/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

First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    debug,
 
47
    delta,
 
48
    dirstate,
 
49
    errors,
 
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    trace,
 
59
    transform,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
 
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
 
67
""")
 
68
 
 
69
from bzrlib import symbol_versioning
 
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
 
76
from bzrlib.osutils import (
 
77
    file_kind,
 
78
    isdir,
 
79
    normpath,
 
80
    pathjoin,
 
81
    rand_chars,
 
82
    realpath,
 
83
    safe_unicode,
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
 
87
from bzrlib.transport.local import LocalTransport
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
 
120
    def __init__(self, basedir,
 
121
                 branch,
 
122
                 _control_files=None,
 
123
                 _format=None,
 
124
                 _bzrdir=None):
 
125
        """Construct a WorkingTree for basedir.
 
126
 
 
127
        If the branch is not supplied, it is opened automatically.
 
128
        If the branch is supplied, it must be the branch for this basedir.
 
129
        (branch.base is not cross checked, because for remote branches that
 
130
        would be meaningless).
 
131
        """
 
132
        self._format = _format
 
133
        self.bzrdir = _bzrdir
 
134
        basedir = safe_unicode(basedir)
 
135
        mutter("opening working tree %r", basedir)
 
136
        self._branch = branch
 
137
        self.basedir = realpath(basedir)
 
138
        # if branch is at our basedir and is a format 6 or less
 
139
        # assume all other formats have their own control files.
 
140
        self._control_files = _control_files
 
141
        self._dirty = None
 
142
        #-------------
 
143
        # during a read or write lock these objects are set, and are
 
144
        # None the rest of the time.
 
145
        self._dirstate = None
 
146
        self._inventory = None
 
147
        #-------------
 
148
        self._setup_directory_is_tree_reference()
 
149
        self._detect_case_handling()
 
150
 
 
151
    @needs_tree_write_lock
 
152
    def _add(self, files, ids, kinds):
 
153
        """See MutableTree._add."""
 
154
        state = self.current_dirstate()
 
155
        for f, file_id, kind in zip(files, ids, kinds):
 
156
            f = f.strip('/')
 
157
            if self.path2id(f):
 
158
                # special case tree root handling.
 
159
                if f == '' and self.path2id(f) == ROOT_ID:
 
160
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
161
                continue
 
162
            if file_id is None:
 
163
                file_id = generate_ids.gen_file_id(f)
 
164
            # deliberately add the file with no cached stat or sha1
 
165
            # - on the first access it will be gathered, and we can
 
166
            # always change this once tests are all passing.
 
167
            state.add(f, file_id, kind, None, '')
 
168
        self._make_dirty(reset_inventory=True)
 
169
 
 
170
    def _make_dirty(self, reset_inventory):
 
171
        """Make the tree state dirty.
 
172
 
 
173
        :param reset_inventory: True if the cached inventory should be removed
 
174
            (presuming there is one).
 
175
        """
 
176
        self._dirty = True
 
177
        if reset_inventory and self._inventory is not None:
 
178
            self._inventory = None
 
179
 
 
180
    @needs_tree_write_lock
 
181
    def add_reference(self, sub_tree):
 
182
        # use standard implementation, which calls back to self._add
 
183
        # 
 
184
        # So we don't store the reference_revision in the working dirstate,
 
185
        # it's just recorded at the moment of commit. 
 
186
        self._add_reference(sub_tree)
 
187
 
 
188
    def break_lock(self):
 
189
        """Break a lock if one is present from another instance.
 
190
 
 
191
        Uses the ui factory to ask for confirmation if the lock may be from
 
192
        an active process.
 
193
 
 
194
        This will probe the repository for its lock as well.
 
195
        """
 
196
        # if the dirstate is locked by an active process, reject the break lock
 
197
        # call.
 
198
        try:
 
199
            if self._dirstate is None:
 
200
                clear = True
 
201
            else:
 
202
                clear = False
 
203
            state = self._current_dirstate()
 
204
            if state._lock_token is not None:
 
205
                # we already have it locked. sheese, cant break our own lock.
 
206
                raise errors.LockActive(self.basedir)
 
207
            else:
 
208
                try:
 
209
                    # try for a write lock - need permission to get one anyhow
 
210
                    # to break locks.
 
211
                    state.lock_write()
 
212
                except errors.LockContention:
 
213
                    # oslocks fail when a process is still live: fail.
 
214
                    # TODO: get the locked lockdir info and give to the user to
 
215
                    # assist in debugging.
 
216
                    raise errors.LockActive(self.basedir)
 
217
                else:
 
218
                    state.unlock()
 
219
        finally:
 
220
            if clear:
 
221
                self._dirstate = None
 
222
        self._control_files.break_lock()
 
223
        self.branch.break_lock()
 
224
 
 
225
    def _comparison_data(self, entry, path):
 
226
        kind, executable, stat_value = \
 
227
            WorkingTree3._comparison_data(self, entry, path)
 
228
        # it looks like a plain directory, but it's really a reference -- see
 
229
        # also kind()
 
230
        if (self._repo_supports_tree_reference and
 
231
            kind == 'directory' and
 
232
            self._directory_is_tree_reference(path)):
 
233
            kind = 'tree-reference'
 
234
        return kind, executable, stat_value
 
235
 
 
236
    @needs_write_lock
 
237
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
238
        # mark the tree as dirty post commit - commit
 
239
        # can change the current versioned list by doing deletes.
 
240
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
241
        self._make_dirty(reset_inventory=True)
 
242
        return result
 
243
 
 
244
    def current_dirstate(self):
 
245
        """Return the current dirstate object.
 
246
 
 
247
        This is not part of the tree interface and only exposed for ease of
 
248
        testing.
 
249
 
 
250
        :raises errors.NotWriteLocked: when not in a lock.
 
251
        """
 
252
        self._must_be_locked()
 
253
        return self._current_dirstate()
 
254
 
 
255
    def _current_dirstate(self):
 
256
        """Internal function that does not check lock status.
 
257
 
 
258
        This is needed for break_lock which also needs the dirstate.
 
259
        """
 
260
        if self._dirstate is not None:
 
261
            return self._dirstate
 
262
        local_path = self.bzrdir.get_workingtree_transport(None
 
263
            ).local_abspath('dirstate')
 
264
        self._dirstate = dirstate.DirState.on_file(local_path)
 
265
        return self._dirstate
 
266
 
 
267
    def filter_unversioned_files(self, paths):
 
268
        """Filter out paths that are versioned.
 
269
 
 
270
        :return: set of paths.
 
271
        """
 
272
        # TODO: make a generic multi-bisect routine roughly that should list
 
273
        # the paths, then process one half at a time recursively, and feed the
 
274
        # results of each bisect in further still
 
275
        paths = sorted(paths)
 
276
        result = set()
 
277
        state = self.current_dirstate()
 
278
        # TODO we want a paths_to_dirblocks helper I think
 
279
        for path in paths:
 
280
            dirname, basename = os.path.split(path.encode('utf8'))
 
281
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
282
                dirname, basename, 0)
 
283
            if not path_is_versioned:
 
284
                result.add(path)
 
285
        return result
 
286
 
 
287
    def flush(self):
 
288
        """Write all cached data to disk."""
 
289
        if self._control_files._lock_mode != 'w':
 
290
            raise errors.NotWriteLocked(self)
 
291
        self.current_dirstate().save()
 
292
        self._inventory = None
 
293
        self._dirty = False
 
294
 
 
295
    @needs_tree_write_lock
 
296
    def _gather_kinds(self, files, kinds):
 
297
        """See MutableTree._gather_kinds."""
 
298
        for pos, f in enumerate(files):
 
299
            if kinds[pos] is None:
 
300
                kinds[pos] = self._kind(f)
 
301
 
 
302
    def _generate_inventory(self):
 
303
        """Create and set self.inventory from the dirstate object.
 
304
        
 
305
        This is relatively expensive: we have to walk the entire dirstate.
 
306
        Ideally we would not, and can deprecate this function.
 
307
        """
 
308
        #: uncomment to trap on inventory requests.
 
309
        # import pdb;pdb.set_trace()
 
310
        state = self.current_dirstate()
 
311
        state._read_dirblocks_if_needed()
 
312
        root_key, current_entry = self._get_entry(path='')
 
313
        current_id = root_key[2]
 
314
        if not (current_entry[0][0] == 'd'): # directory
 
315
            raise AssertionError(current_entry)
 
316
        inv = Inventory(root_id=current_id)
 
317
        # Turn some things into local variables
 
318
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
319
        factory = entry_factory
 
320
        utf8_decode = cache_utf8._utf8_decode
 
321
        inv_byid = inv._byid
 
322
        # we could do this straight out of the dirstate; it might be fast
 
323
        # and should be profiled - RBC 20070216
 
324
        parent_ies = {'' : inv.root}
 
325
        for block in state._dirblocks[1:]: # skip the root
 
326
            dirname = block[0]
 
327
            try:
 
328
                parent_ie = parent_ies[dirname]
 
329
            except KeyError:
 
330
                # all the paths in this block are not versioned in this tree
 
331
                continue
 
332
            for key, entry in block[1]:
 
333
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
334
                if minikind in ('a', 'r'): # absent, relocated
 
335
                    # a parent tree only entry
 
336
                    continue
 
337
                name = key[1]
 
338
                name_unicode = utf8_decode(name)[0]
 
339
                file_id = key[2]
 
340
                kind = minikind_to_kind[minikind]
 
341
                inv_entry = factory[kind](file_id, name_unicode,
 
342
                                          parent_ie.file_id)
 
343
                if kind == 'file':
 
344
                    # This is only needed on win32, where this is the only way
 
345
                    # we know the executable bit.
 
346
                    inv_entry.executable = executable
 
347
                    # not strictly needed: working tree
 
348
                    #inv_entry.text_size = size
 
349
                    #inv_entry.text_sha1 = sha1
 
350
                elif kind == 'directory':
 
351
                    # add this entry to the parent map.
 
