # dirblocks)
 
1199
                block_index += 1
 
1200
                continue
 
1201
            entry_index = 0
 
1202
            while entry_index < len(block[1]):
 
1203
                entry = block[1][entry_index]
 
1204
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1205
                    # ^ some parent row.
 
1206
                    entry[0][2] not in ids_to_unversion):
 
1207
                    # ^ not an id to unversion
 
1208
                    entry_index += 1
 
1209
                    continue
 
1210
                if entry[1][0][0] == 'd':
 
1211
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1212
                if not state._make_absent(entry):
 
1213
                    entry_index += 1
 
1214
                # we have unversioned this id
 
1215
                ids_to_unversion.remove(entry[0][2])
 
1216
            block_index += 1
 
1217
        if ids_to_unversion:
 
1218
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1219
        self._make_dirty(reset_inventory=False)
 
1220
        # have to change the legacy inventory too.
 
1221
        if self._inventory is not None:
 
1222
            for file_id in file_ids:
 
1223
                self._inventory.remove_recursive_id(file_id)
 
1224
 
 
1225
    @needs_tree_write_lock
 
1226
    def apply_inventory_delta(self, changes):
 
1227
        """See MutableTree.apply_inventory_delta"""
 
1228
        self.current_dirstate().update_by_delta(changes, self.basedir)
 
1229
 
 
1230
    def update_basis_by_delta(self, new_revid, delta):
 
1231
        """See MutableTree.update_basis_by_delta."""
 
1232
        assert self.last_revision() != new_revid
 
1233
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1234
 
 
1235
    @needs_read_lock
 
1236
    def _validate(self):
 
1237
        self._dirstate._validate()
 
1238
 
 
1239
    @needs_tree_write_lock
 
1240
    def _write_inventory(self, inv):
 
1241
        """Write inventory as the current inventory."""
 
1242
        assert not self._dirty, ("attempting to write an inventory when the "
 
1243
            "dirstate is dirty will cause data loss")
 
1244
        self.current_dirstate().set_state_from_inventory(inv)
 
1245
        self._make_dirty(reset_inventory=False)
 
1246
        if self._inventory is not None:
 
1247
            self._inventory = inv
 
1248
        self.flush()
 
1249
 
 
1250
 
 
1251
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1252
    """The first consolidated dirstate working tree format.
 
1253
 
 
1254
    This format:
 
1255
        - exists within a metadir controlling .bzr
 
1256
        - includes an explicit version marker for the workingtree control
 
1257
          files, separate from the BzrDir format
 
1258
        - modifies the hash cache format
 
1259
        - is new in bzr 0.15
 
1260
        - uses a LockDir to guard access to it.
 
1261
    """
 
1262
 
 
1263
    upgrade_recommended = False
 
1264
 
 
1265
    def get_format_string(self):
 
1266
        """See WorkingTreeFormat.get_format_string()."""
 
1267
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1268
 
 
1269
    def get_format_description(self):
 
1270
        """See WorkingTreeFormat.get_format_description()."""
 
1271
        return "Working tree format 4"
 
1272
 
 
1273
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1274
                   accelerator_tree=None):
 
1275
        """See WorkingTreeFormat.initialize().
 
1276
 
 
1277
        :param revision_id: allows creating a working tree at a different
 
1278
        revision than the branch is at.
 
1279
        :param accelerator_tree: A tree which can be used for retrieving file
 
1280
            contents more quickly than the revision tree, i.e. a workingtree.
 
1281
            The revision tree will be used for cases where accelerator_tree's
 
1282
            content is different.
 
1283
 
 
1284
        These trees get an initial random root id, if their repository supports
 
1285
        rich root data, TREE_ROOT otherwise.
 
1286
        """
 
1287
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1288
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1289
        transport = a_bzrdir.get_workingtree_transport(self)
 
1290
        control_files = self._open_control_files(a_bzrdir)
 
1291
        control_files.create_lock()
 
1292
        control_files.lock_write()
 
1293
        control_files.put_utf8('format', self.get_format_string())
 
1294
        if from_branch is not None:
 
1295
            branch = from_branch
 
1296
        else:
 
1297
            branch = a_bzrdir.open_branch()
 
1298
        if revision_id is None:
 
1299
            revision_id = branch.last_revision()
 
1300
        local_path = transport.local_abspath('dirstate')
 
1301
        # write out new dirstate (must exist when we create the tree)
 
1302
        state = dirstate.DirState.initialize(local_path)
 
1303
        state.unlock()
 
1304
        del state
 
1305
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1306
                         branch,
 
1307
                         _format=self,
 
1308
                         _bzrdir=a_bzrdir,
 
1309
                         _control_files=control_files)
 
1310
        wt._new_tree()
 
1311
        wt.lock_tree_write()
 
1312
        try:
 
1313
            if revision_id in (None, NULL_REVISION):
 
1314
                if branch.repository.supports_rich_root():
 
1315
                    wt._set_root_id(generate_ids.gen_root_id())
 
1316
                else:
 
1317
                    wt._set_root_id(ROOT_ID)
 
1318
                wt.flush()
 
1319
            basis = None
 
1320
            # frequently, we will get here due to branching.  The accelerator
 
1321
            # tree will be the tree from the branch, so the desired basis
 
1322
            # tree will often be a parent of the accelerator tree.
 
1323
            if accelerator_tree is not None:
 
1324
                try:
 
1325
                    basis = accelerator_tree.revision_tree(revision_id)
 
1326
                except errors.NoSuchRevision:
 
1327
                    pass
 
1328
            if basis is None:
 
1329
                basis = branch.repository.revision_tree(revision_id)
 
1330
            if revision_id == NULL_REVISION:
 
1331
                parents_list = []
 
1332
            else:
 
1333
                parents_list = [(revision_id, basis)]
 
1334
            basis.lock_read()
 
1335
            wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1336
            wt.flush()
 
1337
            # if the basis has a root id we have to use that; otherwise we use
 
1338
            # a new random one
 
1339
            basis_root_id = basis.get_root_id()
 
1340
            if basis_root_id is not None:
 
1341
                wt._set_root_id(basis_root_id)
 
1342
                wt.flush()
 
1343
            transform.build_tree(basis, wt, accelerator_tree)
 
1344
            basis.unlock()
 
1345
        finally:
 
1346
            control_files.unlock()
 
1347
            wt.unlock()
 
1348
        return wt
 
1349
 
 
1350
    def _open(self, a_bzrdir, control_files):
 
1351
        """Open the tree itself.
 
1352
 
 
1353
        :param a_bzrdir: the dir for the tree.
 
1354
        :param control_files: the control files for the tree.
 
