/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Alexander Belchenko
  • Date: 2007-09-27 12:45:51 UTC
  • mto: (2881.2.1 trivial)
  • mto: This revision was merged to the branch mainline in revision 2883.
  • Revision ID: bialix@ukr.net-20070927124551-q3vv29cy3ubidvk1
check bzrlib version early

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    debug,
 
47
    delta,
 
48
    dirstate,
 
49
    errors,
 
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    trace,
 
59
    transform,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
 
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
 
67
""")
 
68
 
 
69
from bzrlib import symbol_versioning
 
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
 
76
from bzrlib.osutils import (
 
77
    file_kind,
 
78
    isdir,
 
79
    normpath,
 
80
    pathjoin,
 
81
    rand_chars,
 
82
    realpath,
 
83
    safe_unicode,
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
 
87
from bzrlib.transport.local import LocalTransport
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
 
120
    def __init__(self, basedir,
 
121
                 branch,
 
122
                 _control_files=None,
 
123
                 _format=None,
 
124
                 _bzrdir=None):
 
125
        """Construct a WorkingTree for basedir.
 
126
 
 
127
        If the branch is not supplied, it is opened automatically.
 
128
        If the branch is supplied, it must be the branch for this basedir.
 
129
        (branch.base is not cross checked, because for remote branches that
 
130
        would be meaningless).
 
131
        """
 
132
        self._format = _format
 
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
 
136
        basedir = safe_unicode(basedir)
 
137
        mutter("opening working tree %r", basedir)
 
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
 
141
        self.basedir = realpath(basedir)
 
142
        # if branch is at our basedir and is a format 6 or less
 
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
 
146
        self._control_files = _control_files
 
147
        self._dirty = None
 
148
        #-------------
 
149
        # during a read or write lock these objects are set, and are
 
150
        # None the rest of the time.
 
151
        self._dirstate = None
 
152
        self._inventory = None
 
153
        #-------------
 
154
 
 
155
    @needs_tree_write_lock
 
156
    def _add(self, files, ids, kinds):
 
157
        """See MutableTree._add."""
 
158
        state = self.current_dirstate()
 
159
        for f, file_id, kind in zip(files, ids, kinds):
 
160
            f = f.strip('/')
 
161
            assert '//' not in f
 
162
            assert '..' not in f
 
163
            if self.path2id(f):
 
164
                # special case tree root handling.
 
165
                if f == '' and self.path2id(f) == ROOT_ID:
 
166
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
167
                continue
 
168
            if file_id is None:
 
169
                file_id = generate_ids.gen_file_id(f)
 
170
            # deliberately add the file with no cached stat or sha1
 
171
            # - on the first access it will be gathered, and we can
 
172
            # always change this once tests are all passing.
 
173
            state.add(f, file_id, kind, None, '')
 
174
        self._make_dirty(reset_inventory=True)
 
175
 
 
176
    def _make_dirty(self, reset_inventory):
 
177
        """Make the tree state dirty.
 
178
 
 
179
        :param reset_inventory: True if the cached inventory should be removed
 
180
            (presuming there is one).
 
181
        """
 
182
        self._dirty = True
 
183
        if reset_inventory and self._inventory is not None:
 
184
            self._inventory = None
 
185
 
 
186
    @needs_tree_write_lock
 
187
    def add_reference(self, sub_tree):
 
188
        # use standard implementation, which calls back to self._add
 
189
        # 
 
190
        # So we don't store the reference_revision in the working dirstate,
 
191
        # it's just recorded at the moment of commit. 
 
192
        self._add_reference(sub_tree)
 
193
 
 
194
    def break_lock(self):
 
195
        """Break a lock if one is present from another instance.
 
196
 
 
197
        Uses the ui factory to ask for confirmation if the lock may be from
 
198
        an active process.
 
199
 
 
200
        This will probe the repository for its lock as well.
 
201
        """
 
202
        # if the dirstate is locked by an active process, reject the break lock
 
203
        # call.
 
204
        try:
 
205
            if self._dirstate is None:
 
206
                clear = True
 
207
            else:
 
208
                clear = False
 
209
            state = self._current_dirstate()
 
210
            if state._lock_token is not None:
 
211
                # we already have it locked. sheese, cant break our own lock.
 