352
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
353
                elif kind == 'tree-reference':
 
354
                    if not self._repo_supports_tree_reference:
 
355
                        raise AssertionError(
 
356
                            "repository of %r "
 
357
                            "doesn't support tree references "
 
358
                            "required by entry %r"
 
359
                            % (self, name))
 
360
                    inv_entry.reference_revision = link_or_sha1 or None
 
361
                elif kind != 'symlink':
 
362
                    raise AssertionError("unknown kind %r" % kind)
 
363
                # These checks cost us around 40ms on a 55k entry tree
 
364
                if file_id in inv_byid:
 
365
                    raise AssertionError('file_id %s already in'
 
366
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
367
                if name_unicode in parent_ie.children:
 
368
                    raise AssertionError('name %r already in parent'
 
369
                        % (name_unicode,))
 
370
                inv_byid[file_id] = inv_entry
 
371
                parent_ie.children[name_unicode] = inv_entry
 
372
        self._inventory = inv
 
373
 
 
374
    def _get_entry(self, file_id=None, path=None):
 
375
        """Get the dirstate row for file_id or path.
 
376
 
 
377
        If either file_id or path is supplied, it is used as the key to lookup.
 
378
        If both are supplied, the fastest lookup is used, and an error is
 
379
        raised if they do not both point at the same row.
 
380
        
 
381
        :param file_id: An optional unicode file_id to be looked up.
 
382
        :param path: An optional unicode path to be looked up.
 
383
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
384
        """
 
385
        if file_id is None and path is None:
 
386
            raise errors.BzrError('must supply file_id or path')
 
387
        state = self.current_dirstate()
 
388
        if path is not None:
 
389
            path = path.encode('utf8')
 
390
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
391
 
 
392
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
393
        # check file id is valid unconditionally.
 
394
        entry = self._get_entry(file_id=file_id, path=path)
 
395
        if entry[0] is None:
 
396
            raise errors.NoSuchId(self, file_id)
 
397
        if path is None:
 
398
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
399
 
 
400
        file_abspath = self.abspath(path)
 
401
        state = self.current_dirstate()
 
402
        if stat_value is None:
 
403
            try:
 
404
                stat_value = os.lstat(file_abspath)
 
405
            except OSError, e:
 
406
                if e.errno == errno.ENOENT:
 
407
                    return None
 
408
                else:
 
409
                    raise
 
410
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
411
                                          stat_value=stat_value)
 
412
        if entry[1][0][0] == 'f':
 
413
            return link_or_sha1
 
414
        return None
 
415
 
 
416
    def _get_inventory(self):
 
417
        """Get the inventory for the tree. This is only valid within a lock."""
 
418
        if 'evil' in debug.debug_flags:
 
419
            trace.mutter_callsite(2,
 
420
                "accessing .inventory forces a size of tree translation.")
 
421
        if self._inventory is not None:
 
422
            return self._inventory
 
423
        self._must_be_locked()
 
424
        self._generate_inventory()
 
425
        return self._inventory
 
426
 
 
427
    inventory = property(_get_inventory,
 
428
                         doc="Inventory of this Tree")
 
429
 
 
430
    @needs_read_lock
 
431
    def get_parent_ids(self):
 
432
        """See Tree.get_parent_ids.
 
433
        
 
434
        This implementation requests the ids list from the dirstate file.
 
435
        """
 
436
        return self.current_dirstate().get_parent_ids()
 
437
 
 
438
    def get_reference_revision(self, file_id, path=None):
 
439
        # referenced tree's revision is whatever's currently there
 
440
        return self.get_nested_tree(file_id, path).last_revision()
 
441
 
 
442
    def get_nested_tree(self, file_id, path=None):
 
443
        if path is None:
 
444
            path = self.id2path(file_id)
 
445
        # else: check file_id is at path?
 
446
        return WorkingTree.open(self.abspath(path))
 
447
 
 
448
    @needs_read_lock
 
449
    def get_root_id(self):
 
450
        """Return the id of this trees root"""
 
451
        return self._get_entry(path='')[0][2]
 
452
 
 
453
    def has_id(self, file_id):
 
454
        state = self.current_dirstate()
 
455
        row, parents = self._get_entry(file_id=file_id)
 
456
        if row is None:
 
457
            return False
 
458
        return osutils.lexists(pathjoin(
 
459
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
460
 
 
461
    @needs_read_lock
 
462
    def id2path(self, file_id):
 
463
        "Convert a file-id to a path."
 
464
        state = self.current_dirstate()
 
465
        entry = self._get_entry(file_id=file_id)
 
466
        if entry == (None, None):
 
467
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
468
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
469
        return path_utf8.decode('utf8')
 
470
 
 
471
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
472
        entry = self._get_entry(path=path)
 
473
        if entry == (None, None):
 
474
            return False # Missing entries are not executable
 
475
        return entry[1][0][3] # Executable?
 
476
 
 
477
    if not osutils.supports_executable():
 
478
        def is_executable(self, file_id, path=None):
 
479
            """Test if a file is executable or not.
 
480
 
 
481
            Note: The caller is expected to take a read-lock before calling this.
 
482
            """
 
483
            entry = self._get_entry(file_id=file_id, path=path)
 
484
            if entry == (None, None):
 
485
                return False
 
486
            return entry[1][0][3]
 
487
 
 
488
        _is_executable_from_path_and_stat = \
 
489
            _is_executable_from_path_and_stat_from_basis
 
490
    else:
 
491
        def is_executable(self, file_id, path=None):
 
492
            """Test if a file is executable or not.
 
493
 
 
494
            Note: The caller is expected to take a read-lock before calling this.
 
495
            """
 
496
            self._must_be_locked()
 
497
            if not path:
 
498
                path = self.id2path(file_id)
 
499
            mode = os.lstat(self.abspath(path)).st_mode
 
500
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
501
 
 
502
    def all_file_ids(self):
 
503
        """See Tree.iter_all_file_ids"""
 
504
        self._must_be_locked()
 
505
        result = set()
 
506
        for key, tree_details in self.current_dirstate()._iter_entries():
 
507
            if tree_details[0][0] in ('a', 'r'): # relocated
 
508
                continue
 
509
            result.add(key[2])
 
510
        return result
 
511
 
 
512
    @needs_read_lock
 
513
    def __iter__(self):
 
514
        """Iterate through file_ids for this tree.
 
515
 
 
516
        file_ids are in a WorkingTree if they are in the working inventory
 
517
        and the working file exists.
 
518
        """
 
519
        result = []
 
520
        for key, tree_details in self.current_dirstate()._iter_entries():
 
521
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
522
                # not relevant to the working tree
 
523
                continue
 
524
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
525
            if osutils.lexists(path):
 
526
                result.append(key[2])
 
527
        return iter(result)
 
528
 
 
529
    def iter_references(self):
 
530
        for key, tree_details in self.current_dirstate()._iter_entries():
 
531
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
532
                # not relevant to the working tree
 
533
                continue
 
534
            if not key[1]:
 
535
                # the root is not a reference.
 
536
                continue
 
537
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
538
            try:
 
539
                if self._kind(path) == 'tree-reference':
 
540
                    yield path, key[2]
 
541
            except errors.NoSuchFile:
 
542
                # path is missing on disk.
 
543
                continue
 
544
 
 
545
    def kind(self, file_id):
 
546
        """Return the kind of a file.
 
547
 
 
548
        This is always the actual kind that's on disk, regardless of what it
 
549
        was added as.
 
550
 
 
551
        Note: The caller is expected to take a read-lock before calling this.
 
552
        """
 
553
        relpath = self.id2path(file_id)
 
554
        if relpath is None:
 
555
            raise AssertionError(
 
556
                "path for id {%s} is None!" % file_id)
 
557
        return self._kind(relpath)
 
558
 
 
559
    def _kind(self, relpath):
 
560
        abspath = self.abspath(relpath)
 
561
        kind = file_kind(abspath)
 
562
        if (self._repo_supports_tree_reference and
 
563
            kind == 'directory' and
 
564
            self._directory_is_tree_reference(relpath)):
 
565
            kind = 'tree-reference'
 
566
        return kind
 
567
 
 
568
    @needs_read_lock
 
569
    def _last_revision(self):
 
570
        """See Mutable.last_revision."""
 
571
        parent_ids = self.current_dirstate().get_parent_ids()
 
572
        if parent_ids:
 
573
            return parent_ids[0]
 
574
        else:
 
575
            return _mod_revision.NULL_REVISION
 
576
 
 
577
    def lock_read(self):
 
578
        """See Branch.lock_read, and WorkingTree.unlock."""
 
579
        self.branch.lock_read()
 
580
        try:
 
581
            self._control_files.lock_read()
 
582
            try:
 
583
                state = self.current_dirstate()
 
584
                if not state._lock_token:
 
585
                    state.lock_read()
 
586
                # set our support for tree references from the repository in
 
587
                # use.
 
588
                self._repo_supports_tree_reference = getattr(
 
589
                    self.branch.repository._format, "supports_tree_reference",
 
590
                    False)
 
591
            except:
 
592
                self._control_files.unlock()
 
593
                raise
 
594
        except:
 
595
            self.branch.unlock()
 
596
            raise
 
597
 
 
598
    def _lock_self_write(self):
 
599
        """This should be called after the branch is locked."""
 
600
        try:
 
601
            self._control_files.lock_write()
 
602
            try:
 
603
                state = self.current_dirstate()
 
604
                if not state._lock_token:
 
605
                    state.lock_write()
 
606
                # set our support for tree references from the repository in
 
607
                # use.
 
608
                self._repo_supports_tree_reference = getattr(
 
609
                    self.branch.repository._format, "supports_tree_reference",
 
610
                    False)
 
611
            except:
 
612
                self._control_files.unlock()
 
613
                raise
 
614
        except:
 
615
            self.branch.unlock()
 
616
            raise
 
617
 
 
618
    def lock_tree_write(self):
 
619
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
620
        self.branch.lock_read()
 
621
        self._lock_self_write()
 
622
 
 
623
    def lock_write(self):
 
624
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
625
        self.branch.lock_write()
 
626
        self._lock_self_write()
 
627
 
 
628
    @needs_tree_write_lock
 
629
    def move(self, from_paths, to_dir, after=False):
 
630
        """See WorkingTree.move()."""
 