1355
        """
 
1356
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1357
                           branch=a_bzrdir.open_branch(),
 
1358
                           _format=self,
 
1359
                           _bzrdir=a_bzrdir,
 
1360
                           _control_files=control_files)
 
1361
 
 
1362
    def __get_matchingbzrdir(self):
 
1363
        # please test against something that will let us do tree references
 
1364
        return bzrdir.format_registry.make_bzrdir(
 
1365
            'dirstate-with-subtree')
 
1366
 
 
1367
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1368
 
 
1369
 
 
1370
class DirStateRevisionTree(Tree):
 
1371
    """A revision tree pulling the inventory from a dirstate."""
 
1372
 
 
1373
    def __init__(self, dirstate, revision_id, repository):
 
1374
        self._dirstate = dirstate
 
1375
        self._revision_id = revision_id
 
1376
        self._repository = repository
 
1377
        self._inventory = None
 
1378
        self._locked = 0
 
1379
        self._dirstate_locked = False
 
1380
 
 
1381
    def __repr__(self):
 
1382
        return "<%s of %s in %s>" % \
 
1383
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1384
 
 
1385
    def annotate_iter(self, file_id,
 
1386
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1387
        """See Tree.annotate_iter"""
 
1388
        w = self._get_weave(file_id)
 
1389
        return w.annotate_iter(self.inventory[file_id].revision)
 
1390
 
 
1391
    def _get_ancestors(self, default_revision):
 
1392
        return set(self._repository.get_ancestry(self._revision_id,
 
1393
                                                 topo_sorted=False))
 
1394
    def _comparison_data(self, entry, path):
 
1395
        """See Tree._comparison_data."""
 
1396
        if entry is None:
 
1397
            return None, False, None
 
1398
        # trust the entry as RevisionTree does, but this may not be
 
1399
        # sensible: the entry might not have come from us?
 
1400
        return entry.kind, entry.executable, None
 
1401
 
 
1402
    def _file_size(self, entry, stat_value):
 
1403
        return entry.text_size
 
1404
 
 
1405
    def filter_unversioned_files(self, paths):
 
1406
        """Filter out paths that are not versioned.
 
1407
 
 
1408
        :return: set of paths.
 
1409
        """
 
1410
        pred = self.has_filename
 
1411
        return set((p for p in paths if not pred(p)))
 
1412
 
 
1413
    def get_root_id(self):
 
1414
        return self.path2id('')
 
1415
 
 
1416
    def id2path(self, file_id):
 
1417
        "Convert a file-id to a path."
 
1418
        entry = self._get_entry(file_id=file_id)
 
1419
        if entry == (None, None):
 
1420
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1421
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1422
        return path_utf8.decode('utf8')
 
1423
 
 
1424
    def _get_parent_index(self):
 
1425
        """Return the index in the dirstate referenced by this tree."""
 
1426
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1427
 
 
1428
    def _get_entry(self, file_id=None, path=None):
 
1429
        """Get the dirstate row for file_id or path.
 
1430
 
 
1431
        If either file_id or path is supplied, it is used as the key to lookup.
 
1432
        If both are supplied, the fastest lookup is used, and an error is
 
1433
        raised if they do not both point at the same row.
 
1434
        
 
1435
        :param file_id: An optional unicode file_id to be looked up.
 
1436
        :param path: An optional unicode path to be looked up.
 
1437
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1438
        """
 
1439
        if file_id is None and path is None:
 
1440
            raise errors.BzrError('must supply file_id or path')
 
1441
        if path is not None:
 
1442
            path = path.encode('utf8')
 
1443
        parent_index = self._get_parent_index()
 
1444
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1445
 
 
1446
    def _generate_inventory(self):
 
1447
        """Create and set self.inventory from the dirstate object.
 
1448
 
 
1449
        (So this is only called the first time the inventory is requested for
 
1450
        this tree; it then remains in memory until it's out of date.)
 
1451
 
 
1452
        This is relatively expensive: we have to walk the entire dirstate.
 
1453
        """
 
1454
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1455
            'dirstate revision tree'
 
1456
        # separate call for profiling - makes it clear where the costs are.
 
1457
        self._dirstate._read_dirblocks_if_needed()
 
1458
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1459
            'parent %s has disappeared from %s' % (
 
1460
            self._revision_id, self._dirstate.get_parent_ids())
 
1461
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1462
        # This is identical now to the WorkingTree _generate_inventory except
 
1463
        # for the tree index use.
 
1464
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1465
        current_id = root_key[2]
 
1466
        assert current_entry[parent_index][0] == 'd'
 
1467
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1468
        inv.root.revision = current_entry[parent_index][4]
 
1469
        # Turn some things into local variables
 
1470
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1471
        factory = entry_factory
 
1472
        utf8_decode = cache_utf8._utf8_decode
 
1473
        inv_byid = inv._byid
 
1474
        # we could do this straight out of the dirstate; it might be fast
 
1475
        # and should be profiled - RBC 20070216
 
1476
        parent_ies = {'' : inv.root}
 
1477
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1478
            dirname = block[0]
 
1479
            try:
 
1480
                parent_ie = parent_ies[dirname]
 
1481
            except KeyError:
 
1482
                # all the paths in this block are not versioned in this tree
 
1483
                continue
 
1484
            for key, entry in block[1]:
 
1485
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1486
                if minikind in ('a', 'r'): # absent, relocated
 
1487
                    # not this tree
 
1488
                    continue
 
1489
                name = key[1]
 
1490
                name_unicode = utf8_decode(name)[0]
 
1491
                file_id = key[2]
 
1492
                kind = minikind_to_kind[minikind]
 
1493
                inv_entry = factory[kind](file_id, name_unicode,
 
1494
                                          parent_ie.file_id)
 
1495
                inv_entry.revision = revid
 
1496
                if kind == 'file':
 
1497
                    inv_entry.executable = executable
 
1498
                    inv_entry.text_size = size
 
1499
                    inv_entry.text_sha1 = fingerprint
 
1500
                elif kind == 'directory':
 
1501
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1502
                elif kind == 'symlink':
 
1503
                    inv_entry.executable = False
 
1504
                    inv_entry.text_size = None
 
1505
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1506
                elif kind == 'tree-reference':
 
1507
                    inv_entry.reference_revision = fingerprint or None
 
1508
                else:
 
1509
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1510
                            % entry)
 
1511
                # These checks cost us around 40ms on a 55k entry tree
 
1512
                assert file_id not in inv_byid
 
1513
                assert name_unicode not in parent_ie.children
 
1514
                inv_byid[file_id] = inv_entry
 
1515
                parent_ie.children[name_unicode] = inv_entry
 
1516
        self._inventory = inv
 
1517
 
 
1518
    def get_file_mtime(self, file_id, path=None):
 
1519
        """Return the modification time for this record.
 
1520
 
 
1521
        We return the timestamp of the last-changed revision.
 
1522
        """
 
1523
        # Make sure the file exists
 
1524
        entry = self._get_entry(file_id, path=path)
 
1525
        if entry == (None, None): # do we raise?
 
