/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

Deprecate LockableFiles.put_utf8

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._transport = self._control_files._transport
 
148
        self._dirty = None
 
149
        #-------------
 
150
        # during a read or write lock these objects are set, and are
 
151
        # None the rest of the time.
 
152
        self._dirstate = None
 
153
        self._inventory = None
 
154
        #-------------
 
155
        self._setup_directory_is_tree_reference()
 
156
        self._detect_case_handling()
 
157
 
 
158
    @needs_tree_write_lock
 
159
    def _add(self, files, ids, kinds):
 
160
        """See MutableTree._add."""
 
161
        state = self.current_dirstate()
 
162
        for f, file_id, kind in zip(files, ids, kinds):
 
163
            f = f.strip('/')
 
164
            if self.path2id(f):
 
165
                # special case tree root handling.
 
166
                if f == '' and self.path2id(f) == ROOT_ID:
 
167
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
168
                continue
 
169
            if file_id is None:
 
170
                file_id = generate_ids.gen_file_id(f)
 
171
            # deliberately add the file with no cached stat or sha1
 
172
            # - on the first access it will be gathered, and we can
 
173
            # always change this once tests are all passing.
 
174
            state.add(f, file_id, kind, None, '')
 
175
        self._make_dirty(reset_inventory=True)
 
176
 
 
177
    def _make_dirty(self, reset_inventory):
 
178
        """Make the tree state dirty.
 
179
 
 
180
        :param reset_inventory: True if the cached inventory should be removed
 
181
            (presuming there is one).
 
182
        """
 
183
        self._dirty = True
 
184
        if reset_inventory and self._inventory is not None:
 
185
            self._inventory = None
 
186
 
 
187
    @needs_tree_write_lock
 
188
    def add_reference(self, sub_tree):
 
189
        # use standard implementation, which calls back to self._add
 
190
        # 
 
191
        # So we don't store the reference_revision in the working dirstate,
 
192
        # it's just recorded at the moment of commit. 
 
193
        self._add_reference(sub_tree)
 
194
 
 
195
    def break_lock(self):
 
196
        """Break a lock if one is present from another instance.
 
197
 
 
198
        Uses the ui factory to ask for confirmation if the lock may be from
 
199
        an active process.
 
200
 
 
201
        This will probe the repository for its lock as well.
 
202
        """
 
203
        # if the dirstate is locked by an active process, reject the break lock
 
204
        # call.
 
205
        try:
 
206
            if self._dirstate is None:
 
207
                clear = True
 
208
            else:
 
209
                clear = False
 
210
            state = self._current_dirstate()
 
211
            if state._lock_token is not None:
 
212
                # we already have it locked. sheese, cant break our own lock.
 
213
                raise errors.LockActive(self.basedir)
 
214
            else:
 
215
                try:
 
216
                    # try for a write lock - need permission to get one anyhow
 
217
                    # to break locks.
 
218
                    state.lock_write()
 
219
                except errors.LockContention:
 
220
                    # oslocks fail when a process is still live: fail.
 
221
                    # TODO: get the locked lockdir info and give to the user to
 
222
                    # assist in debugging.
 
223
                    raise errors.LockActive(self.basedir)
 
224
                else:
 
225
                    state.unlock()
 
226
        finally:
 
227
            if clear:
 
228
                self._dirstate = None
 
229
        self._control_files.break_lock()
 
230
        self.branch.break_lock()
 
231
 
 
232
    def _comparison_data(self, entry, path):
 
233
        kind, executable, stat_value = \
 
234
            WorkingTree3._comparison_data(self, entry, path)
 
235
        # it looks like a plain directory, but it's really a reference -- see
 
236
        # also kind()
 
237
        if (self._repo_supports_tree_reference and
 
238
            kind == 'directory' and
 
239
            self._directory_is_tree_reference(path)):
 
240
            kind = 'tree-reference'
 
241
        return kind, executable, stat_value
 
242
 
 
243
    @needs_write_lock
 
244
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
245
        # mark the tree as dirty post commit - commit
 
246
        # can change the current versioned list by doing deletes.
 
247
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
248
        self._make_dirty(reset_inventory=True)
 
249
        return result
 
250
 
 
251
    def current_dirstate(self):
 
252
        """Return the current dirstate object.
 
253
 
 
254
        This is not part of the tree interface and only exposed for ease of
 
255
        testing.
 
256
 
 
257
        :raises errors.NotWriteLocked: when not in a lock.
 
258
        """
 
259
        self._must_be_locked()
 
260
        return self._current_dirstate()
 
261
 
 
262
    def _current_dirstate(self):
 
263
        """Internal function that does not check lock status.
 
264
 
 
265
        This is needed for break_lock which also needs the dirstate.
 
266
        """
 
267
        if self._dirstate is not None:
 
268
            return self._dirstate
 
269
        local_path = self.bzrdir.get_workingtree_transport(None
 
270
            ).local_abspath('dirstate')
 
271
        self._dirstate = dirstate.DirState.on_file(local_path)
 
272
        return self._dirstate
 
273
 
 
274
    def filter_unversioned_files(self, paths):
 
275
        """Filter out paths that are versioned.
 
276
 
 
277
        :return: set of paths.
 
278
        """
 
279
        # TODO: make a generic multi-bisect routine roughly that should list
 
280
        # the paths, then process one half at a time recursively, and feed the
 
281
        # results of each bisect in further still
 
282
        paths = sorted(paths)
 
283
        result = set()
 
284
        state = self.current_dirstate()
 
285
        # TODO we want a paths_to_dirblocks helper I think
 
286
        for path in paths:
 
287
            dirname, basename = os.path.split(path.encode('utf8'))
 
288
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
289
                dirname, basename, 0)
 
290
            if not path_is_versioned:
 
291
                result.add(path)
 
292
        return result
 
293
 
 
294
    def flush(self):
 
295
        """Write all cached data to disk."""
 
296
        if self._control_files._lock_mode != 'w':
 
297
            raise errors.NotWriteLocked(self)
 
298
        self.current_dirstate().save()
 
299
        self._inventory = None
 
300
        self._dirty = False
 
301
 
 
302
    @needs_tree_write_lock
 
303
    def _gather_kinds(self, files, kinds):
 
304
        """See MutableTree._gather_kinds."""
 
305
        for pos, f in enumerate(files):
 
306
            if kinds[pos] is None:
 
307
                kinds[pos] = self._kind(f)
 
308
 
 
309
    def _generate_inventory(self):
 
310
        """Create and set self.inventory from the dirstate object.
 
311
        
 
312
        This is relatively expensive: we have to walk the entire dirstate.
 
313
        Ideally we would not, and can deprecate this function.
 
314
        """
 
315
        #: uncomment to trap on inventory requests.
 
316
        # import pdb;pdb.set_trace()
 
317
        state = self.current_dirstate()
 
318
        state._read_dirblocks_if_needed()
 
319
        root_key, current_entry = self._get_entry(path='')
 
320
        current_id = root_key[2]
 
321
        assert current_entry[0][0] == 'd' # directory
 
322
        inv = Inventory(root_id=current_id)
 
323
        # Turn some things into local variables
 
324
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
325
        factory = entry_factory
 
326
        utf8_decode = cache_utf8._utf8_decode
 
327
        inv_byid = inv._byid
 
328
        # we could do this straight out of the dirstate; it might be fast
 
329
        # and should be profiled - RBC 20070216
 
330
        parent_ies = {'' : inv.root}
 
331
        for block in state._dirblocks[1:]: # skip the root
 
332
            dirname = block[0]
 
333
            try:
 
334
                parent_ie = parent_ies[dirname]
 
335
            except KeyError:
 
336
                # all the paths in this block are not versioned in this tree
 
337
                continue
 
338
            for key, entry in block[1]:
 
339
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
340
                if minikind in ('a', 'r'): # absent, relocated
 
341
                    # a parent tree only entry
 
342
                    continue
 
343
                name = key[1]
 
344
                name_unicode = utf8_decode(name)[0]
 
345
                file_id = key[2]
 
346
                kind = minikind_to_kind[minikind]
 
347
                inv_entry = factory[kind](file_id, name_unicode,
 
348
                                          parent_ie.file_id)
 
349
                if kind == 'file':
 
350
                    # This is only needed on win32, where this is the only way
 
351
                    # we know the executable bit.
 
352
                    inv_entry.executable = executable
 
353
                    # not strictly needed: working tree
 
354
                    #inv_entry.text_size = size
 
355
                    #inv_entry.text_sha1 = sha1
 
356
                elif kind == 'directory':
 
357
                    # add this entry to the parent map.
 
358
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
359
                elif kind == 'tree-reference':
 
360
                    assert self._repo_supports_tree_reference, \
 
361
                        "repository of %r " \
 
362
                        "doesn't support tree references " \
 
363
                        "required by entry %r" \
 
364
                        % (self, name)
 
365
                    inv_entry.reference_revision = link_or_sha1 or None
 
366
                elif kind != 'symlink':
 
367
                    raise AssertionError("unknown kind %r" % kind)
 
368
                # These checks cost us around 40ms on a 55k entry tree
 
369
                assert file_id not in inv_byid, ('file_id %s already in'
 
370
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
371
                assert name_unicode not in parent_ie.children
 
372
                inv_byid[file_id] = inv_entry
 
373
                parent_ie.children[name_unicode] = inv_entry
 
374
        self._inventory = inv
 
375
 
 
376
    def _get_entry(self, file_id=None, path=None):
 
377
        """Get the dirstate row for file_id or path.
 
378
 
 
379
        If either file_id or path is supplied, it is used as the key to lookup.
 
380
        If both are supplied, the fastest lookup is used, and an error is
 
381
        raised if they do not both point at the same row.
 
382
        
 
383
        :param file_id: An optional unicode file_id to be looked up.
 
384
        :param path: An optional unicode path to be looked up.
 
385
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
386
        """
 
387
        if file_id is None and path is None:
 
388
            raise errors.BzrError('must supply file_id or path')
 
389
        state = self.current_dirstate()
 
390
        if path is not None:
 
391
            path = path.encode('utf8')
 
392
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
393
 
 
394
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
395
        # check file id is valid unconditionally.
 
396
        entry = self._get_entry(file_id=file_id, path=path)
 
397
        if entry[0] is None:
 
398
            raise errors.NoSuchId(self, file_id)
 
399
        if path is None:
 
400
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
401
 
 
402
        file_abspath = self.abspath(path)
 
403
        state = self.current_dirstate()
 
404
        if stat_value is None:
 
405
            try:
 
406
                stat_value = os.lstat(file_abspath)
 
407
            except OSError, e:
 
408
                if e.errno == errno.ENOENT:
 
409
                    return None
 
410
                else:
 
411
                    raise
 
412
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
413
                                          stat_value=stat_value)
 
414
        if entry[1][0][0] == 'f':
 
415
            return link_or_sha1
 
416
        return None
 
417
 
 
418
    def _get_inventory(self):
 
419
        """Get the inventory for the tree. This is only valid within a lock."""
 
420
        if 'evil' in debug.debug_flags:
 
421
            trace.mutter_callsite(2,
 
422
                "accessing .inventory forces a size of tree translation.")
 
423
        if self._inventory is not None:
 
424
            return self._inventory
 
425
        self._must_be_locked()
 
426
        self._generate_inventory()
 
427
        return self._inventory
 
428
 
 
429
    inventory = property(_get_inventory,
 
430
                         doc="Inventory of this Tree")
 
431
 
 
432
    @needs_read_lock
 
433
    def get_parent_ids(self):
 
434
        """See Tree.get_parent_ids.
 
435
        
 
436
        This implementation requests the ids list from the dirstate file.
 
437
        """
 
438
        return self.current_dirstate().get_parent_ids()
 
439
 
 
440
    def get_reference_revision(self, file_id, path=None):
 
441
        # referenced tree's revision is whatever's currently there
 
442
        return self.get_nested_tree(file_id, path).last_revision()
 
443
 
 
444
    def get_nested_tree(self, file_id, path=None):
 
445
        if path is None:
 
446
            path = self.id2path(file_id)
 
447
        # else: check file_id is at path?
 