212
                raise errors.LockActive(self.basedir)
 
213
            else:
 
214
                try:
 
215
                    # try for a write lock - need permission to get one anyhow
 
216
                    # to break locks.
 
217
                    state.lock_write()
 
218
                except errors.LockContention:
 
219
                    # oslocks fail when a process is still live: fail.
 
220
                    # TODO: get the locked lockdir info and give to the user to
 
221
                    # assist in debugging.
 
222
                    raise errors.LockActive(self.basedir)
 
223
                else:
 
224
                    state.unlock()
 
225
        finally:
 
226
            if clear:
 
227
                self._dirstate = None
 
228
        self._control_files.break_lock()
 
229
        self.branch.break_lock()
 
230
 
 
231
    def _comparison_data(self, entry, path):
 
232
        kind, executable, stat_value = \
 
233
            WorkingTree3._comparison_data(self, entry, path)
 
234
        # it looks like a plain directory, but it's really a reference -- see
 
235
        # also kind()
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
 
239
            kind = 'tree-reference'
 
240
        return kind, executable, stat_value
 
241
 
 
242
    @needs_write_lock
 
243
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
244
        # mark the tree as dirty post commit - commit
 
245
        # can change the current versioned list by doing deletes.
 
246
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
247
        self._make_dirty(reset_inventory=True)
 
248
        return result
 
249
 
 
250
    def current_dirstate(self):
 
251
        """Return the current dirstate object.
 
252
 
 
253
        This is not part of the tree interface and only exposed for ease of
 
254
        testing.
 
255
 
 
256
        :raises errors.NotWriteLocked: when not in a lock.
 
257
        """
 
258
        self._must_be_locked()
 
259
        return self._current_dirstate()
 
260
 
 
261
    def _current_dirstate(self):
 
262
        """Internal function that does not check lock status.
 
263
 
 
264
        This is needed for break_lock which also needs the dirstate.
 
265
        """
 
266
        if self._dirstate is not None:
 
267
            return self._dirstate
 
268
        local_path = self.bzrdir.get_workingtree_transport(None
 
269
            ).local_abspath('dirstate')
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
 
271
        return self._dirstate
 
272
 
 
273
    def filter_unversioned_files(self, paths):
 
274
        """Filter out paths that are versioned.
 
275
 
 
276
        :return: set of paths.
 
277
        """
 
278
        # TODO: make a generic multi-bisect routine roughly that should list
 
279
        # the paths, then process one half at a time recursively, and feed the
 
280
        # results of each bisect in further still
 
281
        paths = sorted(paths)
 
282
        result = set()
 
283
        state = self.current_dirstate()
 
284
        # TODO we want a paths_to_dirblocks helper I think
 
285
        for path in paths:
 
286
            dirname, basename = os.path.split(path.encode('utf8'))
 
287
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
288
                dirname, basename, 0)
 
289
            if not path_is_versioned:
 
290
                result.add(path)
 
291
        return result
 
292
 
 
293
    def flush(self):
 
294
        """Write all cached data to disk."""
 
295
        if self._control_files._lock_mode != 'w':
 
296
            raise errors.NotWriteLocked(self)
 
297
        self.current_dirstate().save()
 
298
        self._inventory = None
 
299
        self._dirty = False
 
300
 
 
301
    @needs_tree_write_lock
 
302
    def _gather_kinds(self, files, kinds):
 
303
        """See MutableTree._gather_kinds."""
 
304
        for pos, f in enumerate(files):
 
305
            if kinds[pos] is None:
 
306
                kinds[pos] = self._kind(f)
 
307
 
 
308
    def _generate_inventory(self):
 
309
        """Create and set self.inventory from the dirstate object.
 
310
        
 
311
        This is relatively expensive: we have to walk the entire dirstate.
 
312
        Ideally we would not, and can deprecate this function.
 
313
        """
 
314
        #: uncomment to trap on inventory requests.
 