1526
            return None
 
1527
        parent_index = self._get_parent_index()
 
1528
        last_changed_revision = entry[1][parent_index][4]
 
1529
        return self._repository.get_revision(last_changed_revision).timestamp
 
1530
 
 
1531
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1532
        entry = self._get_entry(file_id=file_id, path=path)
 
1533
        parent_index = self._get_parent_index()
 
1534
        parent_details = entry[1][parent_index]
 
1535
        if parent_details[0] == 'f':
 
1536
            return parent_details[1]
 
1537
        return None
 
1538
 
 
1539
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1540
    def get_weave(self, file_id):
 
1541
        return self._get_weave(file_id)
 
1542
 
 
1543
    def _get_weave(self, file_id):
 
1544
        return self._repository.weave_store.get_weave(file_id,
 
1545
                self._repository.get_transaction())
 
1546
 
 
1547
    def get_file(self, file_id, path=None):
 
1548
        return StringIO(self.get_file_text(file_id))
 
1549
 
 
1550
    def get_file_lines(self, file_id):
 
1551
        entry = self._get_entry(file_id=file_id)[1]
 
1552
        if entry == None:
 
1553
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1554
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1555
 
 
1556
    def get_file_size(self, file_id):
 
1557
        return self.inventory[file_id].text_size
 
1558
 
 
1559
    def get_file_text(self, file_id):
 
1560
        return ''.join(self.get_file_lines(file_id))
 
1561
 
 
1562
    def get_reference_revision(self, file_id, path=None):
 
1563
        return self.inventory[file_id].reference_revision
 
1564
 
 
1565
    def iter_files_bytes(self, desired_files):
 
1566
        """See Tree.iter_files_bytes.
 
1567
 
 
1568
        This version is implemented on top of Repository.iter_files_bytes"""
 
1569
        parent_index = self._get_parent_index()
 
1570
        repo_desired_files = []
 
1571
        for file_id, identifier in desired_files:
 
1572
            entry = self._get_entry(file_id)
 
1573
            if entry == (None, None):
 
1574
                raise errors.NoSuchId(self, file_id)
 
1575
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1576
                                       identifier))
 
1577
        return self._repository.iter_files_bytes(repo_desired_files)
 
1578
 
 
1579
    def get_symlink_target(self, file_id):
 
1580
        entry = self._get_entry(file_id=file_id)
 
1581
        parent_index = self._get_parent_index()
 
1582
        if entry[1][parent_index][0] != 'l':
 
1583
            return None
 
1584
        else:
 
1585
            # At present, none of the tree implementations supports non-ascii
 
1586
            # symlink targets. So we will just assume that the dirstate path is
 
1587
            # correct.
 
1588
            return entry[1][parent_index][1]
 
1589
 
 
1590
    def get_revision_id(self):
 
1591
        """Return the revision id for this tree."""
 
1592
        return self._revision_id
 
1593
 
 
1594
    def _get_inventory(self):
 
1595
        if self._inventory is not None:
 
1596
            return self._inventory
 
1597
        self._must_be_locked()
 
1598
        self._generate_inventory()
 
1599
        return self._inventory
 
1600
 
 
1601
    inventory = property(_get_inventory,
 
1602
                         doc="Inventory of this Tree")
 
1603
 
 
1604
    def get_parent_ids(self):
 
1605
        """The parents of a tree in the dirstate are not cached."""
 
1606
        return self._repository.get_revision(self._revision_id).parent_ids
 
1607
 
 
1608
    def has_filename(self, filename):
 
1609
        return bool(self.path2id(filename))
 
1610
 
 
1611
    def kind(self, file_id):
 
1612
        entry = self._get_entry(file_id=file_id)[1]
 
1613
        if entry == None:
 
1614
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1615
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1616
 
 
1617
    def stored_kind(self, file_id):
 
1618
        """See Tree.stored_kind"""
 
1619
        return self.kind(file_id)
 
1620
 
 
1621
    def path_content_summary(self, path):
 
1622
        """See Tree.path_content_summary."""
 
1623
        id = self.inventory.path2id(path)
 
1624
        if id is None:
 
1625
            return ('missing', None, None, None)
 
1626
        entry = self._inventory[id]
 
1627
        kind = entry.kind
 
1628
        if kind == 'file':
 
1629
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1630
        elif kind == 'symlink':
 
1631
            return (kind, None, None, entry.symlink_target)
 
1632
        else:
 
1633
            return (kind, None, None, None)
 
1634
 
 
1635
    def is_executable(self, file_id, path=None):
 
1636
        ie = self.inventory[file_id]
 
1637
        if ie.kind != "file":
 
1638
            return None
 
1639
        return ie.executable
 
1640
 
 
1641
    def list_files(self, include_root=False):
 
1642
        # We use a standard implementation, because DirStateRevisionTree is
 
1643
        # dealing with one of the parents of the current state
 
1644
        inv = self._get_inventory()
 
1645
        entries = inv.iter_entries()
 
1646
        if self.inventory.root is not None and not include_root:
 
1647
            entries.next()
 
1648
        for path, entry in entries:
 
1649
            yield path, 'V', entry.kind, entry.file_id, entry
 
1650
 
 
1651
    def lock_read(self):
 
1652
        """Lock the tree for a set of operations."""
 
1653
        if not self._locked:
 
1654
            self._repository.lock_read()
 
1655
            if self._dirstate._lock_token is None:
 
1656
                self._dirstate.lock_read()
 
1657
                self._dirstate_locked = True
 
1658
        self._locked += 1
 
1659
 
 
1660
    def _must_be_locked(self):
 
1661
        if not self._locked:
 
1662
            raise errors.ObjectNotLocked(self)
 
1663
 
 
1664
    @needs_read_lock
 
1665
    def path2id(self, path):
 
1666
        """Return the id for path in this tree."""
 
1667
        # lookup by path: faster than splitting and walking the ivnentory.
 
1668
        entry = self._get_entry(path=path)
 
1669
        if entry == (None, None):
 
1670
            return None
 
1671
        return entry[0][2]
 
1672
 
 
1673
    def unlock(self):
 
1674
        """Unlock, freeing any cache memory used during the lock."""
 
1675
        # outside of a lock, the inventory is suspect: release it.
 
1676
        self._locked -=1
 
1677
        if not self._locked:
 
1678
            self._inventory = None
 
1679
            self._locked = 0
 
1680
            if self._dirstate_locked:
 
1681
                self._dirstate.unlock()
 
1682
                self._dirstate_locked = False
 
1683
            self._repository.unlock()
 
1684
 
 
1685
    def walkdirs(self, prefix=""):
 
1686
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1687
        # implementation based on an inventory.  
 
1688
        # This should be cleaned up to use the much faster Dirstate code
 
1689
        # So for now, we just build up the parent inventory, and extract
 
1690
        # it the same way RevisionTree does.
 