631
        result = []
 
632
        if not from_paths:
 
633
            return result
 
634
        state = self.current_dirstate()
 
635
        if isinstance(from_paths, basestring):
 
636
            raise ValueError()
 
637
        to_dir_utf8 = to_dir.encode('utf8')
 
638
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
639
        id_index = state._get_id_index()
 
640
        # check destination directory
 
641
        # get the details for it
 
642
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
643
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
644
        if not entry_present:
 
645
            raise errors.BzrMoveFailedError('', to_dir,
 
646
                errors.NotVersionedError(to_dir))
 
647
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
648
        # get a handle on the block itself.
 
649
        to_block_index = state._ensure_block(
 
650
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
651
        to_block = state._dirblocks[to_block_index]
 
652
        to_abs = self.abspath(to_dir)
 
653
        if not isdir(to_abs):
 
654
            raise errors.BzrMoveFailedError('',to_dir,
 
655
                errors.NotADirectory(to_abs))
 
656
 
 
657
        if to_entry[1][0][0] != 'd':
 
658
            raise errors.BzrMoveFailedError('',to_dir,
 
659
                errors.NotADirectory(to_abs))
 
660
 
 
661
        if self._inventory is not None:
 
662
            update_inventory = True
 
663
            inv = self.inventory
 
664
            to_dir_id = to_entry[0][2]
 
665
            to_dir_ie = inv[to_dir_id]
 
666
        else:
 
667
            update_inventory = False
 
668
 
 
669
        rollbacks = []
 
670
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
671
                     fingerprint, packed_stat, size,
 
672
                     to_block, to_key, to_path_utf8):
 
673
            state._make_absent(old_entry)
 
674
            from_key = old_entry[0]
 
675
            rollbacks.append(
 
676
                lambda:state.update_minimal(from_key,
 
677
                    minikind,
 
678
                    executable=executable,
 
679
                    fingerprint=fingerprint,
 
680
                    packed_stat=packed_stat,
 
681
                    size=size,
 
682
                    path_utf8=from_path_utf8))
 
683
            state.update_minimal(to_key,
 
684
                    minikind,
 
685
                    executable=executable,
 
686
                    fingerprint=fingerprint,
 
687
                    packed_stat=packed_stat,
 
688
                    size=size,
 
689
                    path_utf8=to_path_utf8)
 
690
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
691
            new_entry = to_block[1][added_entry_index]
 
692
            rollbacks.append(lambda:state._make_absent(new_entry))
 
693
 
 
694
        for from_rel in from_paths:
 
695
            # from_rel is 'pathinroot/foo/bar'
 
696
            from_rel_utf8 = from_rel.encode('utf8')
 
697
            from_dirname, from_tail = osutils.split(from_rel)
 
698
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
699
            from_entry = self._get_entry(path=from_rel)
 
700
            if from_entry == (None, None):
 
701
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
702
                    errors.NotVersionedError(path=str(from_rel)))
 
703
 
 
704
            from_id = from_entry[0][2]
 
705
            to_rel = pathjoin(to_dir, from_tail)
 
706
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
707
            item_to_entry = self._get_entry(path=to_rel)
 
708
            if item_to_entry != (None, None):
 
709
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
710
                    "Target is already versioned.")
 
711
 
 
712
            if from_rel == to_rel:
 
713
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
714
                    "Source and target are identical.")
 
715
 
 
716
            from_missing = not self.has_filename(from_rel)
 
717
            to_missing = not self.has_filename(to_rel)
 
718
            if after:
 
719
                move_file = False
 
720
            else:
 
721
                move_file = True
 
722
            if to_missing:
 
723
                if not move_file:
 
724
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
725
                        errors.NoSuchFile(path=to_rel,
 
726
                        extra="New file has not been created yet"))
 
727
                elif from_missing:
 
728
                    # neither path exists
 
729
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
730
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
731
            else:
 
732
                if from_missing: # implicitly just update our path mapping
 
733
                    move_file = False
 
734
                elif not after:
 
735
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
736
 
 
737
            rollbacks = []
 
738
            def rollback_rename():
 
739
                """A single rename has failed, roll it back."""
 
740
                # roll back everything, even if we encounter trouble doing one
 
741
                # of them.
 
742
                #
 
743
                # TODO: at least log the other exceptions rather than just
 
744
                # losing them mbp 20070307
 
745
                exc_info = None
 
746
                for rollback in reversed(rollbacks):
 
747
                    try:
 
748
                        rollback()
 
749
                    except Exception, e:
 
750
                        exc_info = sys.exc_info()
 
751
                if exc_info:
 
752
                    raise exc_info[0], exc_info[1], exc_info[2]
 
753
 
 
754
            # perform the disk move first - its the most likely failure point.
 
755
            if move_file:
 
756
                from_rel_abs = self.abspath(from_rel)
 
757
                to_rel_abs = self.abspath(to_rel)
 
758
                try:
 
759
                    osutils.rename(from_rel_abs, to_rel_abs)
 
760
                except OSError, e:
 
761
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
762
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
763
            try:
 
764
                # perform the rename in the inventory next if needed: its easy
 
765
                # to rollback
 
766
                if update_inventory:
 
767
                    # rename the entry
 
768
                    from_entry = inv[from_id]
 
769
                    current_parent = from_entry.parent_id
 
770
                    inv.rename(from_id, to_dir_id, from_tail)
 
771
                    rollbacks.append(
 
772
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
773
                # finally do the rename in the dirstate, which is a little
 
774
                # tricky to rollback, but least likely to need it.
 
775
                old_block_index, old_entry_index, dir_present, file_present = \
 
776
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
777
                old_block = state._dirblocks[old_block_index][1]
 
778
                old_entry = old_block[old_entry_index]
 
779
                from_key, old_entry_details = old_entry
 
780
                cur_details = old_entry_details[0]
 
781
                # remove the old row
 
782
                to_key = ((to_block[0],) + from_key[1:3])
 
783
                minikind = cur_details[0]
 
784
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
785
                         minikind=minikind,
 
786
                         executable=cur_details[3],
 
787
                         fingerprint=cur_details[1],
 
788
                         packed_stat=cur_details[4],
 
789
                         size=cur_details[2],
 
790
                         to_block=to_block,
 
791
                         to_key=to_key,
 
792
                         to_path_utf8=to_rel_utf8)
 
793
 
 
794
                if minikind == 'd':
 
795
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
796
                        """Recursively update all entries in this dirblock."""
 
797
                        if from_dir == '':
 
798
                            raise AssertionError("renaming root not supported")
 
799
                        from_key = (from_dir, '')
 
800
                        from_block_idx, present = \
 
801
                            state._find_block_index_from_key(from_key)
 
802
                        if not present:
 
803
                            # This is the old record, if it isn't present, then
 
804
                            # there is theoretically nothing to update.
 
805
                            # (Unless it isn't present because of lazy loading,
 
806
                            # but we don't do that yet)
 
807
                            return
 
808
                        from_block = state._dirblocks[from_block_idx]
 
809
                        to_block_index, to_entry_index, _, _ = \
 
810
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
811
                        to_block_index = state._ensure_block(
 
812
                            to_block_index, to_entry_index, to_dir_utf8)
 
813
                        to_block = state._dirblocks[to_block_index]
 
814
 
 
815
                        # Grab a copy since move_one may update the list.
 
816
                        for entry in from_block[1][:]:
 
817
                            if not (entry[0][0] == from_dir):
 
818
                                raise AssertionError()
 
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
            if len(parents) >= 2:
 
1038
                raise AssertionError(
 
1039
                    "setting the last parent to none with a pending merge is "
 
1040
                    "unsupported.")
 
1041
            self.set_parent_ids([])
 
1042
        else:
 
1043
            self.set_parent_ids([new_revision] + parents[1:],
 
1044
                allow_leftmost_as_ghost=True)
 
1045
 
 
1046
    @needs_tree_write_lock
 
1047
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1048
        """Set the parent ids to revision_ids.
 
1049
        
 
1050
        See also set_parent_trees. This api will try to retrieve the tree data
 
1051
        for each element of revision_ids from the trees repository. If you have
 
1052
        tree data already available, it is more efficient to use
 
1053
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1054
        an easier API to use.
 
1055
 
 
1056
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1057
            working tree. Any of these may be ghosts.
 
1058
        """
 
1059
        trees = []
 
1060
        for revision_id in revision_ids:
 
1061
            try:
 
1062
                revtree = self.branch.repository.revision_tree(revision_id)
 
1063
                # TODO: jam 20070213 KnitVersionedFile raises
 
1064
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1065
                #       given revision_id is not present. Should Repository be
 
1066
                #       catching it and re-raising NoSuchRevision?
 
1067
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1068
                revtree = None
 
1069
            trees.append((revision_id, revtree))
 
1070
        self.set_parent_trees(trees,
 
1071
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1072
 
 
1073
    @needs_tree_write_lock
 
1074
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1075
        """Set the parents of the working tree.
 
1076
 
 
1077
        :param parents_list: A list of (revision_id, tree) tuples.
 
1078
            If tree is None, then that element is treated as an unreachable
 
1079
            parent tree - i.e. a ghost.
 
1080
        """
 
1081
        dirstate = self.current_dirstate()
 
1082
        if len(parents_list) > 0:
 
1083
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1084
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1085
        real_trees = []
 
1086
        ghosts = []
 
1087
        # convert absent trees to the null tree, which we convert back to
 
1088
        # missing on access.
 
1089
        for rev_id, tree in parents_list:
 
1090
            _mod_revision.check_not_reserved_id(rev_id)
 
1091
            if tree is not None:
 
1092
                real_trees.append((rev_id, tree))
 
1093
            else:
 
1094
                real_trees.append((rev_id,
 
1095
                    self.branch.repository.revision_tree(None)))
 
1096
                ghosts.append(rev_id)
 
1097
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1098
        self._make_dirty(reset_inventory=False)
 
1099
 
 
1100
    def _set_root_id(self, file_id):
 
1101
        """See WorkingTree.set_root_id."""
 