448
        return WorkingTree.open(self.abspath(path))
 
449
 
 
450
    @needs_read_lock
 
451
    def get_root_id(self):
 
452
        """Return the id of this trees root"""
 
453
        return self._get_entry(path='')[0][2]
 
454
 
 
455
    def has_id(self, file_id):
 
456
        state = self.current_dirstate()
 
457
        row, parents = self._get_entry(file_id=file_id)
 
458
        if row is None:
 
459
            return False
 
460
        return osutils.lexists(pathjoin(
 
461
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
462
 
 
463
    @needs_read_lock
 
464
    def id2path(self, file_id):
 
465
        "Convert a file-id to a path."
 
466
        state = self.current_dirstate()
 
467
        entry = self._get_entry(file_id=file_id)
 
468
        if entry == (None, None):
 
469
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
470
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
471
        return path_utf8.decode('utf8')
 
472
 
 
473
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
474
        entry = self._get_entry(path=path)
 
475
        if entry == (None, None):
 
476
            return False # Missing entries are not executable
 
477
        return entry[1][0][3] # Executable?
 
478
 
 
479
    if not osutils.supports_executable():
 
480
        def is_executable(self, file_id, path=None):
 
481
            """Test if a file is executable or not.
 
482
 
 
483
            Note: The caller is expected to take a read-lock before calling this.
 
484
            """
 
485
            entry = self._get_entry(file_id=file_id, path=path)
 
486
            if entry == (None, None):
 
487
                return False
 
488
            return entry[1][0][3]
 
489
 
 
490
        _is_executable_from_path_and_stat = \
 
491
            _is_executable_from_path_and_stat_from_basis
 
492
    else:
 
493
        def is_executable(self, file_id, path=None):
 
494
            """Test if a file is executable or not.
 
495
 
 
496
            Note: The caller is expected to take a read-lock before calling this.
 
497
            """
 
498
            self._must_be_locked()
 
499
            if not path:
 
500
                path = self.id2path(file_id)
 
501
            mode = os.lstat(self.abspath(path)).st_mode
 
502
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
503
 
 
504
    def all_file_ids(self):
 
505
        """See Tree.iter_all_file_ids"""
 
506
        self._must_be_locked()
 
507
        result = set()
 
508
        for key, tree_details in self.current_dirstate()._iter_entries():
 
509
            if tree_details[0][0] in ('a', 'r'): # relocated
 
510
                continue
 
511
            result.add(key[2])
 
512
        return result
 
513
 
 
514
    @needs_read_lock
 
515
    def __iter__(self):
 
516
        """Iterate through file_ids for this tree.
 
517
 
 
518
        file_ids are in a WorkingTree if they are in the working inventory
 
519
        and the working file exists.
 
520
        """
 
521
        result = []
 
522
        for key, tree_details in self.current_dirstate()._iter_entries():
 
523
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
524
                # not relevant to the working tree
 
525
                continue
 
526
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
527
            if osutils.lexists(path):
 
528
                result.append(key[2])
 
529
        return iter(result)
 
530
 
 
531
    def iter_references(self):
 
532
        for key, tree_details in self.current_dirstate()._iter_entries():
 
533
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
534
                # not relevant to the working tree
 
535
                continue
 
536
            if not key[1]:
 
537
                # the root is not a reference.
 
538
                continue
 
539
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
540
            try:
 
541
                if self._kind(path) == 'tree-reference':
 
542
                    yield path, key[2]
 
543
            except errors.NoSuchFile:
 
544
                # path is missing on disk.
 
545
                continue
 
546
 
 
547
    def kind(self, file_id):
 
548
        """Return the kind of a file.
 
549
 
 
550
        This is always the actual kind that's on disk, regardless of what it
 
551
        was added as.
 
552
 
 
553
        Note: The caller is expected to take a read-lock before calling this.
 
554
        """
 
555
        relpath = self.id2path(file_id)
 
556
        assert relpath != None, \
 
557
            "path for id {%s} is None!" % file_id
 
558
        return self._kind(relpath)
 
559
 
 
560
    def _kind(self, relpath):
 
561
        abspath = self.abspath(relpath)
 
562
        kind = file_kind(abspath)
 
563
        if (self._repo_supports_tree_reference and
 
564
            kind == 'directory' and
 
565
            self._directory_is_tree_reference(relpath)):
 
566
            kind = 'tree-reference'
 
567
        return kind
 
568
 
 
569
    @needs_read_lock
 
570
    def _last_revision(self):
 
571
        """See Mutable.last_revision."""
 
572
        parent_ids = self.current_dirstate().get_parent_ids()
 
573
        if parent_ids:
 
574
            return parent_ids[0]
 
575
        else:
 
576
            return _mod_revision.NULL_REVISION
 
577
 
 
578
    def lock_read(self):
 
579
        """See Branch.lock_read, and WorkingTree.unlock."""
 
580
        self.branch.lock_read()
 
581
        try:
 
582
            self._control_files.lock_read()
 
583
            try:
 
584
                state = self.current_dirstate()
 
585
                if not state._lock_token:
 
586
                    state.lock_read()
 
587
                # set our support for tree references from the repository in
 
588
                # use.
 
589
                self._repo_supports_tree_reference = getattr(
 
590
                    self.branch.repository._format, "supports_tree_reference",
 
591
                    False)
 
592
            except:
 
593
                self._control_files.unlock()
 
594
                raise
 
595
        except:
 
596
            self.branch.unlock()
 
597
            raise
 
598
 
 
599
    def _lock_self_write(self):
 
600
        """This should be called after the branch is locked."""
 
601
        try:
 
602
            self._control_files.lock_write()
 
603
            try:
 
604
                state = self.current_dirstate()
 
605
                if not state._lock_token:
 
606
                    state.lock_write()
 
607
                # set our support for tree references from the repository in
 
608
                # use.
 
609
                self._repo_supports_tree_reference = getattr(
 
610
                    self.branch.repository._format, "supports_tree_reference",
 
611
                    False)
 
612
            except:
 
613
                self._control_files.unlock()
 
614
                raise
 
615
        except:
 
616
            self.branch.unlock()
 
617
            raise
 
618
 
 
619
    def lock_tree_write(self):
 
620
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
621
        self.branch.lock_read()
 
622
        self._lock_self_write()
 
623
 
 
624
    def lock_write(self):
 
625
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
626
        self.branch.lock_write()
 
627
        self._lock_self_write()
 
628
 
 
629
    @needs_tree_write_lock
 
630
    def move(self, from_paths, to_dir, after=False):
 
631
        """See WorkingTree.move()."""
 
632
        result = []
 
633
        if not from_paths:
 
634
            return result
 
635
 
 
636
        state = self.current_dirstate()
 
637
 
 
638
        assert not isinstance(from_paths, basestring)
 
639
        to_dir_utf8 = to_dir.encode('utf8')
 
640
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
641
        id_index = state._get_id_index()
 
642
        # check destination directory
 
643
        # get the details for it
 
644
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
645
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
646
        if not entry_present:
 
647
            raise errors.BzrMoveFailedError('', to_dir,
 
648
                errors.NotVersionedError(to_dir))
 
649
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
650
        # get a handle on the block itself.
 
651
        to_block_index = state._ensure_block(
 
652
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
653
        to_block = state._dirblocks[to_block_index]
 
654
        to_abs = self.abspath(to_dir)
 
655
        if not isdir(to_abs):
 
656
            raise errors.BzrMoveFailedError('',to_dir,
 
657
                errors.NotADirectory(to_abs))
 
658
 
 
659
        if to_entry[1][0][0] != 'd':
 
660
            raise errors.BzrMoveFailedError('',to_dir,
 
661
                errors.NotADirectory(to_abs))
 
662
 
 
663
        if self._inventory is not None:
 
664
            update_inventory = True
 
665
            inv = self.inventory
 
666
            to_dir_id = to_entry[0][2]
 
667
            to_dir_ie = inv[to_dir_id]
 
668
        else:
 
669
            update_inventory = False
 
670
 
 
671
        rollbacks = []
 
672
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
673
                     fingerprint, packed_stat, size,
 
674
                     to_block, to_key, to_path_utf8):
 
675
            state._make_absent(old_entry)
 
676
            from_key = old_entry[0]
 
677
            rollbacks.append(
 
678
                lambda:state.update_minimal(from_key,
 
679
                    minikind,
 
680
                    executable=executable,
 
681
                    fingerprint=fingerprint,
 
682
                    packed_stat=packed_stat,
 
683
                    size=size,
 
684
                    path_utf8=from_path_utf8))
 
685
            state.update_minimal(to_key,
 
686
                    minikind,
 
687
                    executable=executable,
 
688
                    fingerprint=fingerprint,
 
689
                    packed_stat=packed_stat,
 
690
                    size=size,
 
691
                    path_utf8=to_path_utf8)
 
692
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
693
            new_entry = to_block[1][added_entry_index]
 
694
            rollbacks.append(lambda:state._make_absent(new_entry))
 
695
 
 
696
        for from_rel in from_paths:
 
697
            # from_rel is 'pathinroot/foo/bar'
 
698
            from_rel_utf8 = from_rel.encode('utf8')
 
699
            from_dirname, from_tail = osutils.split(from_rel)
 
700
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
701
            from_entry = self._get_entry(path=from_rel)
 
702
            if from_entry == (None, None):
 
703
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
704
                    errors.NotVersionedError(path=str(from_rel)))
 
705
 
 
706
            from_id = from_entry[0][2]
 
707
            to_rel = pathjoin(to_dir, from_tail)
 
708
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
709
            item_to_entry = self._get_entry(path=to_rel)
 
710
            if item_to_entry != (None, None):
 
711
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
712
                    "Target is already versioned.")
 
713
 
 
714
            if from_rel == to_rel:
 
715
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
716
                    "Source and target are identical.")
 
717
 
 
718
            from_missing = not self.has_filename(from_rel)
 
719
            to_missing = not self.has_filename(to_rel)
 
720
            if after:
 
721
                move_file = False
 
722
            else:
 
723
                move_file = True
 
724
            if to_missing:
 
725
                if not move_file:
 
726
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
727
                        errors.NoSuchFile(path=to_rel,
 
728
                        extra="New file has not been created yet"))
 
729
                elif from_missing:
 
730
                    # neither path exists
 
731
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
732
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
733
            else:
 
734
                if from_missing: # implicitly just update our path mapping
 
735
                    move_file = False
 
736
                elif not after:
 
737
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
738
 
 
739
            rollbacks = []
 
740
            def rollback_rename():
 
741
                """A single rename has failed, roll it back."""
 
742
                # roll back everything, even if we encounter trouble doing one
 
743
                # of them.
 
744
                #
 
745
                # TODO: at least log the other exceptions rather than just
 
746
                # losing them mbp 20070307
 
747
                exc_info = None
 
748
                for rollback in reversed(rollbacks):
 
749
                    try:
 
750
                        rollback()
 
751
                    except Exception, e:
 
752
                        exc_info = sys.exc_info()
 
753
                if exc_info:
 
754
                    raise exc_info[0], exc_info[1], exc_info[2]
 
755
 
 
756
            # perform the disk move first - its the most likely failure point.
 
757
            if move_file:
 
758
                from_rel_abs = self.abspath(from_rel)
 
759
                to_rel_abs = self.abspath(to_rel)
 
760
                try:
 
761
                    osutils.rename(from_rel_abs, to_rel_abs)
 
762
                except OSError, e:
 
763
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
764
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
765
            try:
 
766
                # perform the rename in the inventory next if needed: its easy
 
767
                # to rollback
 
768
                if update_inventory:
 
769
                    # rename the entry
 
770
                    from_entry = inv[from_id]
 
771
                    current_parent = from_entry.parent_id
 
772
                    inv.rename(from_id, to_dir_id, from_tail)
 
773
                    rollbacks.append(
 
774
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
775
                # finally do the rename in the dirstate, which is a little
 
776
                # tricky to rollback, but least likely to need it.
 