1691
        _directory = 'directory'
 
1692
        inv = self._get_inventory()
 
1693
        top_id = inv.path2id(prefix)
 
1694
        if top_id is None:
 
1695
            pending = []
 
1696
        else:
 
1697
            pending = [(prefix, top_id)]
 
1698
        while pending:
 
1699
            dirblock = []
 
1700
            relpath, file_id = pending.pop()
 
1701
            # 0 - relpath, 1- file-id
 
1702
            if relpath:
 
1703
                relroot = relpath + '/'
 
1704
            else:
 
1705
                relroot = ""
 
1706
            # FIXME: stash the node in pending
 
1707
            entry = inv[file_id]
 
1708
            for name, child in entry.sorted_children():
 
1709
                toppath = relroot + name
 
1710
                dirblock.append((toppath, name, child.kind, None,
 
1711
                    child.file_id, child.kind
 
1712
                    ))
 
1713
            yield (relpath, entry.file_id), dirblock
 
1714
            # push the user specified dirs from dirblock
 
1715
            for dir in reversed(dirblock):
 
1716
                if dir[2] == _directory:
 
1717
                    pending.append((dir[0], dir[4]))
 
1718
 
 
1719
 
 
1720
class InterDirStateTree(InterTree):
 
1721
    """Fast path optimiser for changes_from with dirstate trees.
 
1722
    
 
1723
    This is used only when both trees are in the dirstate working file, and 
 
1724
    the source is any parent within the dirstate, and the destination is 
 
1725
    the current working tree of the same dirstate.
 
1726
    """
 
1727
    # this could be generalized to allow comparisons between any trees in the
 
1728
    # dirstate, and possibly between trees stored in different dirstates.
 
1729
 
 
1730
    def __init__(self, source, target):
 
1731
        super(InterDirStateTree, self).__init__(source, target)
 
1732
        if not InterDirStateTree.is_compatible(source, target):
 
1733
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1734
 
 
1735
    @staticmethod
 
1736
    def make_source_parent_tree(source, target):
 
1737
        """Change the source tree into a parent of the target."""
 
1738
        revid = source.commit('record tree')
 
1739
        target.branch.repository.fetch(source.branch.repository, revid)
 
1740
        target.set_parent_ids([revid])
 
1741
        return target.basis_tree(), target
 
1742
 
 
1743
    _matching_from_tree_format = WorkingTreeFormat4()
 
1744
    _matching_to_tree_format = WorkingTreeFormat4()
 
1745
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1746
 
 
1747
    def _iter_changes(self, include_unchanged=False,
 
1748
                      specific_files=None, pb=None, extra_trees=[],
 
1749
                      require_versioned=True, want_unversioned=False):
 
1750
        """Return the changes from source to target.
 
1751
 
 
1752
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1753
            for details.
 
1754
        :param specific_files: An optional list of file paths to restrict the
 
1755
            comparison to. When mapping filenames to ids, all matches in all
 
1756
            trees (including optional extra_trees) are used, and all children of
 
1757
            matched directories are included.
 
1758
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1759
            unchanged entries in the result.
 
1760
        :param extra_trees: An optional list of additional trees to use when
 
1761
            mapping the contents of specific_files (paths) to file_ids.
 
1762
        :param require_versioned: If True, all files in specific_files must be
 
1763
            versioned in one of source, target, extra_trees or
 
1764
            PathsNotVersionedError is raised.
 
1765
        :param want_unversioned: Should unversioned files be returned in the
 
1766
            output. An unversioned file is defined as one with (False, False)
 
1767
            for the versioned pair.
 
1768
        """
 
1769
        utf8_decode = cache_utf8._utf8_decode
 
1770
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1771
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1772
        # NB: show_status depends on being able to pass in non-versioned files
 
1773
        # and report them as unknown
 
1774
        # TODO: handle extra trees in the dirstate.
 
1775
        if (extra_trees or specific_files == []):
 
1776
            # we can't fast-path these cases (yet)
 
1777
            for f in super(InterDirStateTree, self)._iter_changes(
 
1778
                include_unchanged, specific_files, pb, extra_trees,
 
1779
                require_versioned, want_unversioned=want_unversioned):
 
1780
                yield f
 
1781
            return
 
1782
        parent_ids = self.target.get_parent_ids()
 
1783
        assert (self.source._revision_id in parent_ids
 
1784
                or self.source._revision_id == NULL_REVISION), \
 
1785
                "revision {%s} is not stored in {%s}, but %s " \
 
1786
                "can only be used for trees stored in the dirstate" \
 
1787
                % (self.source._revision_id, self.target, self._iter_changes)
 
1788
        target_index = 0
 
1789
        if self.source._revision_id == NULL_REVISION:
 
1790
            source_index = None
 
1791
            indices = (target_index,)
 
1792
        else:
 
1793
            assert (self.source._revision_id in parent_ids), \
 
1794
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1795
                self.source._revision_id, parent_ids)
 
1796
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1797
            indices = (source_index, target_index)
 
1798
        # -- make all specific_files utf8 --
 
1799
        if specific_files:
 
1800
            specific_files_utf8 = set()
 
1801
            for path in specific_files:
 
1802
                specific_files_utf8.add(path.encode('utf8'))
 
1803
            specific_files = specific_files_utf8
 
1804
        else:
 
1805
            specific_files = set([''])
 
1806
        # -- specific_files is now a utf8 path set --
 
1807
        # -- get the state object and prepare it.
 
1808
        state = self.target.current_dirstate()
 
1809
        state._read_dirblocks_if_needed()
 
1810
        def _entries_for_path(path):
 
1811
            """Return a list with all the entries that match path for all ids.
 
1812
            """
 
1813
            dirname, basename = os.path.split(path)
 
1814
            key = (dirname, basename, '')
 
1815
            block_index, present = state._find_block_index_from_key(key)
 
1816
            if not present:
 
1817
                # the block which should contain path is absent.
 
1818
                return []
 
1819
            result = []
 
1820
            block = state._dirblocks[block_index][1]
 
1821
            entry_index, _ = state._find_entry_index(key, block)
 
1822
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1823
            while (entry_index < len(block) and
 
1824
                block[entry_index][0][0:2] == key[0:2]):
 
1825
                result.append(block[entry_index])
 
1826
                entry_index += 1
 
1827
            return result
 
1828
        if require_versioned:
 
1829
            # -- check all supplied paths are versioned in a search tree. --
 
1830
            all_versioned = True
 
1831
            for path in specific_files:
 
1832
                path_entries = _entries_for_path(path)
 
1833
                if not path_entries:
 
1834
                    # this specified path is not present at all: error
 
1835
                    all_versioned = False
 
1836
                    break
 
1837
                found_versioned = False
 
1838
                # for each id at this path
 
1839
                for entry in path_entries:
 
1840
                    # for each tree.
 
1841
                    for index in indices:
 
1842
                        if entry[1][index][0] != 'a': # absent
 
1843
                            found_versioned = True
 
1844
                            # all good: found a versioned cell
 
1845
                            break
 
1846
                if not found_versioned:
 
1847
                    # none of the indexes was not 'absent' at all ids for this
 
1848
                    # path.
 