315
        # import pdb;pdb.set_trace()
 
316
        state = self.current_dirstate()
 
317
        state._read_dirblocks_if_needed()
 
318
        root_key, current_entry = self._get_entry(path='')
 
319
        current_id = root_key[2]
 
320
        assert current_entry[0][0] == 'd' # directory
 
321
        inv = Inventory(root_id=current_id)
 
322
        # Turn some things into local variables
 
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
324
        factory = entry_factory
 
325
        utf8_decode = cache_utf8._utf8_decode
 
326
        inv_byid = inv._byid
 
327
        # we could do this straight out of the dirstate; it might be fast
 
328
        # and should be profiled - RBC 20070216
 
329
        parent_ies = {'' : inv.root}
 
330
        for block in state._dirblocks[1:]: # skip the root
 
331
            dirname = block[0]
 
332
            try:
 
333
                parent_ie = parent_ies[dirname]
 
334
            except KeyError:
 
335
                # all the paths in this block are not versioned in this tree
 
336
                continue
 
337
            for key, entry in block[1]:
 
338
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
339
                if minikind in ('a', 'r'): # absent, relocated
 
340
                    # a parent tree only entry
 
341
                    continue
 
342
                name = key[1]
 
343
                name_unicode = utf8_decode(name)[0]
 
344
                file_id = key[2]
 
345
                kind = minikind_to_kind[minikind]
 
346
                inv_entry = factory[kind](file_id, name_unicode,
 
347
                                          parent_ie.file_id)
 
348
                if kind == 'file':
 
349
                    # This is only needed on win32, where this is the only way
 
350
                    # we know the executable bit.
 
351
                    inv_entry.executable = executable
 
352
                    # not strictly needed: working tree
 
353
                    #inv_entry.text_size = size
 
354
                    #inv_entry.text_sha1 = sha1
 
355
                elif kind == 'directory':
 
356
                    # add this entry to the parent map.
 
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
358
                elif kind == 'tree-reference':
 
359
                    assert self._repo_supports_tree_reference, \
 
360
                        "repository of %r " \
 
361
                        "doesn't support tree references " \
 
362
                        "required by entry %r" \
 
363
                        % (self, name)
 
364
                    inv_entry.reference_revision = link_or_sha1 or None
 
365
                elif kind != 'symlink':
 
366
                    raise AssertionError("unknown kind %r" % kind)
 
367
                # These checks cost us around 40ms on a 55k entry tree
 
368
                assert file_id not in inv_byid, ('file_id %s already in'
 
369
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
370
                assert name_unicode not in parent_ie.children
 
371
                inv_byid[file_id] = inv_entry
 
372
                parent_ie.children[name_unicode] = inv_entry
 
373
        self._inventory = inv
 
374
 
 
375
    def _get_entry(self, file_id=None, path=None):
 
376
        """Get the dirstate row for file_id or path.
 
377
 
 
378
        If either file_id or path is supplied, it is used as the key to lookup.
 
379
        If both are supplied, the fastest lookup is used, and an error is
 
380
        raised if they do not both point at the same row.
 
381
        
 
382
        :param file_id: An optional unicode file_id to be looked up.
 
383
        :param path: An optional unicode path to be looked up.
 
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
385
        """
 
386
        if file_id is None and path is None:
 
387
            raise errors.BzrError('must supply file_id or path')
 
388
        state = self.current_dirstate()
 
389
        if path is not None:
 
390
            path = path.encode('utf8')
 
391
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
392
 
 
393
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
394
        # check file id is valid unconditionally.
 
395
        entry = self._get_entry(file_id=file_id, path=path)
 
396
        if entry[0] is None:
 
397
            raise errors.NoSuchId(self, file_id)
 
398
        if path is None:
 
399
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
400
 
 
401
        file_abspath = self.abspath(path)
 
402
        state = self.current_dirstate()
 
403
        if stat_value is None:
 
404
            try:
 
405
                stat_value = os.lstat(file_abspath)
 
406
            except OSError, e:
 
407
                if e.errno == errno.ENOENT:
 
408
                    return None
 
409
                else:
 
410
                    raise
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
 
413
        if entry[1][0][0] == 'f':
 
414
            return link_or_sha1
 
415
        return None
 
416
 
 
417
    def _get_inventory(self):
 
418
        """Get the inventory for the tree. This is only valid within a lock."""
 