777
                old_block_index, old_entry_index, dir_present, file_present = \
 
778
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
779
                old_block = state._dirblocks[old_block_index][1]
 
780
                old_entry = old_block[old_entry_index]
 
781
                from_key, old_entry_details = old_entry
 
782
                cur_details = old_entry_details[0]
 
783
                # remove the old row
 
784
                to_key = ((to_block[0],) + from_key[1:3])
 
785
                minikind = cur_details[0]
 
786
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
787
                         minikind=minikind,
 
788
                         executable=cur_details[3],
 
789
                         fingerprint=cur_details[1],
 
790
                         packed_stat=cur_details[4],
 
791
                         size=cur_details[2],
 
792
                         to_block=to_block,
 
793
                         to_key=to_key,
 
794
                         to_path_utf8=to_rel_utf8)
 
795
 
 
796
                if minikind == 'd':
 
797
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
798
                        """Recursively update all entries in this dirblock."""
 
799
                        assert from_dir != '', "renaming root not supported"
 
800
                        from_key = (from_dir, '')
 
801
                        from_block_idx, present = \
 
802
                            state._find_block_index_from_key(from_key)
 
803
                        if not present:
 
804
                            # This is the old record, if it isn't present, then
 
805
                            # there is theoretically nothing to update.
 
806
                            # (Unless it isn't present because of lazy loading,
 
807
                            # but we don't do that yet)
 
808
                            return
 
809
                        from_block = state._dirblocks[from_block_idx]
 
810
                        to_block_index, to_entry_index, _, _ = \
 
811
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
812
                        to_block_index = state._ensure_block(
 
813
                            to_block_index, to_entry_index, to_dir_utf8)
 
814
                        to_block = state._dirblocks[to_block_index]
 
815
 
 
816
                        # Grab a copy since move_one may update the list.
 
817
                        for entry in from_block[1][:]:
 
818
                            assert entry[0][0] == from_dir
 
819
                            cur_details = entry[1][0]
 
820
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
821
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
822
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
823
                            minikind = cur_details[0]
 
824
                            if minikind in 'ar':
 
825
                                # Deleted children of a renamed directory
 
826
                                # Do not need to be updated.
 
827
                                # Children that have been renamed out of this
 
828
                                # directory should also not be updated
 
829
                                continue
 
830
                            move_one(entry, from_path_utf8=from_path_utf8,
 
831
                                     minikind=minikind,
 
832
                                     executable=cur_details[3],
 
833
                                     fingerprint=cur_details[1],
 
834
                                     packed_stat=cur_details[4],
 
835
                                     size=cur_details[2],
 
836
                                     to_block=to_block,
 
837
                                     to_key=to_key,
 
838
                                     to_path_utf8=to_path_utf8)
 
839
                            if minikind == 'd':
 
840
                                # We need to move all the children of this
 
841
                                # entry
 
842
                                update_dirblock(from_path_utf8, to_key,
 
843
                                                to_path_utf8)
 
844
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
845
            except:
 
846
                rollback_rename()
 
847
                raise
 
848
            result.append((from_rel, to_rel))
 
849
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
850
            self._make_dirty(reset_inventory=False)
 
851
 
 
852
        return result
 
853
 
 
854
    def _must_be_locked(self):
 
855
        if not self._control_files._lock_count:
 
856
            raise errors.ObjectNotLocked(self)
 
857
 
 
858
    def _new_tree(self):
 
859
        """Initialize the state in this tree to be a new tree."""
 
860
        self._dirty = True
 
861
 
 
862
    @needs_read_lock
 
863
    def path2id(self, path):
 
864
        """Return the id for path in this tree."""
 
865
        path = path.strip('/')
 
866
        entry = self._get_entry(path=path)
 
867
        if entry == (None, None):
 
868
            return None
 
869
        return entry[0][2]
 
870
 
 
871
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
872
        """See Tree.paths2ids().
 
873
 
 
874
        This specialisation fast-paths the case where all the trees are in the
 
875
        dirstate.
 
876
        """
 
877
        if paths is None:
 
878
            return None
 
879
        parents = self.get_parent_ids()
 
880
        for tree in trees:
 
881
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
882
                parents):
 
883
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
884
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
885
        # -- make all paths utf8 --
 
886
        paths_utf8 = set()
 
887
        for path in paths:
 
888
            paths_utf8.add(path.encode('utf8'))
 
889
        paths = paths_utf8
 
890
        # -- paths is now a utf8 path set --
 
891
        # -- get the state object and prepare it.
 
892
        state = self.current_dirstate()
 
893
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
894
            and '' not in paths):
 
895
            paths2ids = self._paths2ids_using_bisect
 
896
        else:
 
897
            paths2ids = self._paths2ids_in_memory
 
898
        return paths2ids(paths, search_indexes,
 
899
                         require_versioned=require_versioned)
 
900
 
 
901
    def _paths2ids_in_memory(self, paths, search_indexes,
 
902
                             require_versioned=True):
 
903
        state = self.current_dirstate()
 
904
        state._read_dirblocks_if_needed()
 
905
        def _entries_for_path(path):
 
906
            """Return a list with all the entries that match path for all ids.
 
907
            """
 
908
            dirname, basename = os.path.split(path)
 
909
            key = (dirname, basename, '')
 
910
            block_index, present = state._find_block_index_from_key(key)
 
911
            if not present:
 
912
                # the block which should contain path is absent.
 
913
                return []
 
914
            result = []
 
915
            block = state._dirblocks[block_index][1]
 
916
            entry_index, _ = state._find_entry_index(key, block)
 
917
            # we may need to look at multiple entries at this path: walk while the paths match.
 
918
            while (entry_index < len(block) and
 
919
                block[entry_index][0][0:2] == key[0:2]):
 
920
                result.append(block[entry_index])
 
921
                entry_index += 1
 
922
            return result
 
923
        if require_versioned:
 
924
            # -- check all supplied paths are versioned in a search tree. --
 
925
            all_versioned = True
 
926
            for path in paths:
 
927
                path_entries = _entries_for_path(path)
 
928
                if not path_entries:
 
929
                    # this specified path is not present at all: error
 
930
                    all_versioned = False
 
931
                    break
 
932
                found_versioned = False
 
933
                # for each id at this path
 
934
                for entry in path_entries:
 
935
                    # for each tree.
 
936
                    for index in search_indexes:
 
937
                        if entry[1][index][0] != 'a': # absent
 
938
                            found_versioned = True
 
939
                            # all good: found a versioned cell
 
940
                            break
 
941
                if not found_versioned:
 
942
                    # none of the indexes was not 'absent' at all ids for this
 
943
                    # path.
 
944
                    all_versioned = False
 
945
                    break
 
946
            if not all_versioned:
 
947
                raise errors.PathsNotVersionedError(paths)
 
948
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
949
        search_paths = osutils.minimum_path_selection(paths)
 
950
        # sketch: 
 
951
        # for all search_indexs in each path at or under each element of
 
952
        # search_paths, if the detail is relocated: add the id, and add the
 
953
        # relocated path as one to search if its not searched already. If the
 
954
        # detail is not relocated, add the id.
 
955
        searched_paths = set()
 
956
        found_ids = set()
 
957
        def _process_entry(entry):
 
958
            """Look at search_indexes within entry.
 
959
 
 
960
            If a specific tree's details are relocated, add the relocation
 
961
            target to search_paths if not searched already. If it is absent, do
 
962
            nothing. Otherwise add the id to found_ids.
 
963
            """
 
964
            for index in search_indexes:
 
965
                if entry[1][index][0] == 'r': # relocated
 
966
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
967
                        search_paths.add(entry[1][index][1])
 
968
                elif entry[1][index][0] != 'a': # absent
 
969
                    found_ids.add(entry[0][2])
 
970
        while search_paths:
 
971
            current_root = search_paths.pop()
 
972
            searched_paths.add(current_root)
 
973
            # process the entries for this containing directory: the rest will be
 
974
            # found by their parents recursively.
 
975
            root_entries = _entries_for_path(current_root)
 
976
            if not root_entries:
 
977
                # this specified path is not present at all, skip it.
 
978
                continue
 
979
            for entry in root_entries:
 
980
                _process_entry(entry)
 
981
            initial_key = (current_root, '', '')
 
982
            block_index, _ = state._find_block_index_from_key(initial_key)
 
983
            while (block_index < len(state._dirblocks) and
 
984
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
985
                for entry in state._dirblocks[block_index][1]:
 
986
                    _process_entry(entry)
 
987
                block_index += 1
 
988
        return found_ids
 
989
 
 
990
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
991
                                require_versioned=True):
 
992
        state = self.current_dirstate()
 
993
        found_ids = set()
 
994
 
 
995
        split_paths = sorted(osutils.split(p) for p in paths)
 
996
        found = state._bisect_recursive(split_paths)
 
997
 
 
998
        if require_versioned:
 
999
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
1000
            for dir_name in split_paths:
 
1001
                if dir_name not in found_dir_names:
 
1002
                    raise errors.PathsNotVersionedError(paths)
 
1003
 
 
1004
        for dir_name_id, trees_info in found.iteritems():
 
1005
            for index in search_indexes:
 
1006
                if trees_info[index][0] not in ('r', 'a'):
 
1007
                    found_ids.add(dir_name_id[2])
 
1008
        return found_ids
 
1009
 
 
1010
    def read_working_inventory(self):
 
1011
        """Read the working inventory.
 
1012
        
 
1013
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1014
        """
 
1015
        return self.inventory
 
1016
 
 
1017
    @needs_read_lock
 
1018
    def revision_tree(self, revision_id):
 
1019
        """See Tree.revision_tree.
 
1020
 
 
1021
        WorkingTree4 supplies revision_trees for any basis tree.
 
1022
        """
 
1023
        dirstate = self.current_dirstate()
 
1024
        parent_ids = dirstate.get_parent_ids()
 
1025
        if revision_id not in parent_ids:
 
1026
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1027
        if revision_id in dirstate.get_ghosts():
 
1028
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1029
        return DirStateRevisionTree(dirstate, revision_id,
 
1030
            self.branch.repository)
 
1031
 
 
1032
    @needs_tree_write_lock
 
1033
    def set_last_revision(self, new_revision):
 
1034
        """Change the last revision in the working tree."""
 
1035
        parents = self.get_parent_ids()
 
1036
        if new_revision in (NULL_REVISION, None):
 
1037
            assert len(parents) < 2, (
 
1038
                "setting the last parent to none with a pending merge is "
 
1039
                "unsupported.")
 
1040
            self.set_parent_ids([])
 
1041
        else:
 
1042
            self.set_parent_ids([new_revision] + parents[1:],
 
1043
                allow_leftmost_as_ghost=True)
 
1044
 
 
1045
    @needs_tree_write_lock
 
1046
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1047
        """Set the parent ids to revision_ids.
 
1048
        
 
1049
        See also set_parent_trees. This api will try to retrieve the tree data
 
1050
        for each element of revision_ids from the trees repository. If you have
 
1051
        tree data already available, it is more efficient to use
 
1052
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1053
        an easier API to use.
 
1054
 
 
1055
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1056
            working tree. Any of these may be ghosts.
 
1057
        """
 
1058
        trees = []
 
1059
        for revision_id in revision_ids:
 
1060
            try:
 
1061
                revtree = self.branch.repository.revision_tree(revision_id)
 
1062
                # TODO: jam 20070213 KnitVersionedFile raises
 
1063
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1064
                #       given revision_id is not present. Should Repository be
 
1065
                #       catching it and re-raising NoSuchRevision?
 
1066
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1067
                revtree = None
 
1068
            trees.append((revision_id, revtree))
 
1069
        self.set_parent_trees(trees,
 
1070
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1071
 
 
1072
    @needs_tree_write_lock
 
1073
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1074
        """Set the parents of the working tree.
 
1075
 
 
1076
        :param parents_list: A list of (revision_id, tree) tuples.
 
1077
            If tree is None, then that element is treated as an unreachable
 
1078
            parent tree - i.e. a ghost.
 
1079
        """
 
1080
        dirstate = self.current_dirstate()
 
1081
        if len(parents_list) > 0:
 
1082
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1083
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1084
        real_trees = []
 
1085
        ghosts = []
 
1086
        # convert absent trees to the null tree, which we convert back to
 
1087
        # missing on access.
 