1849
                    all_versioned = False
 
1850
                    break
 
1851
            if not all_versioned:
 
1852
                raise errors.PathsNotVersionedError(specific_files)
 
1853
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1854
        search_specific_files = set()
 
1855
        for path in specific_files:
 
1856
            other_specific_files = specific_files.difference(set([path]))
 
1857
            if not osutils.is_inside_any(other_specific_files, path):
 
1858
                # this is a top level path, we must check it.
 
1859
                search_specific_files.add(path)
 
1860
        # sketch: 
 
1861
        # compare source_index and target_index at or under each element of search_specific_files.
 
1862
        # follow the following comparison table. Note that we only want to do diff operations when
 
1863
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1864
        # for the target.
 
1865
        # cases:
 
1866
        # 
 
1867
        # Source | Target | disk | action
 
1868
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1869
        #        |        |      | diff check on source-target
 
1870
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1871
        #        |        |      | ???
 
1872
        #   r    |  a     |      | add source to search
 
1873
        #   r    |  a     |  a   | 
 
1874
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1875
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1876
        #   a    | fdlt   |      | add new id
 
1877
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1878
        #   a    |  a     |      | not present in either tree, skip
 
1879
        #   a    |  a     |  a   | not present in any tree, skip
 
1880
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1881
        #        |        |      | may not be selected by the users list of paths.
 
1882
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1883
        #        |        |      | may not be selected by the users list of paths.
 
1884
        #  fdlt  | fdlt   |      | content in both: diff them
 
1885
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1886
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1887
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1888
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1889
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1890
        #        |        |      | this id at the other path.
 
1891
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1892
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1893
        #        |        |      | this id at the other path.
 
1894
 
 
1895
        # for all search_indexs in each path at or under each element of
 
1896
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1897
        # relocated path as one to search if its not searched already. If the
 
1898
        # detail is not relocated, add the id.
 
1899
        searched_specific_files = set()
 
1900
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1901
        # Using a list so that we can access the values and change them in
 
1902
        # nested scope. Each one is [path, file_id, entry]
 
1903
        last_source_parent = [None, None]
 
1904
        last_target_parent = [None, None]
 
1905
 
 
1906
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1907
 
 
1908
        # Just a sentry, so that _process_entry can say that this
 
1909
        # record is handled, but isn't interesting to process (unchanged)
 
1910
        uninteresting = object()
 
1911
 
 
1912
 
 
1913
        old_dirname_to_file_id = {}
 
1914
        new_dirname_to_file_id = {}
 
1915
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1916
        #       keeping a cache of directories that we have seen.
 
1917
 
 
1918
        def _process_entry(entry, path_info):
 
1919
            """Compare an entry and real disk to generate delta information.
 
1920
 
 
1921
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1922
                the path of entry. If None, then the path is considered absent.
 
1923
                (Perhaps we should pass in a concrete entry for this ?)
 
1924
                Basename is returned as a utf8 string because we expect this
 
1925
                tuple will be ignored, and don't want to take the time to
 
1926
                decode.
 
1927
            :return: None if these don't match
 
1928
                     A tuple of information about the change, or
 
1929
                     the object 'uninteresting' if these match, but are
 
1930
                     basically identical.
 
1931
            """
 
1932
            if source_index is None:
 
1933
                source_details = NULL_PARENT_DETAILS
 
1934
            else:
 
1935
                source_details = entry[1][source_index]
 
1936
            target_details = entry[1][target_index]
 
1937
            target_minikind = target_details[0]
 
1938
            if path_info is not None and target_minikind in 'fdlt':
 
1939
                assert target_index == 0
 
1940
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1941
                                                  stat_value=path_info[3])
 
1942
                # The entry may have been modified by update_entry
 
1943
                target_details = entry[1][target_index]
 
1944
                target_minikind = target_details[0]
 
1945
            else:
 
1946
                link_or_sha1 = None
 
1947
            file_id = entry[0][2]
 
1948
            source_minikind = source_details[0]
 
1949
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1950
                # claimed content in both: diff
 
1951
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1952
                #        |        |      | diff check on source-target
 
1953
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1954
                #        |        |      | ???
 
1955
                if source_minikind in 'r':
 
1956
                    # add the source to the search path to find any children it
 
1957
                    # has.  TODO ? : only add if it is a container ?
 
1958
                    if not osutils.is_inside_any(searched_specific_files,
 
1959
                                                 source_details[1]):
 
1960
                        search_specific_files.add(source_details[1])
 
1961
                    # generate the old path; this is needed for stating later
 
1962
                    # as well.
 
1963
                    old_path = source_details[1]
 
1964
                    old_dirname, old_basename = os.path.split(old_path)
 
1965
                    path = pathjoin(entry[0][0], entry[0][1])
 
1966
                    old_entry = state._get_entry(source_index,
 
1967
                                                 path_utf8=old_path)
 
1968
                    # update the source details variable to be the real
 
1969
                    # location.
 
1970
                    source_details = old_entry[1][source_index]
 
1971
                    source_minikind = source_details[0]
 
1972
                else:
 
1973
                    old_dirname = entry[0][0]
 
1974
                    old_basename = entry[0][1]
 
1975
                    old_path = path = None
 
1976
                if path_info is None:
 
1977
                    # the file is missing on disk, show as removed.
 
1978
                    content_change = True
 
1979
                    target_kind = None
 
1980
                    target_exec = False
 
1981
                else:
 
1982
                    # source and target are both versioned and disk file is present.
 
1983
                    target_kind = path_info[2]
 
1984
                    if target_kind == 'directory':
 
1985
                        if path is None:
 
1986
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1987
                        new_dirname_to_file_id[path] = file_id
 
1988
                        if source_minikind != 'd':
 
1989
                            content_change = True
 
1990
                        else:
 
1991
                            # directories have no fingerprint
 
1992
                            content_change = False
 
1993
                        target_exec = False
 
1994
                    elif target_kind == 'file':
 
1995
                        if source_minikind != 'f':
 
1996
                            content_change = True
 
1997
                        else:
 
1998
                            # We could check the size, but we already have the
 
1999
                            # sha1 hash.
 
2000
                            content_change = (link_or_sha1 != source_details[1])
 
2001
                        # Target details is updated at update_entry time
 
2002
                        if use_filesystem_for_exec:
 
2003
                            # We don't need S_ISREG here, because we are sure
 
2004
                            # we are dealing with a file.
 