1102
        state = self.current_dirstate()
 
1103
        state.set_path_id('', file_id)
 
1104
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1105
            self._make_dirty(reset_inventory=True)
 
1106
 
 
1107
    def _sha_from_stat(self, path, stat_result):
 
1108
        """Get a sha digest from the tree's stat cache.
 
1109
 
 
1110
        The default implementation assumes no stat cache is present.
 
1111
 
 
1112
        :param path: The path.
 
1113
        :param stat_result: The stat result being looked up.
 
1114
        """
 
1115
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1116
 
 
1117
    @needs_read_lock
 
1118
    def supports_tree_reference(self):
 
1119
        return self._repo_supports_tree_reference
 
1120
 
 
1121
    def unlock(self):
 
1122
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1123
        # do non-implementation specific cleanup
 
1124
        self._cleanup()
 
1125
 
 
1126
        if self._control_files._lock_count == 1:
 
1127
            # eventually we should do signature checking during read locks for
 
1128
            # dirstate updates.
 
1129
            if self._control_files._lock_mode == 'w':
 
1130
                if self._dirty:
 
1131
                    self.flush()
 
1132
            if self._dirstate is not None:
 
1133
                # This is a no-op if there are no modifications.
 
1134
                self._dirstate.save()
 
1135
                self._dirstate.unlock()
 
1136
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1137
            #       point. Instead, it could check if the header has been
 
1138
            #       modified when it is locked, and if not, it can hang on to
 
1139
            #       the data it has in memory.
 
1140
            self._dirstate = None
 
1141
            self._inventory = None
 
1142
        # reverse order of locking.
 
1143
        try:
 
1144
            return self._control_files.unlock()
 
1145
        finally:
 
1146
            self.branch.unlock()
 
1147
 
 
1148
    @needs_tree_write_lock
 
1149
    def unversion(self, file_ids):
 
1150
        """Remove the file ids in file_ids from the current versioned set.
 
1151
 
 
1152
        When a file_id is unversioned, all of its children are automatically
 
1153
        unversioned.
 
1154
 
 
1155
        :param file_ids: The file ids to stop versioning.
 
1156
        :raises: NoSuchId if any fileid is not currently versioned.
 
1157
        """
 
1158
        if not file_ids:
 
1159
            return
 
1160
        state = self.current_dirstate()
 
1161
        state._read_dirblocks_if_needed()
 
1162
        ids_to_unversion = set(file_ids)
 
1163
        paths_to_unversion = set()
 
1164
        # sketch:
 
1165
        # check if the root is to be unversioned, if so, assert for now.
 
1166
        # walk the state marking unversioned things as absent.
 
1167
        # if there are any un-unversioned ids at the end, raise
 
1168
        for key, details in state._dirblocks[0][1]:
 
1169
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1170
                key[2] in ids_to_unversion):
 
1171
                # I haven't written the code to unversion / yet - it should be
 
1172
                # supported.
 
1173
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1174
        block_index = 0
 
1175
        while block_index < len(state._dirblocks):
 
1176
            # process one directory at a time.
 
1177
            block = state._dirblocks[block_index]
 
1178
            # first check: is the path one to remove - it or its children
 
1179
            delete_block = False
 
1180
            for path in paths_to_unversion:
 
1181
                if (block[0].startswith(path) and
 
1182
                    (len(block[0]) == len(path) or
 
1183
                     block[0][len(path)] == '/')):
 
1184
                    # this entire block should be deleted - its the block for a
 
1185
                    # path to unversion; or the child of one
 
1186
                    delete_block = True
 
1187
                    break
 
1188
            # TODO: trim paths_to_unversion as we pass by paths
 
1189
            if delete_block:
 
1190
                # this block is to be deleted: process it.
 
1191
                # TODO: we can special case the no-parents case and
 
1192
                # just forget the whole block.
 
1193
                entry_index = 0
 
1194
                while entry_index < len(block[1]):
 
1195
                    # Mark this file id as having been removed
 
1196
                    entry = block[1][entry_index]
 
1197
                    ids_to_unversion.discard(entry[0][2])
 
1198
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1199
                                               # entries
 
1200
                        or not state._make_absent(entry)):
 
1201
                        entry_index += 1
 
1202
                # go to the next block. (At the moment we dont delete empty
 
1203
                # dirblocks)
 
1204
                block_index += 1
 
1205
                continue
 
1206
            entry_index = 0
 
1207
            while entry_index < len(block[1]):
 
1208
                entry = block[1][entry_index]
 
1209
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1210
                    # ^ some parent row.
 
1211
                    entry[0][2] not in ids_to_unversion):
 
1212
                    # ^ not an id to unversion
 
1213
                    entry_index += 1
 
1214
                    continue
 
1215
                if entry[1][0][0] == 'd':
 
1216
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1217
                if not state._make_absent(entry):
 
1218
                    entry_index += 1
 
1219
                # we have unversioned this id
 
1220
                ids_to_unversion.remove(entry[0][2])
 
1221
            block_index += 1
 
1222
        if ids_to_unversion:
 
1223
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1224
        self._make_dirty(reset_inventory=False)
 
1225
        # have to change the legacy inventory too.
 
1226
        if self._inventory is not None:
 
1227
            for file_id in file_ids:
 
1228
                self._inventory.remove_recursive_id(file_id)
 
1229
 
 
1230
    @needs_tree_write_lock
 
1231
    def rename_one(self, from_rel, to_rel, after=False):
 
1232
        """See WorkingTree.rename_one"""
 
1233
        self.flush()
 
1234
        WorkingTree.rename_one(self, from_rel, to_rel, after)
 
1235
 
 
1236
    @needs_tree_write_lock
 
1237
    def apply_inventory_delta(self, changes):
 
1238
        """See MutableTree.apply_inventory_delta"""
 
1239
        state = self.current_dirstate()
 
1240
        state.update_by_delta(changes)
 
1241
        self._make_dirty(reset_inventory=True)
 
1242
 
 
1243
    def update_basis_by_delta(self, new_revid, delta):
 
1244
        """See MutableTree.update_basis_by_delta."""
 
1245
        if self.last_revision() == new_revid:
 
1246
            raise AssertionError()
 
1247
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1248
 
 
1249
    @needs_read_lock
 
1250
    def _validate(self):
 
1251
        self._dirstate._validate()
 
1252
 
 
1253
    @needs_tree_write_lock
 
1254
    def _write_inventory(self, inv):
 
1255
        """Write inventory as the current inventory."""
 
1256
        if self._dirty:
 
1257
            raise AssertionError("attempting to write an inventory when the "
 
1258
                "dirstate is dirty will lose pending changes")
 
1259
        self.current_dirstate().set_state_from_inventory(inv)
 
1260
        self._make_dirty(reset_inventory=False)
 
1261
        if self._inventory is not None:
 
1262
            self._inventory = inv
 
1263
        self.flush()
 
1264
 
 
1265
 
 
1266
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1267
    """The first consolidated dirstate working tree format.
 
1268
 
 
1269
    This format:
 
1270
        - exists within a metadir controlling .bzr
 
1271
        - includes an explicit version marker for the workingtree control
 
1272
          files, separate from the BzrDir format
 
1273
        - modifies the hash cache format
 
1274
        - is new in bzr 0.15
 
1275
        - uses a LockDir to guard access to it.
 
1276
    """
 
1277
 
 
1278
    upgrade_recommended = False
 
1279
 
 
1280
    def get_format_string(self):
 
1281
        """See WorkingTreeFormat.get_format_string()."""
 
1282
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1283
 
 
1284
    def get_format_description(self):
 
1285
        """See WorkingTreeFormat.get_format_description()."""
 
1286
        return "Working tree format 4"
 
1287
 
 
1288
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1289
                   accelerator_tree=None, hardlink=False):
 
1290
        """See WorkingTreeFormat.initialize().
 
1291
 
 
1292
        :param revision_id: allows creating a working tree at a different
 
1293
        revision than the branch is at.
 
1294
        :param accelerator_tree: A tree which can be used for retrieving file
 
1295
            contents more quickly than the revision tree, i.e. a workingtree.
 
1296
            The revision tree will be used for cases where accelerator_tree's
 
1297
            content is different.
 
1298
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1299
            where possible.
 
1300
 
 
1301
        These trees get an initial random root id, if their repository supports
 
1302
        rich root data, TREE_ROOT otherwise.
 
1303
        """
 
1304
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1305
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1306
        transport = a_bzrdir.get_workingtree_transport(self)
 
1307
        control_files = self._open_control_files(a_bzrdir)
 
1308
        control_files.create_lock()
 
1309
        control_files.lock_write()
 
1310
        control_files.put_utf8('format', self.get_format_string())
 
1311
        if from_branch is not None:
 
1312
            branch = from_branch
 
1313
        else:
 
1314
            branch = a_bzrdir.open_branch()
 
1315
        if revision_id is None:
 
1316
            revision_id = branch.last_revision()
 
1317
        local_path = transport.local_abspath('dirstate')
 
1318
        # write out new dirstate (must exist when we create the tree)
 
1319
        state = dirstate.DirState.initialize(local_path)
 
1320
        state.unlock()
 
1321
        del state
 
1322
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1323
                         branch,
 
1324
                         _format=self,
 
1325
                         _bzrdir=a_bzrdir,
 
1326
                         _control_files=control_files)
 
1327
        wt._new_tree()
 
1328
        wt.lock_tree_write()
 
1329
        try:
 
1330
            if revision_id in (None, NULL_REVISION):
 
1331
                if branch.repository.supports_rich_root():
 
1332
                    wt._set_root_id(generate_ids.gen_root_id())
 
1333
                else:
 
1334
                    wt._set_root_id(ROOT_ID)
 
1335
                wt.flush()
 
1336
            basis = None
 
1337
            # frequently, we will get here due to branching.  The accelerator
 
1338
            # tree will be the tree from the branch, so the desired basis
 
1339
            # tree will often be a parent of the accelerator tree.
 