1088
        for rev_id, tree in parents_list:
 
1089
            _mod_revision.check_not_reserved_id(rev_id)
 
1090
            if tree is not None:
 
1091
                real_trees.append((rev_id, tree))
 
1092
            else:
 
1093
                real_trees.append((rev_id,
 
1094
                    self.branch.repository.revision_tree(None)))
 
1095
                ghosts.append(rev_id)
 
1096
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1097
        self._make_dirty(reset_inventory=False)
 
1098
 
 
1099
    def _set_root_id(self, file_id):
 
1100
        """See WorkingTree.set_root_id."""
 
1101
        state = self.current_dirstate()
 
1102
        state.set_path_id('', file_id)
 
1103
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1104
            self._make_dirty(reset_inventory=True)
 
1105
 
 
1106
    def _sha_from_stat(self, path, stat_result):
 
1107
        """Get a sha digest from the tree's stat cache.
 
1108
 
 
1109
        The default implementation assumes no stat cache is present.
 
1110
 
 
1111
        :param path: The path.
 
1112
        :param stat_result: The stat result being looked up.
 
1113
        """
 
1114
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1115
 
 
1116
    @needs_read_lock
 
1117
    def supports_tree_reference(self):
 
1118
        return self._repo_supports_tree_reference
 
1119
 
 
1120
    def unlock(self):
 
1121
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1122
        # do non-implementation specific cleanup
 
1123
        self._cleanup()
 
1124
 
 
1125
        if self._control_files._lock_count == 1:
 
1126
            # eventually we should do signature checking during read locks for
 
1127
            # dirstate updates.
 
1128
            if self._control_files._lock_mode == 'w':
 
1129
                if self._dirty:
 
1130
                    self.flush()
 
1131
            if self._dirstate is not None:
 
1132
                # This is a no-op if there are no modifications.
 
1133
                self._dirstate.save()
 
1134
                self._dirstate.unlock()
 
1135
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1136
            #       point. Instead, it could check if the header has been
 
1137
            #       modified when it is locked, and if not, it can hang on to
 
1138
            #       the data it has in memory.
 
1139
            self._dirstate = None
 
1140
            self._inventory = None
 
1141
        # reverse order of locking.
 
1142
        try:
 
1143
            return self._control_files.unlock()
 
1144
        finally:
 
1145
            self.branch.unlock()
 
1146
 
 
1147
    @needs_tree_write_lock
 
1148
    def unversion(self, file_ids):
 
1149
        """Remove the file ids in file_ids from the current versioned set.
 
1150
 
 
1151
        When a file_id is unversioned, all of its children are automatically
 
1152
        unversioned.
 
1153
 
 
1154
        :param file_ids: The file ids to stop versioning.
 
1155
        :raises: NoSuchId if any fileid is not currently versioned.
 
1156
        """
 
1157
        if not file_ids:
 
1158
            return
 
1159
        state = self.current_dirstate()
 
1160
        state._read_dirblocks_if_needed()
 
1161
        ids_to_unversion = set(file_ids)
 
1162
        paths_to_unversion = set()
 
1163
        # sketch:
 
1164
        # check if the root is to be unversioned, if so, assert for now.
 
1165
        # walk the state marking unversioned things as absent.
 
1166
        # if there are any un-unversioned ids at the end, raise
 
1167
        for key, details in state._dirblocks[0][1]:
 
1168
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1169
                key[2] in ids_to_unversion):
 
1170
                # I haven't written the code to unversion / yet - it should be
 
1171
                # supported.
 
1172
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1173
        block_index = 0
 
1174
        while block_index < len(state._dirblocks):
 
1175
            # process one directory at a time.
 
1176
            block = state._dirblocks[block_index]
 
1177
            # first check: is the path one to remove - it or its children
 
1178
            delete_block = False
 
1179
            for path in paths_to_unversion:
 
1180
                if (block[0].startswith(path) and
 
1181
                    (len(block[0]) == len(path) or
 
1182
                     block[0][len(path)] == '/')):
 
1183
                    # this entire block should be deleted - its the block for a
 
1184
                    # path to unversion; or the child of one
 
1185
                    delete_block = True
 
1186
                    break
 
1187
            # TODO: trim paths_to_unversion as we pass by paths
 
1188
            if delete_block:
 
1189
                # this block is to be deleted: process it.
 
1190
                # TODO: we can special case the no-parents case and
 
1191
                # just forget the whole block.
 
1192
                entry_index = 0
 
1193
                while entry_index < len(block[1]):
 
1194
                    # Mark this file id as having been removed
 
1195
                    entry = block[1][entry_index]
 
1196
                    ids_to_unversion.discard(entry[0][2])
 
1197
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1198
                                               # entries
 
1199
                        or not state._make_absent(entry)):
 
1200
                        entry_index += 1
 
1201
                # go to the next block. (At the moment we dont delete empty
 
1202
                # dirblocks)
 
1203
                block_index += 1
 
1204
                continue
 
1205
            entry_index = 0
 
1206
            while entry_index < len(block[1]):
 
1207
                entry = block[1][entry_index]
 
1208
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1209
                    # ^ some parent row.
 
1210
                    entry[0][2] not in ids_to_unversion):
 
1211
                    # ^ not an id to unversion
 
1212
                    entry_index += 1
 
1213
                    continue
 
1214
                if entry[1][0][0] == 'd':
 
1215
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1216
                if not state._make_absent(entry):
 
1217
                    entry_index += 1
 
1218
                # we have unversioned this id
 
1219
                ids_to_unversion.remove(entry[0][2])
 
1220
            block_index += 1
 
1221
        if ids_to_unversion:
 
1222
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1223
        self._make_dirty(reset_inventory=False)
 
1224
        # have to change the legacy inventory too.
 
1225
        if self._inventory is not None:
 
1226
            for file_id in file_ids:
 
1227
                self._inventory.remove_recursive_id(file_id)
 
1228
 
 
1229
    @needs_tree_write_lock
 
1230
    def rename_one(self, from_rel, to_rel, after=False):
 
1231
        """See WorkingTree.rename_one"""
 
1232
        self.flush()
 
1233
        WorkingTree.rename_one(self, from_rel, to_rel, after)
 
1234
 
 
1235
    @needs_tree_write_lock
 
1236
    def apply_inventory_delta(self, changes):
 
1237
        """See MutableTree.apply_inventory_delta"""
 
1238
        state = self.current_dirstate()
 
1239
        state.update_by_delta(changes)
 
1240
        self._make_dirty(reset_inventory=True)
 
1241
 
 
1242
    def update_basis_by_delta(self, new_revid, delta):
 
1243
        """See MutableTree.update_basis_by_delta."""
 
1244
        assert self.last_revision() != new_revid
 
1245
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1246
 
 
1247
    @needs_read_lock
 
1248
    def _validate(self):
 
1249
        self._dirstate._validate()
 
1250
 
 
1251
    @needs_tree_write_lock
 
1252
    def _write_inventory(self, inv):
 
1253
        """Write inventory as the current inventory."""
 
1254
        assert not self._dirty, ("attempting to write an inventory when the "
 
1255
            "dirstate is dirty will cause data loss")
 
1256
        self.current_dirstate().set_state_from_inventory(inv)
 
1257
        self._make_dirty(reset_inventory=False)
 
1258
        if self._inventory is not None:
 
1259
            self._inventory = inv
 
1260
        self.flush()
 
1261
 
 
1262
 
 
1263
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1264
    """The first consolidated dirstate working tree format.
 
1265
 
 
1266
    This format:
 
1267
        - exists within a metadir controlling .bzr
 
1268
        - includes an explicit version marker for the workingtree control
 
1269
          files, separate from the BzrDir format
 
1270
        - modifies the hash cache format
 
1271
        - is new in bzr 0.15
 
1272
        - uses a LockDir to guard access to it.
 
1273
    """
 
1274
 
 
1275
    upgrade_recommended = False
 
1276
 
 
1277
    def get_format_string(self):
 
1278
        """See WorkingTreeFormat.get_format_string()."""
 
1279
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1280
 
 
1281
    def get_format_description(self):
 
1282
        """See WorkingTreeFormat.get_format_description()."""
 
1283
        return "Working tree format 4"
 
1284
 
 
1285
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1286
                   accelerator_tree=None, hardlink=False):
 
1287
        """See WorkingTreeFormat.initialize().
 
1288
 
 
1289
        :param revision_id: allows creating a working tree at a different
 
1290
        revision than the branch is at.
 
1291
        :param accelerator_tree: A tree which can be used for retrieving file
 
1292
            contents more quickly than the revision tree, i.e. a workingtree.
 
1293
            The revision tree will be used for cases where accelerator_tree's
 
1294
            content is different.
 
1295
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1296
            where possible.
 
1297
 
 
1298
        These trees get an initial random root id, if their repository supports
 
1299
        rich root data, TREE_ROOT otherwise.
 
1300
        """
 
1301
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1302
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1303
        transport = a_bzrdir.get_workingtree_transport(self)
 
1304
        control_files = self._open_control_files(a_bzrdir)
 
1305
        control_files.create_lock()
 
1306
        control_files.lock_write()
 
1307
        transport.put_bytes('format', self.get_format_string(),
 
1308
            mode=control_files._file_mode)
 
1309
        if from_branch is not None:
 
1310
            branch = from_branch
 
1311
        else:
 
1312
            branch = a_bzrdir.open_branch()
 
1313
        if revision_id is None:
 
1314
            revision_id = branch.last_revision()
 
1315
        local_path = transport.local_abspath('dirstate')
 
1316
        # write out new dirstate (must exist when we create the tree)
 
1317
        state = dirstate.DirState.initialize(local_path)
 
1318
        state.unlock()
 
1319
        del state
 
1320
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1321
                         branch,
 
1322
                         _format=self,
 
1323
                         _bzrdir=a_bzrdir,
 
1324
                         _control_files=control_files)
 
1325
        wt._new_tree()
 
1326
        wt.lock_tree_write()
 
1327
        try:
 
1328
            if revision_id in (None, NULL_REVISION):
 
1329
                if branch.repository.supports_rich_root():
 
1330
                    wt._set_root_id(generate_ids.gen_root_id())
 
1331
                else:
 
1332
                    wt._set_root_id(ROOT_ID)
 
1333
                wt.flush()
 
1334
            basis = None
 
1335
            # frequently, we will get here due to branching.  The accelerator
 
1336
            # tree will be the tree from the branch, so the desired basis
 
1337
            # tree will often be a parent of the accelerator tree.
 
1338
            if accelerator_tree is not None:
 
1339
                try:
 
1340
                    basis = accelerator_tree.revision_tree(revision_id)
 
1341
                except errors.NoSuchRevision:
 
1342
                    pass
 
1343
            if basis is None:
 
1344
                basis = branch.repository.revision_tree(revision_id)
 
1345
            if revision_id == NULL_REVISION:
 
1346
                parents_list = []
 
1347
            else:
 
1348
                parents_list = [(revision_id, basis)]
 
1349
            basis.lock_read()
 
1350
            try:
 
1351
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1352
                wt.flush()
 
1353
                # if the basis has a root id we have to use that; otherwise we
 
1354
                # use a new random one
 
1355
                basis_root_id = basis.get_root_id()
 
1356
                if basis_root_id is not None:
 
1357
                    wt._set_root_id(basis_root_id)
 
1358
                    wt.flush()
 
1359
                transform.build_tree(basis, wt, accelerator_tree,
 
1360
                                     hardlink=hardlink)
 
1361
            finally:
 
1362
                basis.unlock()
 
1363
        finally:
 
1364
            control_files.unlock()
 
1365
            wt.unlock()
 
1366
        return wt
 
1367
 
 
1368
    def _open(self, a_bzrdir, control_files):
 
1369
        """Open the tree itself.
 
1370
 
 
1371
        :param a_bzrdir: the dir for the tree.
 
1372
        :param control_files: the control files for the tree.
 