2005
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2006
                        else:
 
2007
                            target_exec = target_details[3]
 
2008
                    elif target_kind == 'symlink':
 
2009
                        if source_minikind != 'l':
 
2010
                            content_change = True
 
2011
                        else:
 
2012
                            content_change = (link_or_sha1 != source_details[1])
 
2013
                        target_exec = False
 
2014
                    elif target_kind == 'tree-reference':
 
2015
                        if source_minikind != 't':
 
2016
                            content_change = True
 
2017
                        else:
 
2018
                            content_change = False
 
2019
                        target_exec = False
 
2020
                    else:
 
2021
                        raise Exception, "unknown kind %s" % path_info[2]
 
2022
                if source_minikind == 'd':
 
2023
                    if path is None:
 
2024
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2025
                    old_dirname_to_file_id[old_path] = file_id
 
2026
                # parent id is the entry for the path in the target tree
 
2027
                if old_dirname == last_source_parent[0]:
 
2028
                    source_parent_id = last_source_parent[1]
 
2029
                else:
 
2030
                    try:
 
2031
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2032
                    except KeyError:
 
2033
                        source_parent_entry = state._get_entry(source_index,
 
2034
                                                               path_utf8=old_dirname)
 
2035
                        source_parent_id = source_parent_entry[0][2]
 
2036
                    if source_parent_id == entry[0][2]:
 
2037
                        # This is the root, so the parent is None
 
2038
                        source_parent_id = None
 
2039
                    else:
 
2040
                        last_source_parent[0] = old_dirname
 
2041
                        last_source_parent[1] = source_parent_id
 
2042
                new_dirname = entry[0][0]
 
2043
                if new_dirname == last_target_parent[0]:
 
2044
                    target_parent_id = last_target_parent[1]
 
2045
                else:
 
2046
                    try:
 
2047
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2048
                    except KeyError:
 
2049
                        # TODO: We don't always need to do the lookup, because the
 
2050
                        #       parent entry will be the same as the source entry.
 
2051
                        target_parent_entry = state._get_entry(target_index,
 
2052
                                                               path_utf8=new_dirname)
 
2053
                        assert target_parent_entry != (None, None), (
 
2054
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2055
                            % (new_dirname, entry))
 
2056
                        target_parent_id = target_parent_entry[0][2]
 
2057
                    if target_parent_id == entry[0][2]:
 
2058
                        # This is the root, so the parent is None
 
2059
                        target_parent_id = None
 
2060
                    else:
 
2061
                        last_target_parent[0] = new_dirname
 
2062
                        last_target_parent[1] = target_parent_id
 
2063
 
 
2064
                source_exec = source_details[3]
 
2065
                if (include_unchanged
 
2066
                    or content_change
 
2067
                    or source_parent_id != target_parent_id
 
2068
                    or old_basename != entry[0][1]
 
2069
                    or source_exec != target_exec
 
2070
                    ):
 
2071
                    if old_path is None:
 
2072
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2073
                        old_path_u = utf8_decode(old_path)[0]
 
2074
                        path_u = old_path_u
 
2075
                    else:
 
2076
                        old_path_u = utf8_decode(old_path)[0]
 
2077
                        if old_path == path:
 
2078
                            path_u = old_path_u
 
2079
                        else:
 
2080
                            path_u = utf8_decode(path)[0]
 
2081
                    source_kind = _minikind_to_kind[source_minikind]
 
2082
                    return (entry[0][2],
 
2083
                           (old_path_u, path_u),
 
2084
                           content_change,
 
2085
                           (True, True),
 
2086
                           (source_parent_id, target_parent_id),
 
2087
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2088
                           (source_kind, target_kind),
 
2089
                           (source_exec, target_exec))
 
2090
                else:
 
2091
                    return uninteresting
 
2092
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2093
                # looks like a new file
 
2094
                if path_info is not None:
 
2095
                    path = pathjoin(entry[0][0], entry[0][1])
 
2096
                    # parent id is the entry for the path in the target tree
 
2097
                    # TODO: these are the same for an entire directory: cache em.
 
2098
                    parent_id = state._get_entry(target_index,
 
2099
                                                 path_utf8=entry[0][0])[0][2]
 
2100
                    if parent_id == entry[0][2]:
 
2101
                        parent_id = None
 
2102
                    if use_filesystem_for_exec:
 
2103
                        # We need S_ISREG here, because we aren't sure if this
 
2104
                        # is a file or not.
 
2105
                        target_exec = bool(
 
2106
                            stat.S_ISREG(path_info[3].st_mode)
 
2107
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2108
                    else:
 
2109
                        target_exec = target_details[3]
 
2110
                    return (entry[0][2],
 
2111
                           (None, utf8_decode(path)[0]),
 
2112
                           True,
 
2113
                           (False, True),
 
2114
                           (None, parent_id),
 
2115
                           (None, utf8_decode(entry[0][1])[0]),
 
2116
                           (None, path_info[2]),
 
2117
                           (None, target_exec))
 
2118
                else:
 
2119
                    # but its not on disk: we deliberately treat this as just
 
2120
                    # never-present. (Why ?! - RBC 20070224)
 
2121
                    pass
 
2122
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2123
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2124
                # if its still on disk, *and* theres no other entry at this
 
2125
                # path [we dont know this in this routine at the moment -
 
2126
                # perhaps we should change this - then it would be an unknown.
 
2127
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2128
                # parent id is the entry for the path in the target tree
 
2129
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2130
                if parent_id == entry[0][2]:
 
2131
                    parent_id = None
 
2132
                return (entry[0][2],
 
2133
                       (utf8_decode(old_path)[0], None),
 
2134
                       True,
 
2135
                       (True, False),
 
2136
                       (parent_id, None),
 
2137
                       (utf8_decode(entry[0][1])[0], None),
 
2138
                       (_minikind_to_kind[source_minikind], None),
 
2139
                       (source_details[3], None))
 
2140
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2141
                # a rename; could be a true rename, or a rename inherited from
 
2142
                # a renamed parent. TODO: handle this efficiently. Its not
 
2143
                # common case to rename dirs though, so a correct but slow
 
2144
                # implementation will do.
 
2145
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2146
                    search_specific_files.add(target_details[1])
 
2147
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2148
                # neither of the selected trees contain this file,
 
2149
                # so skip over it. This is not currently directly tested, but
 
2150
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2151
                pass
 
2152
            else:
 
2153
                raise AssertionError("don't know how to compare "
 
2154
                    "source_minikind=%r, target_minikind=%r"
 
2155
                    % (source_minikind, target_minikind))
 
2156
                ## import pdb;pdb.set_trace()
 
2157
            return None
 
2158
 
 
2159
        while search_specific_files:
 
2160
            # TODO: the pending list should be lexically sorted?  the
 
2161
            # interface doesn't require it.
 
2162
            current_root = search_specific_files.pop()
 
2163
            current_root_unicode = current_root.decode('utf8')
 
2164
            searched_specific_files.add(current_root)
 
2165
            # process the entries for this containing directory: the rest will be
 
2166
            # found by their parents recursively.
 
2167
            root_entries = _entries_for_path(current_root)
 
2168
            root_abspath = self.target.abspath(current_root_unicode)
 
2169
            try:
 
2170
                root_stat = os.lstat(root_abspath)
 
2171
            except OSError, e:
 
2172
                if e.errno == errno.ENOENT:
 
2173
                    # the path does not exist: let _process_entry know that.
 