1340
            if accelerator_tree is not None:
 
1341
                try:
 
1342
                    basis = accelerator_tree.revision_tree(revision_id)
 
1343
                except errors.NoSuchRevision:
 
1344
                    pass
 
1345
            if basis is None:
 
1346
                basis = branch.repository.revision_tree(revision_id)
 
1347
            if revision_id == NULL_REVISION:
 
1348
                parents_list = []
 
1349
            else:
 
1350
                parents_list = [(revision_id, basis)]
 
1351
            basis.lock_read()
 
1352
            try:
 
1353
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1354
                wt.flush()
 
1355
                # if the basis has a root id we have to use that; otherwise we
 
1356
                # use a new random one
 
1357
                basis_root_id = basis.get_root_id()
 
1358
                if basis_root_id is not None:
 
1359
                    wt._set_root_id(basis_root_id)
 
1360
                    wt.flush()
 
1361
                transform.build_tree(basis, wt, accelerator_tree,
 
1362
                                     hardlink=hardlink)
 
1363
            finally:
 
1364
                basis.unlock()
 
1365
        finally:
 
1366
            control_files.unlock()
 
1367
            wt.unlock()
 
1368
        return wt
 
1369
 
 
1370
    def _open(self, a_bzrdir, control_files):
 
1371
        """Open the tree itself.
 
1372
 
 
1373
        :param a_bzrdir: the dir for the tree.
 
1374
        :param control_files: the control files for the tree.
 
1375
        """
 
1376
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1377
                           branch=a_bzrdir.open_branch(),
 
1378
                           _format=self,
 
1379
                           _bzrdir=a_bzrdir,
 
1380
                           _control_files=control_files)
 
1381
 
 
1382
    def __get_matchingbzrdir(self):
 
1383
        # please test against something that will let us do tree references
 
1384
        return bzrdir.format_registry.make_bzrdir(
 
1385
            'dirstate-with-subtree')
 
1386
 
 
1387
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1388
 
 
1389
 
 
1390
class DirStateRevisionTree(Tree):
 
1391
    """A revision tree pulling the inventory from a dirstate."""
 
1392
 
 
1393
    def __init__(self, dirstate, revision_id, repository):
 
1394
        self._dirstate = dirstate
 
1395
        self._revision_id = revision_id
 
1396
        self._repository = repository
 
1397
        self._inventory = None
 
1398
        self._locked = 0
 
1399
        self._dirstate_locked = False
 
1400
 
 
1401
    def __repr__(self):
 
1402
        return "<%s of %s in %s>" % \
 
1403
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1404
 
 
1405
    def annotate_iter(self, file_id,
 
1406
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1407
        """See Tree.annotate_iter"""
 
1408
        text_key = (file_id, self.inventory[file_id].revision)
 
1409
        annotations = self._repository.texts.annotate(text_key)
 
1410
        return [(key[-1], line) for key, line in annotations]
 
1411
 
 
1412
    def _get_ancestors(self, default_revision):
 
1413
        return set(self._repository.get_ancestry(self._revision_id,
 
1414
                                                 topo_sorted=False))
 
1415
    def _comparison_data(self, entry, path):
 
1416
        """See Tree._comparison_data."""
 
1417
        if entry is None:
 
1418
            return None, False, None
 
1419
        # trust the entry as RevisionTree does, but this may not be
 
1420
        # sensible: the entry might not have come from us?
 
1421
        return entry.kind, entry.executable, None
 
1422
 
 
1423
    def _file_size(self, entry, stat_value):
 
1424
        return entry.text_size
 
1425
 
 
1426
    def filter_unversioned_files(self, paths):
 
1427
        """Filter out paths that are not versioned.
 
1428
 
 
1429
        :return: set of paths.
 
1430
        """
 
1431
        pred = self.has_filename
 
1432
        return set((p for p in paths if not pred(p)))
 
1433
 
 
1434
    def get_root_id(self):
 
1435
        return self.path2id('')
 
1436
 
 
1437
    def id2path(self, file_id):
 
1438
        "Convert a file-id to a path."
 
1439
        entry = self._get_entry(file_id=file_id)
 
1440
        if entry == (None, None):
 
1441
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1442
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1443
        return path_utf8.decode('utf8')
 
1444
 
 
1445
    def _get_parent_index(self):
 
1446
        """Return the index in the dirstate referenced by this tree."""
 
1447
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1448
 
 
1449
    def _get_entry(self, file_id=None, path=None):
 
1450
        """Get the dirstate row for file_id or path.
 
1451
 
 
1452
        If either file_id or path is supplied, it is used as the key to lookup.
 
1453
        If both are supplied, the fastest lookup is used, and an error is
 
1454
        raised if they do not both point at the same row.
 
1455
        
 
1456
        :param file_id: An optional unicode file_id to be looked up.
 
1457
        :param path: An optional unicode path to be looked up.
 
1458
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1459
        """
 
1460
        if file_id is None and path is None:
 
1461
            raise errors.BzrError('must supply file_id or path')
 
1462
        if path is not None:
 
1463
            path = path.encode('utf8')
 
1464
        parent_index = self._get_parent_index()
 
1465
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1466
 
 
1467
    def _generate_inventory(self):
 
1468
        """Create and set self.inventory from the dirstate object.
 
1469
 
 
1470
        (So this is only called the first time the inventory is requested for
 
1471
        this tree; it then remains in memory until it's out of date.)
 
1472
 
 
1473
        This is relatively expensive: we have to walk the entire dirstate.
 
1474
        """
 
1475
        if not self._locked:
 
1476
            raise AssertionError(
 
1477
                'cannot generate inventory of an unlocked '
 
1478
                'dirstate revision tree')
 
1479
        # separate call for profiling - makes it clear where the costs are.
 
1480
        self._dirstate._read_dirblocks_if_needed()
 
1481
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1482
            raise AssertionError(
 
1483
                'parent %s has disappeared from %s' % (
 
1484
                self._revision_id, self._dirstate.get_parent_ids()))
 
1485
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1486
        # This is identical now to the WorkingTree _generate_inventory except
 
1487
        # for the tree index use.
 
1488
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1489
        current_id = root_key[2]
 
1490
        if current_entry[parent_index][0] != 'd':
 
1491
            raise AssertionError()
 
1492
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1493
        inv.root.revision = current_entry[parent_index][4]
 
1494
        # Turn some things into local variables
 
1495
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1496
        factory = entry_factory
 
1497
        utf8_decode = cache_utf8._utf8_decode
 
1498
        inv_byid = inv._byid
 
1499
        # we could do this straight out of the dirstate; it might be fast
 
1500
        # and should be profiled - RBC 20070216
 
1501
        parent_ies = {'' : inv.root}
 
1502
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1503
            dirname = block[0]
 
1504
            try:
 
1505
                parent_ie = parent_ies[dirname]
 
1506
            except KeyError:
 
1507
                # all the paths in this block are not versioned in this tree
 
1508
                continue
 
1509
            for key, entry in block[1]:
 
1510
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1511
                if minikind in ('a', 'r'): # absent, relocated
 
1512
                    # not this tree
 
1513
                    continue
 
1514
                name = key[1]
 
1515
                name_unicode = utf8_decode(name)[0]
 
1516
                file_id = key[2]
 
1517
                kind = minikind_to_kind[minikind]
 
1518
                inv_entry = factory[kind](file_id, name_unicode,
 
1519
                                          parent_ie.file_id)
 
1520
                inv_entry.revision = revid
 
1521
                if kind == 'file':
 
1522
                    inv_entry.executable = executable
 
1523
                    inv_entry.text_size = size
 
1524
                    inv_entry.text_sha1 = fingerprint
 
1525
                elif kind == 'directory':
 
1526
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1527
                elif kind == 'symlink':
 
1528
                    inv_entry.executable = False
 
1529
                    inv_entry.text_size = None
 
1530
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1531
                elif kind == 'tree-reference':
 
1532
                    inv_entry.reference_revision = fingerprint or None
 
1533
                else:
 
1534
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1535
                            % entry)
 
1536
                # These checks cost us around 40ms on a 55k entry tree
 
1537
                if file_id in inv_byid:
 
1538
                    raise AssertionError('file_id %s already in'
 
1539
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1540
                if name_unicode in parent_ie.children:
 
1541
                    raise AssertionError('name %r already in parent'
 
1542
                        % (name_unicode,))
 
1543
                inv_byid[file_id] = inv_entry
 
1544
                parent_ie.children[name_unicode] = inv_entry
 
1545
        self._inventory = inv
 
1546
 
 
1547
    def get_file_mtime(self, file_id, path=None):
 
1548
        """Return the modification time for this record.
 
1549
 
 
1550
        We return the timestamp of the last-changed revision.
 
1551
        """
 
1552
        # Make sure the file exists
 
1553
        entry = self._get_entry(file_id, path=path)
 
1554
        if entry == (None, None): # do we raise?
 
1555
            return None
 
1556
        parent_index = self._get_parent_index()
 
1557
        last_changed_revision = entry[1][parent_index][4]
 
1558
        return self._repository.get_revision(last_changed_revision).timestamp
 
1559
 
 
1560
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1561
        entry = self._get_entry(file_id=file_id, path=path)
 
1562
        parent_index = self._get_parent_index()
 
1563
        parent_details = entry[1][parent_index]
 
1564
        if parent_details[0] == 'f':
 
1565
            return parent_details[1]
 
1566
        return None
 
1567
 
 
1568
    def get_file(self, file_id, path=None):
 
1569
        return StringIO(self.get_file_text(file_id))
 
1570
 
 
1571
    def get_file_lines(self, file_id):
 
1572
        return osutils.split_lines(self.get_file_text(file_id))
 
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 list(self.iter_files_bytes([(file_id, None)]))[0][1]
 
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 is 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
        if not (self.source._revision_id in parent_ids
 
1803
                or self.source._revision_id == NULL_REVISION):
 
1804
            raise AssertionError(
 
1805
                "revision {%s} is not stored in {%s}, but %s "
 
1806
                "can only be used for trees stored in the dirstate"
 
1807
                % (self.source._revision_id, self.target, self.iter_changes))
 
1808
        target_index = 0
 
1809
        if self.source._revision_id == NULL_REVISION:
 
1810
            source_index = None
 
1811
            indices = (target_index,)
 
1812
        else:
 
1813
            if not (self.source._revision_id in parent_ids):
 
1814
                raise AssertionError(
 
1815
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1816
                    self.source._revision_id, parent_ids))
 
1817
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1818
            indices = (source_index, target_index)
 
1819
        # -- make all specific_files utf8 --
 
1820
        if specific_files:
 
1821
            specific_files_utf8 = set()
 
1822
            for path in specific_files:
 
1823
                specific_files_utf8.add(path.encode('utf8'))
 
1824
            specific_files = specific_files_utf8
 
1825
        else:
 
1826
            specific_files = set([''])
 
1827
        # -- specific_files is now a utf8 path set --
 
1828
        # -- get the state object and prepare it.
 