1373
        """
 
1374
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1375
                           branch=a_bzrdir.open_branch(),
 
1376
                           _format=self,
 
1377
                           _bzrdir=a_bzrdir,
 
1378
                           _control_files=control_files)
 
1379
 
 
1380
    def __get_matchingbzrdir(self):
 
1381
        # please test against something that will let us do tree references
 
1382
        return bzrdir.format_registry.make_bzrdir(
 
1383
            'dirstate-with-subtree')
 
1384
 
 
1385
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1386
 
 
1387
 
 
1388
class DirStateRevisionTree(Tree):
 
1389
    """A revision tree pulling the inventory from a dirstate."""
 
1390
 
 
1391
    def __init__(self, dirstate, revision_id, repository):
 
1392
        self._dirstate = dirstate
 
1393
        self._revision_id = revision_id
 
1394
        self._repository = repository
 
1395
        self._inventory = None
 
1396
        self._locked = 0
 
1397
        self._dirstate_locked = False
 
1398
 
 
1399
    def __repr__(self):
 
1400
        return "<%s of %s in %s>" % \
 
1401
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1402
 
 
1403
    def annotate_iter(self, file_id,
 
1404
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1405
        """See Tree.annotate_iter"""
 
1406
        w = self._get_weave(file_id)
 
1407
        return w.annotate(self.inventory[file_id].revision)
 
1408
 
 
1409
    def _get_ancestors(self, default_revision):
 
1410
        return set(self._repository.get_ancestry(self._revision_id,
 
1411
                                                 topo_sorted=False))
 
1412
    def _comparison_data(self, entry, path):
 
1413
        """See Tree._comparison_data."""
 
1414
        if entry is None:
 
1415
            return None, False, None
 
1416
        # trust the entry as RevisionTree does, but this may not be
 
1417
        # sensible: the entry might not have come from us?
 
1418
        return entry.kind, entry.executable, None
 
1419
 
 
1420
    def _file_size(self, entry, stat_value):
 
1421
        return entry.text_size
 
1422
 
 
1423
    def filter_unversioned_files(self, paths):
 
1424
        """Filter out paths that are not versioned.
 
1425
 
 
1426
        :return: set of paths.
 
1427
        """
 
1428
        pred = self.has_filename
 
1429
        return set((p for p in paths if not pred(p)))
 
1430
 
 
1431
    def get_root_id(self):
 
1432
        return self.path2id('')
 
1433
 
 
1434
    def id2path(self, file_id):
 
1435
        "Convert a file-id to a path."
 
1436
        entry = self._get_entry(file_id=file_id)
 
1437
        if entry == (None, None):
 
1438
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1439
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1440
        return path_utf8.decode('utf8')
 
1441
 
 
1442
    def _get_parent_index(self):
 
1443
        """Return the index in the dirstate referenced by this tree."""
 
1444
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1445
 
 
1446
    def _get_entry(self, file_id=None, path=None):
 
1447
        """Get the dirstate row for file_id or path.
 
1448
 
 
1449
        If either file_id or path is supplied, it is used as the key to lookup.
 
1450
        If both are supplied, the fastest lookup is used, and an error is
 
1451
        raised if they do not both point at the same row.
 
1452
        
 
1453
        :param file_id: An optional unicode file_id to be looked up.
 
1454
        :param path: An optional unicode path to be looked up.
 
1455
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1456
        """
 
1457
        if file_id is None and path is None:
 
1458
            raise errors.BzrError('must supply file_id or path')
 
1459
        if path is not None:
 
1460
            path = path.encode('utf8')
 
1461
        parent_index = self._get_parent_index()
 
1462
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1463
 
 
1464
    def _generate_inventory(self):
 
1465
        """Create and set self.inventory from the dirstate object.
 
1466
 
 
1467
        (So this is only called the first time the inventory is requested for
 
1468
        this tree; it then remains in memory until it's out of date.)
 
1469
 
 
1470
        This is relatively expensive: we have to walk the entire dirstate.
 
1471
        """
 
1472
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1473
            'dirstate revision tree'
 
1474
        # separate call for profiling - makes it clear where the costs are.
 
1475
        self._dirstate._read_dirblocks_if_needed()
 
1476
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1477
            'parent %s has disappeared from %s' % (
 
1478
            self._revision_id, self._dirstate.get_parent_ids())
 
1479
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1480
        # This is identical now to the WorkingTree _generate_inventory except
 
1481
        # for the tree index use.
 
1482
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1483
        current_id = root_key[2]
 
1484
        assert current_entry[parent_index][0] == 'd'
 
1485
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1486
        inv.root.revision = current_entry[parent_index][4]
 
1487
        # Turn some things into local variables
 
1488
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1489
        factory = entry_factory
 
1490
        utf8_decode = cache_utf8._utf8_decode
 
1491
        inv_byid = inv._byid
 
1492
        # we could do this straight out of the dirstate; it might be fast
 
1493
        # and should be profiled - RBC 20070216
 
1494
        parent_ies = {'' : inv.root}
 
1495
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1496
            dirname = block[0]
 
1497
            try:
 
1498
                parent_ie = parent_ies[dirname]
 
1499
            except KeyError:
 
1500
                # all the paths in this block are not versioned in this tree
 
1501
                continue
 
1502
            for key, entry in block[1]:
 
1503
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1504
                if minikind in ('a', 'r'): # absent, relocated
 
1505
                    # not this tree
 
1506
                    continue
 
1507
                name = key[1]
 
1508
                name_unicode = utf8_decode(name)[0]
 
1509
                file_id = key[2]
 
1510
                kind = minikind_to_kind[minikind]
 
1511
                inv_entry = factory[kind](file_id, name_unicode,
 
1512
                                          parent_ie.file_id)
 
1513
                inv_entry.revision = revid
 
1514
                if kind == 'file':
 
1515
                    inv_entry.executable = executable
 
1516
                    inv_entry.text_size = size
 
1517
                    inv_entry.text_sha1 = fingerprint
 
1518
                elif kind == 'directory':
 
1519
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1520
                elif kind == 'symlink':
 
1521
                    inv_entry.executable = False
 
1522
                    inv_entry.text_size = None
 
1523
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1524
                elif kind == 'tree-reference':
 
1525
                    inv_entry.reference_revision = fingerprint or None
 
1526
                else:
 
1527
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1528
                            % entry)
 
1529
                # These checks cost us around 40ms on a 55k entry tree
 
1530
                assert file_id not in inv_byid
 
1531
                assert name_unicode not in parent_ie.children
 
1532
                inv_byid[file_id] = inv_entry
 
1533
                parent_ie.children[name_unicode] = inv_entry
 
1534
        self._inventory = inv
 
1535
 
 
1536
    def get_file_mtime(self, file_id, path=None):
 
1537
        """Return the modification time for this record.
 
1538
 
 
1539
        We return the timestamp of the last-changed revision.
 
1540
        """
 
1541
        # Make sure the file exists
 
1542
        entry = self._get_entry(file_id, path=path)
 
1543
        if entry == (None, None): # do we raise?
 
1544
            return None
 
1545
        parent_index = self._get_parent_index()
 
1546
        last_changed_revision = entry[1][parent_index][4]
 
1547
        return self._repository.get_revision(last_changed_revision).timestamp
 
1548
 
 
1549
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1550
        entry = self._get_entry(file_id=file_id, path=path)
 
1551
        parent_index = self._get_parent_index()
 
1552
        parent_details = entry[1][parent_index]
 
1553
        if parent_details[0] == 'f':
 
1554
            return parent_details[1]
 
1555
        return None
 
1556
 
 
1557
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1558
    def get_weave(self, file_id):
 
1559
        return self._get_weave(file_id)
 
1560
 
 
1561
    def _get_weave(self, file_id):
 
1562
        return self._repository.weave_store.get_weave(file_id,
 
1563
                self._repository.get_transaction())
 
1564
 
 
1565
    def get_file(self, file_id, path=None):
 
1566
        return StringIO(self.get_file_text(file_id))
 
1567
 
 
1568
    def get_file_lines(self, file_id):
 
1569
        entry = self._get_entry(file_id=file_id)[1]
 
1570
        if entry == None:
 
1571
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1572
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1573
 
 
1574
    def get_file_size(self, file_id):
 
1575
        """See Tree.get_file_size"""
 
1576
        return self.inventory[file_id].text_size
 
1577
 
 
1578
    def get_file_text(self, file_id):
 
1579
        return ''.join(self.get_file_lines(file_id))
 
1580
 
 
1581
    def get_reference_revision(self, file_id, path=None):
 
1582
        return self.inventory[file_id].reference_revision
 
1583
 
 
1584
    def iter_files_bytes(self, desired_files):
 
1585
        """See Tree.iter_files_bytes.
 
1586
 
 
1587
        This version is implemented on top of Repository.iter_files_bytes"""
 
1588
        parent_index = self._get_parent_index()
 
1589
        repo_desired_files = []
 
1590
        for file_id, identifier in desired_files:
 
1591
            entry = self._get_entry(file_id)
 
1592
            if entry == (None, None):
 
1593
                raise errors.NoSuchId(self, file_id)
 
1594
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1595
                                       identifier))
 
1596
        return self._repository.iter_files_bytes(repo_desired_files)
 
1597
 
 
1598
    def get_symlink_target(self, file_id):
 
1599
        entry = self._get_entry(file_id=file_id)
 
1600
        parent_index = self._get_parent_index()
 
1601
        if entry[1][parent_index][0] != 'l':
 
1602
            return None
 
1603
        else:
 
1604
            # At present, none of the tree implementations supports non-ascii
 
1605
            # symlink targets. So we will just assume that the dirstate path is
 
1606
            # correct.
 
1607
            return entry[1][parent_index][1]
 
1608
 
 
1609
    def get_revision_id(self):
 
1610
        """Return the revision id for this tree."""
 
1611
        return self._revision_id
 
1612
 
 
1613
    def _get_inventory(self):
 
1614
        if self._inventory is not None:
 
1615
            return self._inventory
 
1616
        self._must_be_locked()
 
1617
        self._generate_inventory()
 
1618
        return self._inventory
 
1619
 
 
1620
    inventory = property(_get_inventory,
 
1621
                         doc="Inventory of this Tree")
 
1622
 
 
1623
    def get_parent_ids(self):
 
1624
        """The parents of a tree in the dirstate are not cached."""
 
1625
        return self._repository.get_revision(self._revision_id).parent_ids
 
1626
 
 
1627
    def has_filename(self, filename):
 
1628
        return bool(self.path2id(filename))
 
1629
 
 
1630
    def kind(self, file_id):
 
1631
        entry = self._get_entry(file_id=file_id)[1]
 
1632
        if entry == None:
 
1633
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1634
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1635
 
 
1636
    def stored_kind(self, file_id):
 
1637
        """See Tree.stored_kind"""
 
1638
        return self.kind(file_id)
 
1639
 
 
1640
    def path_content_summary(self, path):
 
1641
        """See Tree.path_content_summary."""
 
1642
        id = self.inventory.path2id(path)
 
1643
        if id is None:
 
1644
            return ('missing', None, None, None)
 
1645
        entry = self._inventory[id]
 
1646
        kind = entry.kind
 
1647
        if kind == 'file':
 
1648
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1649
        elif kind == 'symlink':
 
1650
            return (kind, None, None, entry.symlink_target)
 
1651
        else:
 
1652
            return (kind, None, None, None)
 
1653
 
 
1654
    def is_executable(self, file_id, path=None):
 
1655
        ie = self.inventory[file_id]
 
1656
        if ie.kind != "file":
 
1657
            return None
 
1658
        return ie.executable
 
1659
 
 
1660
    def list_files(self, include_root=False):
 
1661
        # We use a standard implementation, because DirStateRevisionTree is
 
1662
        # dealing with one of the parents of the current state
 
1663
        inv = self._get_inventory()
 
1664
        entries = inv.iter_entries()
 
1665
        if self.inventory.root is not None and not include_root:
 
1666
            entries.next()
 
1667
        for path, entry in entries:
 
1668
            yield path, 'V', entry.kind, entry.file_id, entry
 
1669
 
 
1670
    def lock_read(self):
 
1671
        """Lock the tree for a set of operations."""
 