2174
                    root_dir_info = None
 
2175
                else:
 
2176
                    # some other random error: hand it up.
 
2177
                    raise
 
2178
            else:
 
2179
                root_dir_info = ('', current_root,
 
2180
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2181
                    root_abspath)
 
2182
                if root_dir_info[2] == 'directory':
 
2183
                    if self.target._directory_is_tree_reference(
 
2184
                        current_root.decode('utf8')):
 
2185
                        root_dir_info = root_dir_info[:2] + \
 
2186
                            ('tree-reference',) + root_dir_info[3:]
 
2187
 
 
2188
            if not root_entries and not root_dir_info:
 
2189
                # this specified path is not present at all, skip it.
 
2190
                continue
 
2191
            path_handled = False
 
2192
            for entry in root_entries:
 
2193
                result = _process_entry(entry, root_dir_info)
 
2194
                if result is not None:
 
2195
                    path_handled = True
 
2196
                    if result is not uninteresting:
 
2197
                        yield result
 
2198
            if want_unversioned and not path_handled and root_dir_info:
 
2199
                new_executable = bool(
 
2200
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2201
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2202
                yield (None,
 
2203
                       (None, current_root_unicode),
 
2204
                       True,
 
2205
                       (False, False),
 
2206
                       (None, None),
 
2207
                       (None, splitpath(current_root_unicode)[-1]),
 
2208
                       (None, root_dir_info[2]),
 
2209
                       (None, new_executable)
 
2210
                      )
 
2211
            initial_key = (current_root, '', '')
 
2212
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2213
            if block_index == 0:
 
2214
                # we have processed the total root already, but because the
 
2215
                # initial key matched it we should skip it here.
 
2216
                block_index +=1
 
2217
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2218
                current_dir_info = None
 
2219
            else:
 
2220
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2221
                try:
 
2222
                    current_dir_info = dir_iterator.next()
 
2223
                except OSError, e:
 
2224
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2225
                    # python 2.5 has e.errno == EINVAL,
 
2226
                    #            and e.winerror == ERROR_DIRECTORY
 
2227
                    e_winerror = getattr(e, 'winerror', None)
 
2228
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2229
                    # there may be directories in the inventory even though
 
2230
                    # this path is not a file on disk: so mark it as end of
 
2231
                    # iterator
 
2232
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2233
                        current_dir_info = None
 
2234
                    elif (sys.platform == 'win32'
 
2235
                          and (e.errno in win_errors
 
2236
                               or e_winerror in win_errors)):
 
2237
                        current_dir_info = None
 
2238
                    else:
 
2239
                        raise
 
2240
                else:
 
2241
                    if current_dir_info[0][0] == '':
 
2242
                        # remove .bzr from iteration
 
2243
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2244
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2245
                        del current_dir_info[1][bzr_index]
 
2246
            # walk until both the directory listing and the versioned metadata
 
2247
            # are exhausted. 
 
2248
            if (block_index < len(state._dirblocks) and
 
2249
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2250
                current_block = state._dirblocks[block_index]
 
2251
            else:
 
2252
                current_block = None
 
2253
            while (current_dir_info is not None or
 
2254
                   current_block is not None):
 
2255
                if (current_dir_info and current_block
 
2256
                    and current_dir_info[0][0] != current_block[0]):
 
2257
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2258
                        # filesystem data refers to paths not covered by the dirblock.
 
2259
                        # this has two possibilities:
 
2260
                        # A) it is versioned but empty, so there is no block for it
 
2261
                        # B) it is not versioned.
 
2262
 
 
2263
                        # if (A) then we need to recurse into it to check for
 
2264
                        # new unknown files or directories.
 
2265
                        # if (B) then we should ignore it, because we don't
 
2266
                        # recurse into unknown directories.
 
2267
                        path_index = 0
 
2268
                        while path_index < len(current_dir_info[1]):
 
2269
                                current_path_info = current_dir_info[1][path_index]
 
2270
                                if want_unversioned:
 
2271
                                    if current_path_info[2] == 'directory':
 
2272
                                        if self.target._directory_is_tree_reference(
 
2273
                                            current_path_info[0].decode('utf8')):
 
2274
                                            current_path_info = current_path_info[:2] + \
 
2275
                                                ('tree-reference',) + current_path_info[3:]
 
2276
                                    new_executable = bool(
 
2277
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2278
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2279
                                    yield (None,
 
2280
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2281
                                        True,
 
2282
                                        (False, False),
 
2283
                                        (None, None),
 
2284
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2285
                                        (None, current_path_info[2]),
 
2286
                                        (None, new_executable))
 
2287
                                # dont descend into this unversioned path if it is
 
2288
                                # a dir
 
2289
                                if current_path_info[2] in ('directory',
 
2290
                                                            'tree-reference'):
 
2291
                                    del current_dir_info[1][path_index]
 
2292
                                    path_index -= 1
 
2293
                                path_index += 1
 
2294
 
 
2295
                        # This dir info has been handled, go to the next
 
2296
                        try:
 
2297
                            current_dir_info = dir_iterator.next()
 
2298
                        except StopIteration:
 
2299
                            current_dir_info = None
 
2300
                    else:
 
2301
                        # We have a dirblock entry for this location, but there
 
2302
                        # is no filesystem path for this. This is most likely
 
2303
                        # because a directory was removed from the disk.
 
2304
                        # We don't have to report the missing directory,
 
2305
                        # because that should have already been handled, but we
 
2306
                        # need to handle all of the files that are contained
 
2307
                        # within.
 
2308
                        for current_entry in current_block[1]:
 
2309
                            # entry referring to file not present on disk.
 
2310
                            # advance the entry only, after processing.
 
2311
                            result = _process_entry(current_entry, None)
 
2312
                            if result is not None:
 
2313
                                if result is not uninteresting:
 
2314
                                    yield result
 
2315
                        block_index +=1
 
2316
                        if (block_index < len(state._dirblocks) and
 
2317
                            osutils.is_inside(current_root,
 
2318
                                              state._dirblocks[block_index][0])):
 
2319
                            current_block = state._dirblocks[block_index]
 
2320
                        else:
 
2321
                            current_block = None
 
2322
                    continue
 
2323
                entry_index = 0
 
2324
                if current_block and entry_index < len(current_block[1]):
 
2325
                    current_entry = current_block[1][entry_index]
 
2326
                else:
 
2327
                    current_entry = None
 
2328
                advance_entry = True
 
2329
                path_index = 0
 
2330
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2331
                    current_path_info = current_dir_info[1][path_index]
 
2332
                    if current_path_info[2] == 'directory':
 
2333
                        if self.target._directory_is_tree_reference(
 
2334
                            current_path_info[0].decode('utf8')):
 
2335
                            current_path_info = current_path_info[:2] + \
 
2336
                                ('tree-reference',) + current_path_info[3:]
 
2337
                else:
 
2338
                    current_path_info = None
 
2339
                advance_path = True
 
2340
                path_handled = False
 
2341
                while (current_entry is not None or
 
2342
                    current_path_info is not None):
 
2343
                    if current_entry is None:
 
2344
                        # the check for path_handled when the path is adnvaced
 
2345
                        # will yield this path if needed.
 