419
        if 'evil' in debug.debug_flags:
 
420
            trace.mutter_callsite(2,
 
421
                "accessing .inventory forces a size of tree translation.")
 
422
        if self._inventory is not None:
 
423
            return self._inventory
 
424
        self._must_be_locked()
 
425
        self._generate_inventory()
 
426
        return self._inventory
 
427
 
 
428
    inventory = property(_get_inventory,
 
429
                         doc="Inventory of this Tree")
 
430
 
 
431
    @needs_read_lock
 
432
    def get_parent_ids(self):
 
433
        """See Tree.get_parent_ids.
 
434
        
 
435
        This implementation requests the ids list from the dirstate file.
 
436
        """
 
437
        return self.current_dirstate().get_parent_ids()
 
438
 
 
439
    def get_reference_revision(self, file_id, path=None):
 
440
        # referenced tree's revision is whatever's currently there
 
441
        return self.get_nested_tree(file_id, path).last_revision()
 
442
 
 
443
    def get_nested_tree(self, file_id, path=None):
 
444
        if path is None:
 
445
            path = self.id2path(file_id)
 
446
        # else: check file_id is at path?
 
447
        return WorkingTree.open(self.abspath(path))
 
448
 
 
449
    @needs_read_lock
 
450
    def get_root_id(self):
 
451
        """Return the id of this trees root"""
 
452
        return self._get_entry(path='')[0][2]
 
453
 
 
454
    def has_id(self, file_id):
 
455
        state = self.current_dirstate()
 
456
        file_id = osutils.safe_file_id(file_id)
 
457
        row, parents = self._get_entry(file_id=file_id)
 
458
        if row is None:
 
459
            return False
 
460
        return osutils.lexists(pathjoin(
 
461
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
462
 
 
463
    @needs_read_lock
 
464
    def id2path(self, file_id):
 
465
        "Convert a file-id to a path."
 
466
        file_id = osutils.safe_file_id(file_id)
 
467
        state = self.current_dirstate()
 
468
        entry = self._get_entry(file_id=file_id)
 
469
        if entry == (None, None):
 
470
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
471
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
472
        return path_utf8.decode('utf8')
 
473
 
 
474
    if not osutils.supports_executable():
 
475
        def is_executable(self, file_id, path=None):
 
476
            """Test if a file is executable or not.
 
477
 
 
478
            Note: The caller is expected to take a read-lock before calling this.
 
479
            """
 
480
            file_id = osutils.safe_file_id(file_id)
 
481
            entry = self._get_entry(file_id=file_id, path=path)
 
482
            if entry == (None, None):
 
483
                return False
 
484
            return entry[1][0][3]
 
485
    else:
 
486
        def is_executable(self, file_id, path=None):
 
487
            """Test if a file is executable or not.
 
488
 
 
489
            Note: The caller is expected to take a read-lock before calling this.
 
490
            """
 
491
            if not path:
 
492
                file_id = osutils.safe_file_id(file_id)
 
493
                path = self.id2path(file_id)
 
494
            mode = os.lstat(self.abspath(path)).st_mode
 
495
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
496
 
 
497
    @needs_read_lock
 
498
    def __iter__(self):
 
499
        """Iterate through file_ids for this tree.
 
500
 
 
501
        file_ids are in a WorkingTree if they are in the working inventory
 
502
        and the working file exists.
 
503
        """
 
504
        result = []
 
505
        for key, tree_details in self.current_dirstate()._iter_entries():
 
506
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
507
                # not relevant to the working tree
 
508
                continue
 
509
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
510
            if osutils.lexists(path):
 
511
                result.append(key[2])
 
512
        return iter(result)
 
513
 
 
514
    def iter_references(self):
 
515
        for key, tree_details in self.current_dirstate()._iter_entries():
 
516
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
517
                # not relevant to the working tree
 
518
                continue
 
519
            if not key[1]:
 
520
                # the root is not a reference.
 
521
                continue
 
522
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
523
            try:
 
524
                if self._kind(path) == 'tree-reference':
 
525
                    yield path, key[2]
 
526
            except errors.NoSuchFile:
 
527
                # path is missing on disk.
 