1672
        if not self._locked:
 
1673
            self._repository.lock_read()
 
1674
            if self._dirstate._lock_token is None:
 
1675
                self._dirstate.lock_read()
 
1676
                self._dirstate_locked = True
 
1677
        self._locked += 1
 
1678
 
 
1679
    def _must_be_locked(self):
 
1680
        if not self._locked:
 
1681
            raise errors.ObjectNotLocked(self)
 
1682
 
 
1683
    @needs_read_lock
 
1684
    def path2id(self, path):
 
1685
        """Return the id for path in this tree."""
 
1686
        # lookup by path: faster than splitting and walking the ivnentory.
 
1687
        entry = self._get_entry(path=path)
 
1688
        if entry == (None, None):
 
1689
            return None
 
1690
        return entry[0][2]
 
1691
 
 
1692
    def unlock(self):
 
1693
        """Unlock, freeing any cache memory used during the lock."""
 
1694
        # outside of a lock, the inventory is suspect: release it.
 
1695
        self._locked -=1
 
1696
        if not self._locked:
 
1697
            self._inventory = None
 
1698
            self._locked = 0
 
1699
            if self._dirstate_locked:
 
1700
                self._dirstate.unlock()
 
1701
                self._dirstate_locked = False
 
1702
            self._repository.unlock()
 
1703
 
 
1704
    def walkdirs(self, prefix=""):
 
1705
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1706
        # implementation based on an inventory.  
 
1707
        # This should be cleaned up to use the much faster Dirstate code
 
1708
        # So for now, we just build up the parent inventory, and extract
 
1709
        # it the same way RevisionTree does.
 
1710
        _directory = 'directory'
 
1711
        inv = self._get_inventory()
 
1712
        top_id = inv.path2id(prefix)
 
1713
        if top_id is None:
 
1714
            pending = []
 
1715
        else:
 
1716
            pending = [(prefix, top_id)]
 
1717
        while pending:
 
1718
            dirblock = []
 
1719
            relpath, file_id = pending.pop()
 
1720
            # 0 - relpath, 1- file-id
 
1721
            if relpath:
 
1722
                relroot = relpath + '/'
 
1723
            else:
 
1724
                relroot = ""
 
1725
            # FIXME: stash the node in pending
 
1726
            entry = inv[file_id]
 
1727
            for name, child in entry.sorted_children():
 
1728
                toppath = relroot + name
 
1729
                dirblock.append((toppath, name, child.kind, None,
 
1730
                    child.file_id, child.kind
 
1731
                    ))
 
1732
            yield (relpath, entry.file_id), dirblock
 
1733
            # push the user specified dirs from dirblock
 
1734
            for dir in reversed(dirblock):
 
1735
                if dir[2] == _directory:
 
1736
                    pending.append((dir[0], dir[4]))
 
1737
 
 
1738
 
 
1739
class InterDirStateTree(InterTree):
 
1740
    """Fast path optimiser for changes_from with dirstate trees.
 
1741
    
 
1742
    This is used only when both trees are in the dirstate working file, and 
 
1743
    the source is any parent within the dirstate, and the destination is 
 
1744
    the current working tree of the same dirstate.
 
1745
    """
 
1746
    # this could be generalized to allow comparisons between any trees in the
 
1747
    # dirstate, and possibly between trees stored in different dirstates.
 
1748
 
 
1749
    def __init__(self, source, target):
 
1750
        super(InterDirStateTree, self).__init__(source, target)
 
1751
        if not InterDirStateTree.is_compatible(source, target):
 
1752
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1753
 
 
1754
    @staticmethod
 
1755
    def make_source_parent_tree(source, target):
 
1756
        """Change the source tree into a parent of the target."""
 
1757
        revid = source.commit('record tree')
 
1758
        target.branch.repository.fetch(source.branch.repository, revid)
 
1759
        target.set_parent_ids([revid])
 
1760
        return target.basis_tree(), target
 
1761
 
 
1762
    _matching_from_tree_format = WorkingTreeFormat4()
 
1763
    _matching_to_tree_format = WorkingTreeFormat4()
 
1764
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1765
 
 
1766
    def iter_changes(self, include_unchanged=False,
 
1767
                      specific_files=None, pb=None, extra_trees=[],
 
1768
                      require_versioned=True, want_unversioned=False):
 
1769
        """Return the changes from source to target.
 
1770
 
 
1771
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1772
            for details.
 
1773
        :param specific_files: An optional list of file paths to restrict the
 
1774
            comparison to. When mapping filenames to ids, all matches in all
 
1775
            trees (including optional extra_trees) are used, and all children of
 
1776
            matched directories are included.
 
1777
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1778
            unchanged entries in the result.
 
1779
        :param extra_trees: An optional list of additional trees to use when
 
1780
            mapping the contents of specific_files (paths) to file_ids.
 
1781
        :param require_versioned: If True, all files in specific_files must be
 
1782
            versioned in one of source, target, extra_trees or
 
1783
            PathsNotVersionedError is raised.
 
1784
        :param want_unversioned: Should unversioned files be returned in the
 
1785
            output. An unversioned file is defined as one with (False, False)
 
1786
            for the versioned pair.
 
1787
        """
 
1788
        utf8_decode = cache_utf8._utf8_decode
 
1789
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1790
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1791
        # NB: show_status depends on being able to pass in non-versioned files
 
1792
        # and report them as unknown
 
1793
        # TODO: handle extra trees in the dirstate.
 
1794
        if (extra_trees or specific_files == []):
 
1795
            # we can't fast-path these cases (yet)
 
1796
            for f in super(InterDirStateTree, self).iter_changes(
 
1797
                include_unchanged, specific_files, pb, extra_trees,
 
1798
                require_versioned, want_unversioned=want_unversioned):
 
1799
                yield f
 
1800
            return
 
1801
        parent_ids = self.target.get_parent_ids()
 
1802
        assert (self.source._revision_id in parent_ids
 
1803
                or self.source._revision_id == NULL_REVISION), \
 
1804
                "revision {%s} is not stored in {%s}, but %s " \
 
1805
                "can only be used for trees stored in the dirstate" \
 
1806
                % (self.source._revision_id, self.target, self.iter_changes)
 
1807
        target_index = 0
 
1808
        if self.source._revision_id == NULL_REVISION:
 
1809
            source_index = None
 
1810
            indices = (target_index,)
 
1811
        else:
 
1812
            assert (self.source._revision_id in parent_ids), \
 
1813
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1814
                self.source._revision_id, parent_ids)
 
1815
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1816
            indices = (source_index, target_index)
 
1817
        # -- make all specific_files utf8 --
 
1818
        if specific_files:
 
1819
            specific_files_utf8 = set()
 
1820
            for path in specific_files:
 
1821
                specific_files_utf8.add(path.encode('utf8'))
 
1822
            specific_files = specific_files_utf8
 
1823
        else:
 
1824
            specific_files = set([''])
 
1825
        # -- specific_files is now a utf8 path set --
 
1826
        # -- get the state object and prepare it.
 
1827
        state = self.target.current_dirstate()
 
1828
        state._read_dirblocks_if_needed()
 
1829
        def _entries_for_path(path):
 
1830
            """Return a list with all the entries that match path for all ids.
 
1831
            """
 
1832
            dirname, basename = os.path.split(path)
 
1833
            key = (dirname, basename, '')
 
1834
            block_index, present = state._find_block_index_from_key(key)
 
1835
            if not present:
 
1836
                # the block which should contain path is absent.
 
1837
                return []
 
1838
            result = []
 
1839
            block = state._dirblocks[block_index][1]
 
1840
            entry_index, _ = state._find_entry_index(key, block)
 
1841
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1842
            while (entry_index < len(block) and
 
1843
                block[entry_index][0][0:2] == key[0:2]):
 
1844
                result.append(block[entry_index])
 
1845
                entry_index += 1
 
1846
            return result
 
1847
        if require_versioned:
 
1848
            # -- check all supplied paths are versioned in a search tree. --
 
1849
            all_versioned = True
 
1850
            for path in specific_files:
 
1851
                path_entries = _entries_for_path(path)
 
1852
                if not path_entries:
 
1853
                    # this specified path is not present at all: error
 
1854
                    all_versioned = False
 
1855
                    break
 
1856
                found_versioned = False
 
1857
                # for each id at this path
 
1858
                for entry in path_entries:
 
1859
                    # for each tree.
 
1860
                    for index in indices:
 
1861
                        if entry[1][index][0] != 'a': # absent
 
1862
                            found_versioned = True
 
1863
                            # all good: found a versioned cell
 
1864
                            break
 
1865
                if not found_versioned:
 
1866
                    # none of the indexes was not 'absent' at all ids for this
 
1867
                    # path.
 
1868
                    all_versioned = False
 
1869
                    break
 
1870
            if not all_versioned:
 
1871
                raise errors.PathsNotVersionedError(specific_files)
 
1872
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1873
        search_specific_files = set()
 
1874
        for path in specific_files:
 
1875
            other_specific_files = specific_files.difference(set([path]))
 
1876
            if not osutils.is_inside_any(other_specific_files, path):
 
1877
                # this is a top level path, we must check it.
 
1878
                search_specific_files.add(path)
 
1879
        # sketch: 
 
1880
        # compare source_index and target_index at or under each element of search_specific_files.
 
1881
        # follow the following comparison table. Note that we only want to do diff operations when
 
1882
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1883
        # for the target.
 
1884
        # cases:
 
1885
        # 
 
1886
        # Source | Target | disk | action
 
1887
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1888
        #        |        |      | diff check on source-target
 
1889
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1890
        #        |        |      | ???
 
1891
        #   r    |  a     |      | add source to search
 
1892
        #   r    |  a     |  a   | 
 
1893
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1894
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1895
        #   a    | fdlt   |      | add new id
 
1896
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1897
        #   a    |  a     |      | not present in either tree, skip
 
1898
        #   a    |  a     |  a   | not present in any tree, skip
 
1899
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1900
        #        |        |      | may not be selected by the users list of paths.
 
1901
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1902
        #        |        |      | may not be selected by the users list of paths.
 
1903
        #  fdlt  | fdlt   |      | content in both: diff them
 
1904
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1905
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1906
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1907
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1908
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1909
        #        |        |      | this id at the other path.
 
1910
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1911
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1912
        #        |        |      | this id at the other path.
 
1913
 
 
1914
        # for all search_indexs in each path at or under each element of
 
1915
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1916
        # relocated path as one to search if its not searched already. If the
 
1917
        # detail is not relocated, add the id.
 
1918
        searched_specific_files = set()
 
1919
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1920
        # Using a list so that we can access the values and change them in
 
1921
        # nested scope. Each one is [path, file_id, entry]
 
1922
        last_source_parent = [None, None]
 
1923
        last_target_parent = [None, None]
 
1924
 
 
1925
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1926
 
 
1927
        # Just a sentry, so that _process_entry can say that this
 
1928
        # record is handled, but isn't interesting to process (unchanged)
 
1929
        uninteresting = object()
 
1930
 
 
1931
 
 
1932
        old_dirname_to_file_id = {}
 
1933
        new_dirname_to_file_id = {}
 
1934
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1935
        #       keeping a cache of directories that we have seen.
 
1936
 
 
1937
        def _process_entry(entry, path_info):
 
1938
            """Compare an entry and real disk to generate delta information.
 
1939
 
 
1940
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1941
                the path of entry. If None, then the path is considered absent.
 
1942
                (Perhaps we should pass in a concrete entry for this ?)
 
1943
                Basename is returned as a utf8 string because we expect this
 
1944
                tuple will be ignored, and don't want to take the time to
 
1945
                decode.
 
1946
            :return: None if these don't match
 
1947
                     A tuple of information about the change, or
 
1948
                     the object 'uninteresting' if these match, but are
 
1949
                     basically identical.
 