2346
                        pass
 
2347
                    elif current_path_info is None:
 
2348
                        # no path is fine: the per entry code will handle it.
 
2349
                        result = _process_entry(current_entry, current_path_info)
 
2350
                        if result is not None:
 
2351
                            if result is not uninteresting:
 
2352
                                yield result
 
2353
                    elif (current_entry[0][1] != current_path_info[1]
 
2354
                          or current_entry[1][target_index][0] in 'ar'):
 
2355
                        # The current path on disk doesn't match the dirblock
 
2356
                        # record. Either the dirblock is marked as absent, or
 
2357
                        # the file on disk is not present at all in the
 
2358
                        # dirblock. Either way, report about the dirblock
 
2359
                        # entry, and let other code handle the filesystem one.
 
2360
 
 
2361
                        # Compare the basename for these files to determine
 
2362
                        # which comes first
 
2363
                        if current_path_info[1] < current_entry[0][1]:
 
2364
                            # extra file on disk: pass for now, but only
 
2365
                            # increment the path, not the entry
 
2366
                            advance_entry = False
 
2367
                        else:
 
2368
                            # entry referring to file not present on disk.
 
2369
                            # advance the entry only, after processing.
 
2370
                            result = _process_entry(current_entry, None)
 
2371
                            if result is not None:
 
2372
                                if result is not uninteresting:
 
2373
                                    yield result
 
2374
                            advance_path = False
 
2375
                    else:
 
2376
                        result = _process_entry(current_entry, current_path_info)
 
2377
                        if result is not None:
 
2378
                            path_handled = True
 
2379
                            if result is not uninteresting:
 
2380
                                yield result
 
2381
                    if advance_entry and current_entry is not None:
 
2382
                        entry_index += 1
 
2383
                        if entry_index < len(current_block[1]):
 
2384
                            current_entry = current_block[1][entry_index]
 
2385
                        else:
 
2386
                            current_entry = None
 
2387
                    else:
 
2388
                        advance_entry = True # reset the advance flaga
 
2389
                    if advance_path and current_path_info is not None:
 
2390
                        if not path_handled:
 
2391
                            # unversioned in all regards
 
2392
                            if want_unversioned:
 
2393
                                new_executable = bool(
 
2394
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2395
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2396
                                yield (None,
 
2397
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2398
                                    True,
 
2399
                                    (False, False),
 
2400
                                    (None, None),
 
2401
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2402
                                    (None, current_path_info[2]),
 
2403
                                    (None, new_executable))
 
2404
                            # dont descend into this unversioned path if it is
 
2405
                            # a dir
 
2406
                            if current_path_info[2] in ('directory'):
 
2407
                                del current_dir_info[1][path_index]
 
2408
                                path_index -= 1
 
2409
                        # dont descend the disk iterator into any tree 
 
2410
                        # paths.
 
2411
                        if current_path_info[2] == 'tree-reference':
 
2412
                            del current_dir_info[1][path_index]
 
2413
                            path_index -= 1
 
2414
                        path_index += 1
 
2415
                        if path_index < len(current_dir_info[1]):
 
2416
                            current_path_info = current_dir_info[1][path_index]
 
2417
                            if current_path_info[2] == 'directory':
 
2418
                                if self.target._directory_is_tree_reference(
 
2419
                                    current_path_info[0].decode('utf8')):
 
2420
                                    current_path_info = current_path_info[:2] + \
 
2421
                                        ('tree-reference',) + current_path_info[3:]
 
2422
                        else:
 
2423
                            current_path_info = None
 
2424
                        path_handled = False
 
2425
                    else:
 
2426
                        advance_path = True # reset the advance flagg.
 
2427
                if current_block is not None:
 
2428
                    block_index += 1
 
2429
                    if (block_index < len(state._dirblocks) and
 
2430
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2431
                        current_block = state._dirblocks[block_index]
 
2432
                    else:
 
2433
                        current_block = None
 
2434
                if current_dir_info is not None:
 
2435
                    try:
 
2436
                        current_dir_info = dir_iterator.next()
 
2437
                    except StopIteration:
 
2438
                        current_dir_info = None
 
2439
 
 
2440
 
 
2441
    @staticmethod
 
2442
    def is_compatible(source, target):
 
2443
        # the target must be a dirstate working tree
 
2444
        if not isinstance(target, WorkingTree4):
 
2445
            return False
 
2446
        # the source must be a revtreee or dirstate rev tree.
 
2447
        if not isinstance(source,
 
2448
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2449
            return False
 
2450
        # the source revid must be in the target dirstate
 
2451
        if not (source._revision_id == NULL_REVISION or
 
2452
            source._revision_id in target.get_parent_ids()):
 
2453
            # TODO: what about ghosts? it may well need to 
 
2454
            # check for them explicitly.
 
2455
            return False
 
2456
        return True
 
2457
 
 
2458
InterTree.register_optimiser(InterDirStateTree)
 
2459
 
 
2460
 
 
2461
class Converter3to4(object):
 
2462
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2463
 
 
2464
    def __init__(self):
 
2465
        self.target_format = WorkingTreeFormat4()
 
2466
 
 
2467
    def convert(self, tree):
 
2468
        # lock the control files not the tree, so that we dont get tree
 
2469
        # on-unlock behaviours, and so that noone else diddles with the 
 
2470
        # tree during upgrade.
 
2471
        tree._control_files.lock_write()
 
2472
        try:
 
2473
            tree.read_working_inventory()
 
2474
            self.create_dirstate_data(tree)
 
2475
            self.update_format(tree)
 
2476
            self.remove_xml_files(tree)
 
2477
        finally:
 
2478
            tree._control_files.unlock()
 
2479
 
 
2480
    def create_dirstate_data(self, tree):
 
2481
        """Create the dirstate based data for tree."""
 
2482
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2483
            ).local_abspath('dirstate')
 
2484
        state = dirstate.DirState.from_tree(tree, local_path)
 
2485
        state.save()
 
2486
        state.unlock()
 
2487
 
 
2488
    def remove_xml_files(self, tree):
 
2489
        """Remove the oldformat 3 data."""
 
2490
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2491
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2492
            'pending-merges', 'stat-cache']:
 
2493
            try:
 
2494
                transport.delete(path)
 
2495
            except errors.NoSuchFile:
 
2496
                # some files are optional - just deal.
 
2497
                pass
 
2498
 
 
2499
    def update_format(self, tree):
 
2500
        """Change the format marker."""
 
2501
        tree._control_files.put_utf8('format',
 
2502
            self.target_format.get_format_string())