1829
        state = self.target.current_dirstate()
 
1830
        state._read_dirblocks_if_needed()
 
1831
        def _entries_for_path(path):
 
1832
            """Return a list with all the entries that match path for all ids.
 
1833
            """
 
1834
            dirname, basename = os.path.split(path)
 
1835
            key = (dirname, basename, '')
 
1836
            block_index, present = state._find_block_index_from_key(key)
 
1837
            if not present:
 
1838
                # the block which should contain path is absent.
 
1839
                return []
 
1840
            result = []
 
1841
            block = state._dirblocks[block_index][1]
 
1842
            entry_index, _ = state._find_entry_index(key, block)
 
1843
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1844
            while (entry_index < len(block) and
 
1845
                block[entry_index][0][0:2] == key[0:2]):
 
1846
                result.append(block[entry_index])
 
1847
                entry_index += 1
 
1848
            return result
 
1849
        if require_versioned:
 
1850
            # -- check all supplied paths are versioned in a search tree. --
 
1851
            all_versioned = True
 
1852
            for path in specific_files:
 
1853
                path_entries = _entries_for_path(path)
 
1854
                if not path_entries:
 
1855
                    # this specified path is not present at all: error
 
1856
                    all_versioned = False
 
1857
                    break
 
1858
                found_versioned = False
 
1859
                # for each id at this path
 
1860
                for entry in path_entries:
 
1861
                    # for each tree.
 
1862
                    for index in indices:
 
1863
                        if entry[1][index][0] != 'a': # absent
 
1864
                            found_versioned = True
 
1865
                            # all good: found a versioned cell
 
1866
                            break
 
1867
                if not found_versioned:
 
1868
                    # none of the indexes was not 'absent' at all ids for this
 
1869
                    # path.
 
1870
                    all_versioned = False
 
1871
                    break
 
1872
            if not all_versioned:
 
1873
                raise errors.PathsNotVersionedError(specific_files)
 
1874
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1875
        search_specific_files = set()
 
1876
        for path in specific_files:
 
1877
            other_specific_files = specific_files.difference(set([path]))
 
1878
            if not osutils.is_inside_any(other_specific_files, path):
 
1879
                # this is a top level path, we must check it.
 
1880
                search_specific_files.add(path)
 
1881
        # sketch: 
 
1882
        # compare source_index and target_index at or under each element of search_specific_files.
 
1883
        # follow the following comparison table. Note that we only want to do diff operations when
 
1884
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1885
        # for the target.
 
1886
        # cases:
 
1887
        # 
 
1888
        # Source | Target | disk | action
 
1889
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1890
        #        |        |      | diff check on source-target
 
1891
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1892
        #        |        |      | ???
 
1893
        #   r    |  a     |      | add source to search
 
1894
        #   r    |  a     |  a   | 
 
1895
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1896
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1897
        #   a    | fdlt   |      | add new id
 
1898
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1899
        #   a    |  a     |      | not present in either tree, skip
 
1900
        #   a    |  a     |  a   | not present in any tree, skip
 
1901
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1902
        #        |        |      | may not be selected by the users list of paths.
 
1903
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1904
        #        |        |      | may not be selected by the users list of paths.
 
1905
        #  fdlt  | fdlt   |      | content in both: diff them
 
1906
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1907
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1908
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1909
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1910
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1911
        #        |        |      | this id at the other path.
 
1912
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1913
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1914
        #        |        |      | this id at the other path.
 
1915
 
 
1916
        # for all search_indexs in each path at or under each element of
 
1917
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1918
        # relocated path as one to search if its not searched already. If the
 
1919
        # detail is not relocated, add the id.
 
1920
        searched_specific_files = set()
 
1921
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1922
        # Using a list so that we can access the values and change them in
 
1923
        # nested scope. Each one is [path, file_id, entry]
 
1924
        last_source_parent = [None, None]
 
1925
        last_target_parent = [None, None]
 
1926
 
 
1927
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1928
 
 
1929
        # Just a sentry, so that _process_entry can say that this
 
1930
        # record is handled, but isn't interesting to process (unchanged)
 
1931
        uninteresting = object()
 
1932
 
 
1933
 
 
1934
        old_dirname_to_file_id = {}
 
1935
        new_dirname_to_file_id = {}
 
1936
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1937
        #       keeping a cache of directories that we have seen.
 
1938
 
 
1939
        def _process_entry(entry, path_info):
 
1940
            """Compare an entry and real disk to generate delta information.
 
1941
 
 
1942
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1943
                the path of entry. If None, then the path is considered absent.
 
1944
                (Perhaps we should pass in a concrete entry for this ?)
 
1945
                Basename is returned as a utf8 string because we expect this
 
1946
                tuple will be ignored, and don't want to take the time to
 
1947
                decode.
 
1948
            :return: None if these don't match
 
1949
                     A tuple of information about the change, or
 
1950
                     the object 'uninteresting' if these match, but are
 
1951
                     basically identical.
 
1952
            """
 
1953
            if source_index is None:
 
1954
                source_details = NULL_PARENT_DETAILS
 
1955
            else:
 
1956
                source_details = entry[1][source_index]
 
1957
            target_details = entry[1][target_index]
 
1958
            target_minikind = target_details[0]
 
1959
            if path_info is not None and target_minikind in 'fdlt':
 
1960
                if not (target_index == 0):
 
1961
                    raise AssertionError()
 
1962
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1963
                                                  stat_value=path_info[3])
 
1964
                # The entry may have been modified by update_entry
 
1965
                target_details = entry[1][target_index]
 
1966
                target_minikind = target_details[0]
 
1967
            else:
 
1968
                link_or_sha1 = None
 
1969
            file_id = entry[0][2]
 
1970
            source_minikind = source_details[0]
 
1971
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1972
                # claimed content in both: diff
 
1973
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1974
                #        |        |      | diff check on source-target
 
1975
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1976
                #        |        |      | ???
 
1977
                if source_minikind in 'r':
 
1978
                    # add the source to the search path to find any children it
 
1979
                    # has.  TODO ? : only add if it is a container ?
 
1980
                    if not osutils.is_inside_any(searched_specific_files,
 
1981
                                                 source_details[1]):
 
1982
                        search_specific_files.add(source_details[1])
 
1983
                    # generate the old path; this is needed for stating later
 
1984
                    # as well.
 
1985
                    old_path = source_details[1]
 
1986
                    old_dirname, old_basename = os.path.split(old_path)
 
1987
                    path = pathjoin(entry[0][0], entry[0][1])
 
1988
                    old_entry = state._get_entry(source_index,
 
1989
                                                 path_utf8=old_path)
 
1990
                    # update the source details variable to be the real
 
1991
                    # location.
 
1992
                    if old_entry == (None, None):
 
1993
                        raise errors.CorruptDirstate(state._filename,
 
1994
                            "entry '%s/%s' is considered renamed from %r"
 
1995
                            " but source does not exist\n"
 
1996
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
1997
                    source_details = old_entry[1][source_index]
 
1998
                    source_minikind = source_details[0]
 
1999
                else:
 
2000
                    old_dirname = entry[0][0]
 
2001
                    old_basename = entry[0][1]
 
2002
                    old_path = path = None
 
2003
                if path_info is None:
 
2004
                    # the file is missing on disk, show as removed.
 
2005
                    content_change = True
 
2006
                    target_kind = None
 
2007
                    target_exec = False
 
2008
                else:
 
2009
                    # source and target are both versioned and disk file is present.
 
2010
                    target_kind = path_info[2]
 
2011
                    if target_kind == 'directory':
 
2012
                        if path is None:
 
2013
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2014
                        new_dirname_to_file_id[path] = file_id
 
2015
                        if source_minikind != 'd':
 
2016
                            content_change = True
 
2017
                        else:
 
2018
                            # directories have no fingerprint
 
2019
                            content_change = False
 
2020
                        target_exec = False
 
2021
                    elif target_kind == 'file':
 
2022
                        if source_minikind != 'f':
 
2023
                            content_change = True
 
2024
                        else:
 
2025
                            # We could check the size, but we already have the
 
2026
                            # sha1 hash.
 
2027
                            content_change = (link_or_sha1 != source_details[1])
 
2028
                        # Target details is updated at update_entry time
 
2029
                        if use_filesystem_for_exec:
 
2030
                            # We don't need S_ISREG here, because we are sure
 
2031
                            # we are dealing with a file.
 
2032
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2033
                        else:
 
2034
                            target_exec = target_details[3]
 
2035
                    elif target_kind == 'symlink':
 
2036
                        if source_minikind != 'l':
 
2037
                            content_change = True
 
2038
                        else:
 
2039
                            content_change = (link_or_sha1 != source_details[1])
 
2040
                        target_exec = False
 
2041
                    elif target_kind == 'tree-reference':
 
2042
                        if source_minikind != 't':
 
2043
                            content_change = True
 
2044
                        else:
 
2045
                            content_change = False
 
2046
                        target_exec = False
 
2047
                    else:
 
2048
                        raise Exception, "unknown kind %s" % path_info[2]
 
2049
                if source_minikind == 'd':
 
2050
                    if path is None:
 
2051
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2052
                    old_dirname_to_file_id[old_path] = file_id
 
2053
                # parent id is the entry for the path in the target tree
 
2054
                if old_dirname == last_source_parent[0]:
 
2055
                    source_parent_id = last_source_parent[1]
 
2056
                else:
 
2057
                    try:
 
2058
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2059
                    except KeyError:
 
2060
                        source_parent_entry = state._get_entry(source_index,
 
2061
                                                               path_utf8=old_dirname)
 
2062
                        source_parent_id = source_parent_entry[0][2]
 
2063
                    if source_parent_id == entry[0][2]:
 
2064
                        # This is the root, so the parent is None
 
2065
                        source_parent_id = None
 
2066
                    else:
 
2067
                        last_source_parent[0] = old_dirname
 
2068
                        last_source_parent[1] = source_parent_id
 
2069
                new_dirname = entry[0][0]
 
2070
                if new_dirname == last_target_parent[0]:
 
2071
                    target_parent_id = last_target_parent[1]
 
2072
                else:
 
2073
                    try:
 
2074
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2075
                    except KeyError:
 
2076
                        # TODO: We don't always need to do the lookup, because the
 
2077
                        #       parent entry will be the same as the source entry.
 