528
                continue
 
529
 
 
530
    def kind(self, file_id):
 
531
        """Return the kind of a file.
 
532
 
 
533
        This is always the actual kind that's on disk, regardless of what it
 
534
        was added as.
 
535
 
 
536
        Note: The caller is expected to take a read-lock before calling this.
 
537
        """
 
538
        relpath = self.id2path(file_id)
 
539
        assert relpath != None, \
 
540
            "path for id {%s} is None!" % file_id
 
541
        return self._kind(relpath)
 
542
 
 
543
    def _kind(self, relpath):
 
544
        abspath = self.abspath(relpath)
 
545
        kind = file_kind(abspath)
 
546
        if (self._repo_supports_tree_reference and
 
547
            kind == 'directory' and
 
548
            self._directory_is_tree_reference(relpath)):
 
549
            kind = 'tree-reference'
 
550
        return kind
 
551
 
 
552
    @needs_read_lock
 
553
    def _last_revision(self):
 
554
        """See Mutable.last_revision."""
 
555
        parent_ids = self.current_dirstate().get_parent_ids()
 
556
        if parent_ids:
 
557
            return parent_ids[0]
 
558
        else:
 
559
            return _mod_revision.NULL_REVISION
 
560
 
 
561
    def lock_read(self):
 
562
        """See Branch.lock_read, and WorkingTree.unlock."""
 
563
        self.branch.lock_read()
 
564
        try:
 
565
            self._control_files.lock_read()
 
566
            try:
 
567
                state = self.current_dirstate()
 
568
                if not state._lock_token:
 
569
                    state.lock_read()
 
570
                # set our support for tree references from the repository in
 
571
                # use.
 
572
                self._repo_supports_tree_reference = getattr(
 
573
                    self.branch.repository._format, "supports_tree_reference",
 
574
                    False)
 
575
            except:
 
576
                self._control_files.unlock()
 
577
                raise
 
578
        except:
 
579
            self.branch.unlock()
 
580
            raise
 
581
 
 
582
    def _lock_self_write(self):
 
583
        """This should be called after the branch is locked."""
 
584
        try:
 
585
            self._control_files.lock_write()
 
586
            try:
 
587
                state = self.current_dirstate()
 
588
                if not state._lock_token:
 
589
                    state.lock_write()
 
590
                # set our support for tree references from the repository in
 
591
                # use.
 
592
                self._repo_supports_tree_reference = getattr(
 
593
                    self.branch.repository._format, "supports_tree_reference",
 
594
                    False)
 
595
            except:
 
596
                self._control_files.unlock()
 
597
                raise
 
598
        except:
 
599
            self.branch.unlock()
 
600
            raise
 
601
 
 
602
    def lock_tree_write(self):
 
603
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
604
        self.branch.lock_read()
 
605
        self._lock_self_write()
 
606
 
 
607
    def lock_write(self):
 
608
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
609
        self.branch.lock_write()
 
610
        self._lock_self_write()
 
611
 
 
612
    @needs_tree_write_lock
 
613
    def move(self, from_paths, to_dir, after=False):
 
614
        """See WorkingTree.move()."""
 
615
        result = []
 
616
        if not from_paths:
 
617
            return result
 
618
 
 
619
        state = self.current_dirstate()
 
620
 
 
621
        assert not isinstance(from_paths, basestring)
 
622
        to_dir_utf8 = to_dir.encode('utf8')
 
623
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
624
        id_index = state._get_id_index()
 
625
        # check destination directory
 
626
        # get the details for it
 
627
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
628
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
629
        if not entry_present:
 
630
            raise errors.BzrMoveFailedError('', to_dir,
 
631
                errors.NotVersionedError(to_dir))
 
632
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
633
        # get a handle on the block itself.
 