1950
            """
 
1951
            if source_index is None:
 
1952
                source_details = NULL_PARENT_DETAILS
 
1953
            else:
 
1954
                source_details = entry[1][source_index]
 
1955
            target_details = entry[1][target_index]
 
1956
            target_minikind = target_details[0]
 
1957
            if path_info is not None and target_minikind in 'fdlt':
 
1958
                assert target_index == 0
 
1959
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1960
                                                  stat_value=path_info[3])
 
1961
                # The entry may have been modified by update_entry
 
1962
                target_details = entry[1][target_index]
 
1963
                target_minikind = target_details[0]
 
1964
            else:
 
1965
                link_or_sha1 = None
 
1966
            file_id = entry[0][2]
 
1967
            source_minikind = source_details[0]
 
1968
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1969
                # claimed content in both: diff
 
1970
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1971
                #        |        |      | diff check on source-target
 
1972
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1973
                #        |        |      | ???
 
1974
                if source_minikind in 'r':
 
1975
                    # add the source to the search path to find any children it
 
1976
                    # has.  TODO ? : only add if it is a container ?
 
1977
                    if not osutils.is_inside_any(searched_specific_files,
 
1978
                                                 source_details[1]):
 
1979
                        search_specific_files.add(source_details[1])
 
1980
                    # generate the old path; this is needed for stating later
 
1981
                    # as well.
 
1982
                    old_path = source_details[1]
 
1983
                    old_dirname, old_basename = os.path.split(old_path)
 
1984
                    path = pathjoin(entry[0][0], entry[0][1])
 
1985
                    old_entry = state._get_entry(source_index,
 
1986
                                                 path_utf8=old_path)
 
1987
                    # update the source details variable to be the real
 
1988
                    # location.
 
1989
                    if old_entry == (None, None):
 
1990
                        raise errors.CorruptDirstate(state._filename,
 
1991
                            "entry '%s/%s' is considered renamed from %r"
 
1992
                            " but source does not exist\n"
 
1993
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
1994
                    source_details = old_entry[1][source_index]
 
1995
                    source_minikind = source_details[0]
 
1996
                else:
 
1997
                    old_dirname = entry[0][0]
 
1998
                    old_basename = entry[0][1]
 
1999
                    old_path = path = None
 
2000
                if path_info is None:
 
2001
                    # the file is missing on disk, show as removed.
 
2002
                    content_change = True
 
2003
                    target_kind = None
 
2004
                    target_exec = False
 
2005
                else:
 
2006
                    # source and target are both versioned and disk file is present.
 
2007
                    target_kind = path_info[2]
 
2008
                    if target_kind == 'directory':
 
2009
                        if path is None:
 
2010
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2011
                        new_dirname_to_file_id[path] = file_id
 
2012
                        if source_minikind != 'd':
 
2013
                            content_change = True
 
2014
                        else:
 
2015
                            # directories have no fingerprint
 
2016
                            content_change = False
 
2017
                        target_exec = False
 
2018
                    elif target_kind == 'file':
 
2019
                        if source_minikind != 'f':
 
2020
                            content_change = True
 
2021
                        else:
 
2022
                            # We could check the size, but we already have the
 
2023
                            # sha1 hash.
 
2024
                            content_change = (link_or_sha1 != source_details[1])
 
2025
                        # Target details is updated at update_entry time
 
2026
                        if use_filesystem_for_exec:
 
2027
                            # We don't need S_ISREG here, because we are sure
 
2028
                            # we are dealing with a file.
 
2029
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2030
                        else:
 
2031
                            target_exec = target_details[3]
 
2032
                    elif target_kind == 'symlink':
 
2033
                        if source_minikind != 'l':
 
2034
                            content_change = True
 
2035
                        else:
 
2036
                            content_change = (link_or_sha1 != source_details[1])
 
2037
                        target_exec = False
 
2038
                    elif target_kind == 'tree-reference':
 
2039
                        if source_minikind != 't':
 
2040
                            content_change = True
 
2041
                        else:
 
2042
                            content_change = False
 
2043
                        target_exec = False
 
2044
                    else:
 
2045
                        raise Exception, "unknown kind %s" % path_info[2]
 
2046
                if source_minikind == 'd':
 
2047
                    if path is None:
 
2048
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2049
                    old_dirname_to_file_id[old_path] = file_id
 
2050
                # parent id is the entry for the path in the target tree
 
2051
                if old_dirname == last_source_parent[0]:
 
2052
                    source_parent_id = last_source_parent[1]
 
2053
                else:
 
2054
                    try:
 
2055
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2056
                    except KeyError:
 
2057
                        source_parent_entry = state._get_entry(source_index,
 
2058
                                                               path_utf8=old_dirname)
 
2059
                        source_parent_id = source_parent_entry[0][2]
 
2060
                    if source_parent_id == entry[0][2]:
 
2061
                        # This is the root, so the parent is None
 
2062
                        source_parent_id = None
 
2063
                    else:
 
2064
                        last_source_parent[0] = old_dirname
 
2065
                        last_source_parent[1] = source_parent_id
 
2066
                new_dirname = entry[0][0]
 
2067
                if new_dirname == last_target_parent[0]:
 
2068
                    target_parent_id = last_target_parent[1]
 
2069
                else:
 
2070
                    try:
 
2071
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2072
                    except KeyError:
 
2073
                        # TODO: We don't always need to do the lookup, because the
 
2074
                        #       parent entry will be the same as the source entry.
 
2075
                        target_parent_entry = state._get_entry(target_index,
 
2076
                                                               path_utf8=new_dirname)
 
2077
                        assert target_parent_entry != (None, None), (
 
2078
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2079
                            % (new_dirname, entry))
 
2080
                        target_parent_id = target_parent_entry[0][2]
 
2081
                    if target_parent_id == entry[0][2]:
 
2082
                        # This is the root, so the parent is None
 
2083
                        target_parent_id = None
 
2084
                    else:
 
2085
                        last_target_parent[0] = new_dirname
 
2086
                        last_target_parent[1] = target_parent_id
 
2087
 
 
2088
                source_exec = source_details[3]
 
2089
                if (include_unchanged
 
2090
                    or content_change
 
2091
                    or source_parent_id != target_parent_id
 
2092
                    or old_basename != entry[0][1]
 
2093
                    or source_exec != target_exec
 
2094
                    ):
 
2095
                    if old_path is None:
 
2096
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2097
                        old_path_u = utf8_decode(old_path)[0]
 
2098
                        path_u = old_path_u
 
2099
                    else:
 
2100
                        old_path_u = utf8_decode(old_path)[0]
 
2101
                        if old_path == path:
 
2102
                            path_u = old_path_u
 
2103
                        else:
 
2104
                            path_u = utf8_decode(path)[0]
 
2105
                    source_kind = _minikind_to_kind[source_minikind]
 
2106
                    return (entry[0][2],
 
2107
                           (old_path_u, path_u),
 
2108
                           content_change,
 
2109
                           (True, True),
 
2110
                           (source_parent_id, target_parent_id),
 
2111
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2112
                           (source_kind, target_kind),
 
2113
                           (source_exec, target_exec))
 
2114
                else:
 
2115
                    return uninteresting
 
2116
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2117
                # looks like a new file
 
2118
                if path_info is not None:
 
2119
                    path = pathjoin(entry[0][0], entry[0][1])
 
2120
                    # parent id is the entry for the path in the target tree
 
2121
                    # TODO: these are the same for an entire directory: cache em.
 
2122
                    parent_id = state._get_entry(target_index,
 
2123
                                                 path_utf8=entry[0][0])[0][2]
 
2124
                    if parent_id == entry[0][2]:
 
2125
                        parent_id = None
 
2126
                    if use_filesystem_for_exec:
 
2127
                        # We need S_ISREG here, because we aren't sure if this
 
2128
                        # is a file or not.
 
2129
                        target_exec = bool(
 
2130
                            stat.S_ISREG(path_info[3].st_mode)
 
2131
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2132
                    else:
 
2133
                        target_exec = target_details[3]
 
2134
                    return (entry[0][2],
 
2135
                           (None, utf8_decode(path)[0]),
 
2136
                           True,
 
2137
                           (False, True),
 
2138
                           (None, parent_id),
 
2139
                           (None, utf8_decode(entry[0][1])[0]),
 
2140
                           (None, path_info[2]),
 
2141
                           (None, target_exec))
 
2142
                else:
 
2143
                    # but its not on disk: we deliberately treat this as just
 
2144
                    # never-present. (Why ?! - RBC 20070224)
 
2145
                    pass
 
2146
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2147
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2148
                # if its still on disk, *and* theres no other entry at this
 
2149
                # path [we dont know this in this routine at the moment -
 
2150
                # perhaps we should change this - then it would be an unknown.
 
2151
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2152
                # parent id is the entry for the path in the target tree
 
2153
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2154
                if parent_id == entry[0][2]:
 
2155
                    parent_id = None
 
2156
                return (entry[0][2],
 
2157
                       (utf8_decode(old_path)[0], None),
 
2158
                       True,
 
2159
                       (True, False),
 
2160
                       (parent_id, None),
 
2161
                       (utf8_decode(entry[0][1])[0], None),
 
2162
                       (_minikind_to_kind[source_minikind], None),
 
2163
                       (source_details[3], None))
 
2164
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2165
                # a rename; could be a true rename, or a rename inherited from
 
2166
                # a renamed parent. TODO: handle this efficiently. Its not
 
2167
                # common case to rename dirs though, so a correct but slow
 
2168
                # implementation will do.
 
2169
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2170
                    search_specific_files.add(target_details[1])
 
2171
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2172
                # neither of the selected trees contain this file,
 
2173
                # so skip over it. This is not currently directly tested, but
 
2174
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2175
                pass
 
2176
            else:
 
2177
                raise AssertionError("don't know how to compare "
 
2178
                    "source_minikind=%r, target_minikind=%r"
 
2179
                    % (source_minikind, target_minikind))
 
2180
                ## import pdb;pdb.set_trace()
 
2181
            return None
 
2182
 
 
2183
        while search_specific_files:
 
2184
            # TODO: the pending list should be lexically sorted?  the
 
2185
            # interface doesn't require it.
 
2186
            current_root = search_specific_files.pop()
 
2187
            current_root_unicode = current_root.decode('utf8')
 
2188
            searched_specific_files.add(current_root)
 
2189
            # process the entries for this containing directory: the rest will be
 
2190
            # found by their parents recursively.
 
2191
            root_entries = _entries_for_path(current_root)
 
2192
            root_abspath = self.target.abspath(current_root_unicode)
 
2193
            try:
 
2194
                root_stat = os.lstat(root_abspath)
 
2195
            except OSError, e:
 
2196
                if e.errno == errno.ENOENT:
 
2197
                    # the path does not exist: let _process_entry know that.
 
2198
                    root_dir_info = None
 
2199
                else:
 
2200
                    # some other random error: hand it up.
 
2201
                    raise
 
2202
            else:
 
2203
                root_dir_info = ('', current_root,
 
2204
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2205
                    root_abspath)
 
2206
                if root_dir_info[2] == 'directory':
 
2207
                    if self.target._directory_is_tree_reference(
 
2208
                        current_root.decode('utf8')):
 
2209
                        root_dir_info = root_dir_info[:2] + \
 
2210
                            ('tree-reference',) + root_dir_info[3:]
 
2211
 
 
2212
            if not root_entries and not root_dir_info:
 
2213
                # this specified path is not present at all, skip it.
 
2214
                continue
 
2215
            path_handled = False
 
2216
            for entry in root_entries:
 
2217
                result = _process_entry(entry, root_dir_info)
 
2218
                if result is not None:
 
2219
                    path_handled = True
 
2220
                    if result is not uninteresting:
 
2221
                        yield result
 
2222
            if want_unversioned and not path_handled and root_dir_info:
 
2223
                new_executable = bool(
 
2224
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2225
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2226
                yield (None,
 
2227
                       (None, current_root_unicode),
 
2228
                       True,
 
2229
                       (False, False),
 
2230
                       (None, None),
 
2231
                       (None, splitpath(current_root_unicode)[-1]),
 
2232
                       (None, root_dir_info[2]),
 
2233
                       (None, new_executable)
 
2234
                      )
 
2235
            initial_key = (current_root, '', '')
 
2236
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2237
            if block_index == 0:
 
2238
                # we have processed the total root already, but because the
 
2239
                # initial key matched it we should skip it here.
 