2078
                        target_parent_entry = state._get_entry(target_index,
 
2079
                                                               path_utf8=new_dirname)
 
2080
                        if target_parent_entry == (None, None):
 
2081
                            raise AssertionError(
 
2082
                                "Could not find target parent in wt: %s\nparent of: %s"
 
2083
                                % (new_dirname, entry))
 
2084
                        target_parent_id = target_parent_entry[0][2]
 
2085
                    if target_parent_id == entry[0][2]:
 
2086
                        # This is the root, so the parent is None
 
2087
                        target_parent_id = None
 
2088
                    else:
 
2089
                        last_target_parent[0] = new_dirname
 
2090
                        last_target_parent[1] = target_parent_id
 
2091
 
 
2092
                source_exec = source_details[3]
 
2093
                if (include_unchanged
 
2094
                    or content_change
 
2095
                    or source_parent_id != target_parent_id
 
2096
                    or old_basename != entry[0][1]
 
2097
                    or source_exec != target_exec
 
2098
                    ):
 
2099
                    if old_path is None:
 
2100
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2101
                        old_path_u = utf8_decode(old_path)[0]
 
2102
                        path_u = old_path_u
 
2103
                    else:
 
2104
                        old_path_u = utf8_decode(old_path)[0]
 
2105
                        if old_path == path:
 
2106
                            path_u = old_path_u
 
2107
                        else:
 
2108
                            path_u = utf8_decode(path)[0]
 
2109
                    source_kind = _minikind_to_kind[source_minikind]
 
2110
                    return (entry[0][2],
 
2111
                           (old_path_u, path_u),
 
2112
                           content_change,
 
2113
                           (True, True),
 
2114
                           (source_parent_id, target_parent_id),
 
2115
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2116
                           (source_kind, target_kind),
 
2117
                           (source_exec, target_exec))
 
2118
                else:
 
2119
                    return uninteresting
 
2120
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2121
                # looks like a new file
 
2122
                if path_info is not None:
 
2123
                    path = pathjoin(entry[0][0], entry[0][1])
 
2124
                    # parent id is the entry for the path in the target tree
 
2125
                    # TODO: these are the same for an entire directory: cache em.
 
2126
                    parent_id = state._get_entry(target_index,
 
2127
                                                 path_utf8=entry[0][0])[0][2]
 
2128
                    if parent_id == entry[0][2]:
 
2129
                        parent_id = None
 
2130
                    if use_filesystem_for_exec:
 
2131
                        # We need S_ISREG here, because we aren't sure if this
 
2132
                        # is a file or not.
 
2133
                        target_exec = bool(
 
2134
                            stat.S_ISREG(path_info[3].st_mode)
 
2135
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2136
                    else:
 
2137
                        target_exec = target_details[3]
 
2138
                    return (entry[0][2],
 
2139
                           (None, utf8_decode(path)[0]),
 
2140
                           True,
 
2141
                           (False, True),
 
2142
                           (None, parent_id),
 
2143
                           (None, utf8_decode(entry[0][1])[0]),
 
2144
                           (None, path_info[2]),
 
2145
                           (None, target_exec))
 
2146
                else:
 
2147
                    # but its not on disk: we deliberately treat this as just
 
2148
                    # never-present. (Why ?! - RBC 20070224)
 
2149
                    pass
 
2150
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2151
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2152
                # if its still on disk, *and* theres no other entry at this
 
2153
                # path [we dont know this in this routine at the moment -
 
2154
                # perhaps we should change this - then it would be an unknown.
 
2155
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2156
                # parent id is the entry for the path in the target tree
 
2157
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2158
                if parent_id == entry[0][2]:
 
2159
                    parent_id = None
 
2160
                return (entry[0][2],
 
2161
                       (utf8_decode(old_path)[0], None),
 
2162
                       True,
 
2163
                       (True, False),
 
2164
                       (parent_id, None),
 
2165
                       (utf8_decode(entry[0][1])[0], None),
 
2166
                       (_minikind_to_kind[source_minikind], None),
 
2167
                       (source_details[3], None))
 
2168
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2169
                # a rename; could be a true rename, or a rename inherited from
 
2170
                # a renamed parent. TODO: handle this efficiently. Its not
 
2171
                # common case to rename dirs though, so a correct but slow
 
2172
                # implementation will do.
 
2173
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2174
                    search_specific_files.add(target_details[1])
 
2175
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2176
                # neither of the selected trees contain this file,
 
2177
                # so skip over it. This is not currently directly tested, but
 
2178
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2179
                pass
 
2180
            else:
 
2181
                raise AssertionError("don't know how to compare "
 
2182
                    "source_minikind=%r, target_minikind=%r"
 
2183
                    % (source_minikind, target_minikind))
 
2184
                ## import pdb;pdb.set_trace()
 
2185
            return None
 
2186
 
 
2187
        while search_specific_files:
 
2188
            # TODO: the pending list should be lexically sorted?  the
 
2189
            # interface doesn't require it.
 
2190
            current_root = search_specific_files.pop()
 
2191
            current_root_unicode = current_root.decode('utf8')
 
2192
            searched_specific_files.add(current_root)
 
2193
            # process the entries for this containing directory: the rest will be
 
2194
            # found by their parents recursively.
 
2195
            root_entries = _entries_for_path(current_root)
 
2196
            root_abspath = self.target.abspath(current_root_unicode)
 
2197
            try:
 
2198
                root_stat = os.lstat(root_abspath)
 
2199
            except OSError, e:
 
2200
                if e.errno == errno.ENOENT:
 
2201
                    # the path does not exist: let _process_entry know that.
 
2202
                    root_dir_info = None
 
2203
                else:
 
2204
                    # some other random error: hand it up.
 
2205
                    raise
 
2206
            else:
 
2207
                root_dir_info = ('', current_root,
 
2208
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2209
                    root_abspath)
 
2210
                if root_dir_info[2] == 'directory':
 
2211
                    if self.target._directory_is_tree_reference(
 
2212
                        current_root.decode('utf8')):
 
2213
                        root_dir_info = root_dir_info[:2] + \
 
2214
                            ('tree-reference',) + root_dir_info[3:]
 
2215
 
 
2216
            if not root_entries and not root_dir_info:
 
2217
                # this specified path is not present at all, skip it.
 
2218
                continue
 
2219
            path_handled = False
 
2220
            for entry in root_entries:
 
2221
                result = _process_entry(entry, root_dir_info)
 
2222
                if result is not None:
 
2223
                    path_handled = True
 
2224
                    if result is not uninteresting:
 
2225
                        yield result
 
2226
            if want_unversioned and not path_handled and root_dir_info:
 
2227
                new_executable = bool(
 
2228
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2229
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2230
                yield (None,
 
2231
                       (None, current_root_unicode),
 
2232
                       True,
 
2233
                       (False, False),
 
2234
                       (None, None),
 
2235
                       (None, splitpath(current_root_unicode)[-1]),
 
2236
                       (None, root_dir_info[2]),
 
2237
                       (None, new_executable)
 
2238
                      )
 
2239
            initial_key = (current_root, '', '')
 
2240
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2241
            if block_index == 0:
 
2242
                # we have processed the total root already, but because the
 
2243
                # initial key matched it we should skip it here.
 
2244
                block_index +=1
 
2245
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2246
                current_dir_info = None
 
2247
            else:
 
2248
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2249
                try:
 
2250
                    current_dir_info = dir_iterator.next()
 
2251
                except OSError, e:
 
2252
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2253
                    # python 2.5 has e.errno == EINVAL,
 
2254
                    #            and e.winerror == ERROR_DIRECTORY
 
2255
                    e_winerror = getattr(e, 'winerror', None)
 
2256
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2257
                    # there may be directories in the inventory even though
 
2258
                    # this path is not a file on disk: so mark it as end of
 
2259
                    # iterator
 
2260
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2261
                        current_dir_info = None
 
2262
                    elif (sys.platform == 'win32'
 
2263
                          and (e.errno in win_errors
 
2264
                               or e_winerror in win_errors)):
 
2265
                        current_dir_info = None
 
2266
                    else:
 
2267
                        raise
 
2268
                else:
 
2269
                    if current_dir_info[0][0] == '':
 
2270
                        # remove .bzr from iteration
 
2271
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2272
                        if current_dir_info[1][bzr_index][0] != '.bzr':
 
2273
                            raise AssertionError()
 
2274
                        del current_dir_info[1][bzr_index]
 
2275
            # walk until both the directory listing and the versioned metadata
 
2276
            # are exhausted. 
 
2277
            if (block_index < len(state._dirblocks) and
 
2278
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2279
                current_block = state._dirblocks[block_index]
 
2280
            else:
 
2281
                current_block = None
 
2282
            while (current_dir_info is not None or
 
2283
                   current_block is not None):
 
2284
                if (current_dir_info and current_block
 
2285
                    and current_dir_info[0][0] != current_block[0]):
 
2286
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2287
                        # filesystem data refers to paths not covered by the dirblock.
 
2288
                        # this has two possibilities:
 
2289
                        # A) it is versioned but empty, so there is no block for it
 
2290
                        # B) it is not versioned.
 
2291
 
 
2292
                        # if (A) then we need to recurse into it to check for
 
2293
                        # new unknown files or directories.
 
2294
                        # if (B) then we should ignore it, because we don't
 
2295
                        # recurse into unknown directories.
 
2296
                        path_index = 0
 
2297
                        while path_index < len(current_dir_info[1]):
 
2298
                                current_path_info = current_dir_info[1][path_index]
 
2299
                                if want_unversioned:
 
2300
                                    if current_path_info[2] == 'directory':
 
2301
                                        if self.target._directory_is_tree_reference(
 
2302
                                            current_path_info[0].decode('utf8')):
 
2303
                                            current_path_info = current_path_info[:2] + \
 
2304
                                                ('tree-reference',) + current_path_info[3:]
 
2305
                                    new_executable = bool(
 
2306
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2307
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2308
                                    yield (None,
 
2309
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2310
                                        True,
 
2311
                                        (False, False),
 
2312
                                        (None, None),
 
2313
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2314
                                        (None, current_path_info[2]),
 
2315
                                        (None, new_executable))
 
2316
                                # dont descend into this unversioned path if it is
 
2317
                                # a dir
 
2318
                                if current_path_info[2] in ('directory',
 
2319
                                                            'tree-reference'):
 
2320
                                    del current_dir_info[1][path_index]
 
2321
                                    path_index -= 1
 
2322
                                path_index += 1
 
2323
 
 
2324
                        # This dir info has been handled, go to the next
 
2325
                        try:
 
2326
                            current_dir_info = dir_iterator.next()
 
2327
                        except StopIteration:
 
2328
                            current_dir_info = None
 
2329
                    else:
 
2330
                        # We have a dirblock entry for this location, but there
 
2331
                        # is no filesystem path for this. This is most likely
 
2332
                        # because a directory was removed from the disk.
 
2333
                        # We don't have to report the missing directory,
 
2334
                        # because that should have already been handled, but we
 
2335
                        # need to handle all of the files that are contained
 
2336
                        # within.
 
2337
                        for current_entry in current_block[1]:
 
2338
                            # entry referring to file not present on disk.
 
2339
                            # advance the entry only, after processing.
 
2340
                            result = _process_entry(current_entry, None)
 
2341
                            if result is not None:
 
2342
                                if result is not uninteresting:
 
2343
                                    yield result
 
2344
                        block_index +=1
 
2345
                        if (block_index < len(state._dirblocks) and
 
2346
                            osutils.is_inside(current_root,
 
2347
                                              state._dirblocks[block_index][0])):
 
2348
                            current_block = state._dirblocks[block_index]
 
2349
                        else:
 
2350
                            current_block = None
 
2351
                    continue
 
2352
                entry_index = 0
 
2353
                if current_block and entry_index < len(current_block[1]):
 
2354
                    current_entry = current_block[1][entry_index]
 
2355
                else:
 
2356
                    current_entry = None
 
2357
                advance_entry = True
 
2358
                path_index = 0
 
2359
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2360
                    current_path_info = current_dir_info[1][path_index]
 
2361
                    if current_path_info[2] == 'directory':
 
2362
                        if self.target._directory_is_tree_reference(
 
2363
                            current_path_info[0].decode('utf8')):
 
2364
                            current_path_info = current_path_info[:2] + \
 
2365
                                ('tree-reference',) + current_path_info[3:]
 
2366
                else:
 
2367
                    current_path_info = None
 
2368
                advance_path = True
 
2369
                path_handled = False
 
2370
                while (current_entry is not None or
 
2371
                    current_path_info is not None):
 
2372
                    if current_entry is None:
 
2373
                        # the check for path_handled when the path is adnvaced
 
2374
                        # will yield this path if needed.
 
2375
                        pass
 
2376
                    elif current_path_info is None:
 
2377
                        # no path is fine: the per entry code will handle it.
 
2378
                        result = _process_entry(current_entry, current_path_info)
 
2379
                        if result is not None:
 
2380
                            if result is not uninteresting:
 
2381
                                yield result
 
2382
                    elif (current_entry[0][1] != current_path_info[1]
 
2383
                          or current_entry[1][target_index][0] in 'ar'):
 
2384
                        # The current path on disk doesn't match the dirblock
 
2385
                        # record. Either the dirblock is marked as absent, or
 
2386
                        # the file on disk is not present at all in the
 
2387
                        # dirblock. Either way, report about the dirblock
 
2388
                        # entry, and let other code handle the filesystem one.
 
2389
 
 
2390
                        # Compare the basename for these files to determine
 
2391
                        # which comes first
 
2392
                        if current_path_info[1] < current_entry[0][1]:
 
2393
                            # extra file on disk: pass for now, but only
 
2394
                            # increment the path, not the entry
 
2395
                            advance_entry = False
 
2396
                        else:
 
2397
                            # entry referring to file not present on disk.
 
2398
                            # advance the entry only, after processing.
 
2399
                            result = _process_entry(current_entry, None)
 
2400
                            if result is not None:
 
2401
                                if result is not uninteresting:
 
2402
                                    yield result
 
2403
                            advance_path = False
 
2404
                    else:
 
2405
                        result = _process_entry(current_entry, current_path_info)
 
2406
                        if result is not None:
 
2407
                            path_handled = True
 
2408
                            if result is not uninteresting:
 
2409
                                yield result
 
2410
                    if advance_entry and current_entry is not None:
 
2411
                        entry_index += 1
 
2412
                        if entry_index < len(current_block[1]):
 
2413
                            current_entry = current_block[1][entry_index]
 
2414
                        else:
 
2415
                            current_entry = None
 
2416
                    else:
 
2417
                        advance_entry = True # reset the advance flaga
 
2418
                    if advance_path and current_path_info is not None:
 
2419
                        if not path_handled:
 
2420
                            # unversioned in all regards
 
2421
                            if want_unversioned:
 
2422
                                new_executable = bool(
 
2423
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2424
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2425
                                yield (None,
 
2426
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2427
                                    True,
 
2428
                                    (False, False),
 
2429
                                    (None, None),
 
2430
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2431
                                    (None, current_path_info[2]),
 
2432
                                    (None, new_executable))
 
2433
                            # dont descend into this unversioned path if it is
 
2434
                            # a dir
 
2435
                            if current_path_info[2] in ('directory'):
 
2436
                                del current_dir_info[1][path_index]
 
2437
                                path_index -= 1
 
2438
                        # dont descend the disk iterator into any tree 
 
2439
                        # paths.
 
2440
                        if current_path_info[2] == 'tree-reference':
 
2441
                            del current_dir_info[1][path_index]
 
2442
                            path_index -= 1
 
2443
                        path_index += 1
 
2444
                        if path_index < len(current_dir_info[1]):
 
2445
                            current_path_info = current_dir_info[1][path_index]
 
2446
                            if current_path_info[2] == 'directory':
 
2447
                                if self.target._directory_is_tree_reference(
 
2448
                                    current_path_info[0].decode('utf8')):
 
2449
                                    current_path_info = current_path_info[:2] + \
 
2450
                                        ('tree-reference',) + current_path_info[3:]
 
2451
                        else:
 
2452
                            current_path_info = None
 
2453
                        path_handled = False
 
2454
                    else:
 
2455
                        advance_path = True # reset the advance flagg.
 
2456
                if current_block is not None:
 
2457
                    block_index += 1
 
2458
                    if (block_index < len(state._dirblocks) and
 
2459
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2460
                        current_block = state._dirblocks[block_index]
 
2461
                    else:
 
2462
                        current_block = None
 
2463
                if current_dir_info is not None:
 
2464
                    try:
 
2465
                        current_dir_info = dir_iterator.next()
 
2466
                    except StopIteration:
 
2467
                        current_dir_info = None
 
2468
 
 
2469
 
 
2470
    @staticmethod
 
2471
    def is_compatible(source, target):
 
2472
        # the target must be a dirstate working tree
 
2473
        if not isinstance(target, WorkingTree4):
 
2474
            return False
 
2475
        # the source must be a revtreee or dirstate rev tree.
 
2476
        if not isinstance(source,
 
2477
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2478
            return False
 
2479
        # the source revid must be in the target dirstate
 
2480
        if not (source._revision_id == NULL_REVISION or
 
2481
            source._revision_id in target.get_parent_ids()):
 
2482
            # TODO: what about ghosts? it may well need to 
 
2483
            # check for them explicitly.
 
2484
            return False
 
2485
        return True
 
2486
 
 
2487
InterTree.register_optimiser(InterDirStateTree)
 
2488
 
 
2489
 
 
2490
class Converter3to4(object):
 
2491
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2492
 
 
2493
    def __init__(self):
 
2494
        self.target_format = WorkingTreeFormat4()
 
2495
 
 
2496
    def convert(self, tree):
 
2497
        # lock the control files not the tree, so that we dont get tree
 
2498
        # on-unlock behaviours, and so that noone else diddles with the 
 
2499
        # tree during upgrade.
 
2500
        tree._control_files.lock_write()
 
2501
        try:
 
2502
            tree.read_working_inventory()
 
2503
            self.create_dirstate_data(tree)
 
2504
            self.update_format(tree)
 
2505
            self.remove_xml_files(tree)
 
2506
        finally:
 
2507
            tree._control_files.unlock()
 
2508
 
 
2509
    def create_dirstate_data(self, tree):
 
2510
        """Create the dirstate based data for tree."""
 
2511
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2512
            ).local_abspath('dirstate')
 
2513
        state = dirstate.DirState.from_tree(tree, local_path)
 
2514
        state.save()
 
2515
        state.unlock()
 
2516
 
 
2517
    def remove_xml_files(self, tree):
 
2518
        """Remove the oldformat 3 data."""
 
2519
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2520
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2521
            'pending-merges', 'stat-cache']:
 
2522
            try:
 
2523
                transport.delete(path)
 
2524
            except errors.NoSuchFile:
 
2525
                # some files are optional - just deal.
 
2526
                pass
 
2527
 
 
2528
    def update_format(self, tree):
 
2529
        """Change the format marker."""
 
2530
        tree._control_files.put_utf8('format',
 
2531
            self.target_format.get_format_string())