2240
                block_index +=1
 
2241
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2242
                current_dir_info = None
 
2243
            else:
 
2244
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2245
                try:
 
2246
                    current_dir_info = dir_iterator.next()
 
2247
                except OSError, e:
 
2248
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2249
                    # python 2.5 has e.errno == EINVAL,
 
2250
                    #            and e.winerror == ERROR_DIRECTORY
 
2251
                    e_winerror = getattr(e, 'winerror', None)
 
2252
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2253
                    # there may be directories in the inventory even though
 
2254
                    # this path is not a file on disk: so mark it as end of
 
2255
                    # iterator
 
2256
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2257
                        current_dir_info = None
 
2258
                    elif (sys.platform == 'win32'
 
2259
                          and (e.errno in win_errors
 
2260
                               or e_winerror in win_errors)):
 
2261
                        current_dir_info = None
 
2262
                    else:
 
2263
                        raise
 
2264
                else:
 
2265
                    if current_dir_info[0][0] == '':
 
2266
                        # remove .bzr from iteration
 
2267
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2268
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2269
                        del current_dir_info[1][bzr_index]
 
2270
            # walk until both the directory listing and the versioned metadata
 
2271
            # are exhausted. 
 
2272
            if (block_index < len(state._dirblocks) and
 
2273
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2274
                current_block = state._dirblocks[block_index]
 
2275
            else:
 
2276
                current_block = None
 
2277
            while (current_dir_info is not None or
 
2278
                   current_block is not None):
 
2279
                if (current_dir_info and current_block
 
2280
                    and current_dir_info[0][0] != current_block[0]):
 
2281
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2282
                        # filesystem data refers to paths not covered by the dirblock.
 
2283
                        # this has two possibilities:
 
2284
                        # A) it is versioned but empty, so there is no block for it
 
2285
                        # B) it is not versioned.
 
2286
 
 
2287
                        # if (A) then we need to recurse into it to check for
 
2288
                        # new unknown files or directories.
 
2289
                        # if (B) then we should ignore it, because we don't
 
2290
                        # recurse into unknown directories.
 
2291
                        path_index = 0
 
2292
                        while path_index < len(current_dir_info[1]):
 
2293
                                current_path_info = current_dir_info[1][path_index]
 
2294
                                if want_unversioned:
 
2295
                                    if current_path_info[2] == 'directory':
 
2296
                                        if self.target._directory_is_tree_reference(
 
2297
                                            current_path_info[0].decode('utf8')):
 
2298
                                            current_path_info = current_path_info[:2] + \
 
2299
                                                ('tree-reference',) + current_path_info[3:]
 
2300
                                    new_executable = bool(
 
2301
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2302
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2303
                                    yield (None,
 
2304
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2305
                                        True,
 
2306
                                        (False, False),
 
2307
                                        (None, None),
 
2308
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2309
                                        (None, current_path_info[2]),
 
2310
                                        (None, new_executable))
 
2311
                                # dont descend into this unversioned path if it is
 
2312
                                # a dir
 
2313
                                if current_path_info[2] in ('directory',
 
2314
                                                            'tree-reference'):
 
2315
                                    del current_dir_info[1][path_index]
 
2316
                                    path_index -= 1
 
2317
                                path_index += 1
 
2318
 
 
2319
                        # This dir info has been handled, go to the next
 
2320
                        try:
 
2321
                            current_dir_info = dir_iterator.next()
 
2322
                        except StopIteration:
 
2323
                            current_dir_info = None
 
2324
                    else:
 
2325
                        # We have a dirblock entry for this location, but there
 
2326
                        # is no filesystem path for this. This is most likely
 
2327
                        # because a directory was removed from the disk.
 
2328
                        # We don't have to report the missing directory,
 
2329
                        # because that should have already been handled, but we
 
2330
                        # need to handle all of the files that are contained
 
2331
                        # within.
 
2332
                        for current_entry in current_block[1]:
 
2333
                            # entry referring to file not present on disk.
 
2334
                            # advance the entry only, after processing.
 
2335
                            result = _process_entry(current_entry, None)
 
2336
                            if result is not None:
 
2337
                                if result is not uninteresting:
 
2338
                                    yield result
 
2339
                        block_index +=1
 
2340
                        if (block_index < len(state._dirblocks) and
 
2341
                            osutils.is_inside(current_root,
 
2342
                                              state._dirblocks[block_index][0])):
 
2343
                            current_block = state._dirblocks[block_index]
 
2344
                        else:
 
2345
                            current_block = None
 
2346
                    continue
 
2347
                entry_index = 0
 
2348
                if current_block and entry_index < len(current_block[1]):
 
2349
                    current_entry = current_block[1][entry_index]
 
2350
                else:
 
2351
                    current_entry = None
 
2352
                advance_entry = True
 
2353
                path_index = 0
 
2354
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2355
                    current_path_info = current_dir_info[1][path_index]
 
2356
                    if current_path_info[2] == 'directory':
 
2357
                        if self.target._directory_is_tree_reference(
 
2358
                            current_path_info[0].decode('utf8')):
 
2359
                            current_path_info = current_path_info[:2] + \
 
2360
                                ('tree-reference',) + current_path_info[3:]
 
2361
                else:
 
2362
                    current_path_info = None
 
2363
                advance_path = True
 
2364
                path_handled = False
 
2365
                while (current_entry is not None or
 
2366
                    current_path_info is not None):
 
2367
                    if current_entry is None:
 
2368
                        # the check for path_handled when the path is adnvaced
 
2369
                        # will yield this path if needed.
 
2370
                        pass
 
2371
                    elif current_path_info is None:
 
2372
                        # no path is fine: the per entry code will handle it.
 
2373
                        result = _process_entry(current_entry, current_path_info)
 
2374
                        if result is not None:
 
2375
                            if result is not uninteresting:
 
2376
                                yield result
 
2377
                    elif (current_entry[0][1] != current_path_info[1]
 
2378
                          or current_entry[1][target_index][0] in 'ar'):
 
2379
                        # The current path on disk doesn't match the dirblock
 
2380
                        # record. Either the dirblock is marked as absent, or
 
2381
                        # the file on disk is not present at all in the
 
2382
                        # dirblock. Either way, report about the dirblock
 
2383
                        # entry, and let other code handle the filesystem one.
 
2384
 
 
2385
                        # Compare the basename for these files to determine
 
2386
                        # which comes first
 
2387
                        if current_path_info[1] < current_entry[0][1]:
 
2388
                            # extra file on disk: pass for now, but only
 
2389
                            # increment the path, not the entry
 
2390
                            advance_entry = False
 
2391
                        else:
 
2392
                            # entry referring to file not present on disk.
 
2393
                            # advance the entry only, after processing.
 
2394
                            result = _process_entry(current_entry, None)
 
2395
                            if result is not None:
 
2396
                                if result is not uninteresting:
 
2397
                                    yield result
 
2398
                            advance_path = False
 
2399
                    else:
 
2400
                        result = _process_entry(current_entry, current_path_info)
 
2401
                        if result is not None:
 
2402
                            path_handled = True
 
2403
                            if result is not uninteresting:
 
2404
                                yield result
 
2405
                    if advance_entry and current_entry is not None:
 
2406
                        entry_index += 1
 
2407
                        if entry_index < len(current_block[1]):
 
2408
                            current_entry = current_block[1][entry_index]
 
2409
                        else:
 
2410
                            current_entry = None
 
2411
                    else:
 
2412
                        advance_entry = True # reset the advance flaga
 
2413
                    if advance_path and current_path_info is not None:
 
2414
                        if not path_handled:
 
2415
                            # unversioned in all regards
 
2416
                            if want_unversioned:
 
2417
                                new_executable = bool(
 
2418
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2419
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2420
                                yield (None,
 
2421
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2422
                                    True,
 
2423
                                    (False, False),
 
2424
                                    (None, None),
 
2425
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2426
                                    (None, current_path_info[2]),
 
2427
                                    (None, new_executable))
 
2428
                            # dont descend into this unversioned path if it is
 
2429
                            # a dir
 
2430
                            if current_path_info[2] in ('directory'):
 
2431
                                del current_dir_info[1][path_index]
 
2432
                                path_index -= 1
 
2433
                        # dont descend the disk iterator into any tree 
 
2434
                        # paths.
 
2435
                        if current_path_info[2] == 'tree-reference':
 
2436
                            del current_dir_info[1][path_index]
 
2437
                            path_index -= 1
 
2438
                        path_index += 1
 
2439
                        if path_index < len(current_dir_info[1]):
 
2440
                            current_path_info = current_dir_info[1][path_index]
 
2441
                            if current_path_info[2] == 'directory':
 
2442
                                if self.target._directory_is_tree_reference(
 
2443
                                    current_path_info[0].decode('utf8')):
 
2444
                                    current_path_info = current_path_info[:2] + \
 
2445
                                        ('tree-reference',) + current_path_info[3:]
 
2446
                        else:
 
2447
                            current_path_info = None
 
2448
                        path_handled = False
 
2449
                    else:
 
2450
                        advance_path = True # reset the advance flagg.
 
2451
                if current_block is not None:
 
2452
                    block_index += 1
 
2453
                    if (block_index < len(state._dirblocks) and
 
2454
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2455
                        current_block = state._dirblocks[block_index]
 
2456
                    else:
 
2457
                        current_block = None
 
2458
                if current_dir_info is not None:
 
2459
                    try:
 
2460
                        current_dir_info = dir_iterator.next()
 
2461
                    except StopIteration:
 
2462
                        current_dir_info = None
 
2463
 
 
2464
 
 
2465
    @staticmethod
 
2466
    def is_compatible(source, target):
 
2467
        # the target must be a dirstate working tree
 
2468
        if not isinstance(target, WorkingTree4):
 
2469
            return False
 
2470
        # the source must be a revtreee or dirstate rev tree.
 
2471
        if not isinstance(source,
 
2472
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2473
            return False
 
2474
        # the source revid must be in the target dirstate
 
2475
        if not (source._revision_id == NULL_REVISION or
 
2476
            source._revision_id in target.get_parent_ids()):
 
2477
            # TODO: what about ghosts? it may well need to 
 
2478
            # check for them explicitly.
 
2479
            return False
 
2480
        return True
 
2481
 
 
2482
InterTree.register_optimiser(InterDirStateTree)
 
2483
 
 
2484
 
 
2485
class Converter3to4(object):
 
2486
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2487
 
 
2488
    def __init__(self):
 
2489
        self.target_format = WorkingTreeFormat4()
 
2490
 
 
2491
    def convert(self, tree):
 
2492
        # lock the control files not the tree, so that we dont get tree
 
2493
        # on-unlock behaviours, and so that noone else diddles with the 
 
2494
        # tree during upgrade.
 
2495
        tree._control_files.lock_write()
 
2496
        try:
 
2497
            tree.read_working_inventory()
 
2498
            self.create_dirstate_data(tree)
 
2499
            self.update_format(tree)
 
2500
            self.remove_xml_files(tree)
 
2501
        finally:
 
2502
            tree._control_files.unlock()
 
2503
 
 
2504
    def create_dirstate_data(self, tree):
 
2505
        """Create the dirstate based data for tree."""
 
2506
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2507
            ).local_abspath('dirstate')
 
2508
        state = dirstate.DirState.from_tree(tree, local_path)
 
2509
        state.save()
 
2510
        state.unlock()
 
2511
 
 
2512
    def remove_xml_files(self, tree):
 
2513
        """Remove the oldformat 3 data."""
 
2514
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2515
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2516
            'pending-merges', 'stat-cache']:
 
2517
            try:
 
2518
                transport.delete(path)
 
2519
            except errors.NoSuchFile:
 
2520
                # some files are optional - just deal.
 
2521
                pass
 
2522
 
 
2523
    def update_format(self, tree):
 
2524
        """Change the format marker."""
 
2525
        tree._control_files._transport.put_utf8('format',
 
2526
            self.target_format.get_format_string(),
 
2527
            mode=tree._control_files._file_mode)