634
        to_block_index = state._ensure_block(
 
635
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
636
        to_block = state._dirblocks[to_block_index]
 
637
        to_abs = self.abspath(to_dir)
 
638
        if not isdir(to_abs):
 
639
            raise errors.BzrMoveFailedError('',to_dir,
 
640
                errors.NotADirectory(to_abs))
 
641
 
 
642
        if to_entry[1][0][0] != 'd':
 
643
            raise errors.BzrMoveFailedError('',to_dir,
 
644
                errors.NotADirectory(to_abs))
 
645
 
 
646
        if self._inventory is not None:
 
647
            update_inventory = True
 
648
            inv = self.inventory
 
649
            to_dir_ie = inv[to_dir_id]
 
650
            to_dir_id = to_entry[0][2]
 
651
        else:
 
652
            update_inventory = False
 
653
 
 
654
        rollbacks = []
 
655
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
656
                     fingerprint, packed_stat, size,
 
657
                     to_block, to_key, to_path_utf8):
 
658
            state._make_absent(old_entry)
 
659
            from_key = old_entry[0]
 
660
            rollbacks.append(
 
661
                lambda:state.update_minimal(from_key,
 
662
                    minikind,
 
663
                    executable=executable,
 
664
                    fingerprint=fingerprint,
 
665
                    packed_stat=packed_stat,
 
666
                    size=size,
 
667
                    path_utf8=from_path_utf8))
 
668
            state.update_minimal(to_key,
 
669
                    minikind,
 
670
                    executable=executable,
 
671
                    fingerprint=fingerprint,
 
672
                    packed_stat=packed_stat,
 
673
                    size=size,
 
674
                    path_utf8=to_path_utf8)
 
675
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
676
            new_entry = to_block[1][added_entry_index]
 
677
            rollbacks.append(lambda:state._make_absent(new_entry))
 
678
 
 
679
        for from_rel in from_paths:
 
680
            # from_rel is 'pathinroot/foo/bar'
 
681
            from_rel_utf8 = from_rel.encode('utf8')
 
682
            from_dirname, from_tail = osutils.split(from_rel)
 
683
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
684
            from_entry = self._get_entry(path=from_rel)
 
685
            if from_entry == (None, None):
 
686
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
687
                    errors.NotVersionedError(path=str(from_rel)))
 
688
 
 
689
            from_id = from_entry[0][2]
 
690
            to_rel = pathjoin(to_dir, from_tail)
 
691
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
692
            item_to_entry = self._get_entry(path=to_rel)
 
693
            if item_to_entry != (None, None):
 
694
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
695
                    "Target is already versioned.")
 
696
 
 
697
            if from_rel == to_rel:
 
698
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
699
                    "Source and target are identical.")
 
700
 
 
701
            from_missing = not self.has_filename(from_rel)
 
702
            to_missing = not self.has_filename(to_rel)
 
703
            if after:
 
704
                move_file = False
 
705
            else:
 
706
                move_file = True
 
707
            if to_missing:
 
708
                if not move_file:
 
709
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
710
                        errors.NoSuchFile(path=to_rel,
 
711
                        extra="New file has not been created yet"))
 
712
                elif from_missing:
 
713
                    # neither path exists
 
714
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
715
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
716
            else:
 
717
                if from_missing: # implicitly just update our path mapping
 
718
                    move_file = False
 
719
                elif not after:
 
720
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
721
                        extra="(Use --after to update the Bazaar id)")
 
722
 
 
723
            rollbacks = []
 
724
            def rollback_rename():
 
725
                """A single rename has failed, roll it back."""
 
726
                # roll back everything, even if we encounter trouble doing one
 
727
                # of them.
 
728
                #
 
729
                # TODO: at least log the other exceptions rather than just
 
730
                # losing them mbp 20070307
 
731
                exc_info = None
 
732
                for rollback in reversed(rollbacks):
 
733
                    try:
 
734
                        rollback()
 
735
                    except Exception, e:
 
736
                        exc_info = sys.exc_info()
 
737
                if exc_info:
 
738
                    raise exc_info[0], exc_info[1], exc_info[2]
 
739
 
 
740
            # perform the disk move first - its the most likely failure point.
 
741
            if move_file:
 
742
                from_rel_abs = self.abspath(from_rel)
 
743
                to_rel_abs = self.abspath(to_rel)
 
744
                try:
 
745
                    osutils.rename(from_rel_abs, to_rel_abs)
 
746
                except OSError, e:
 
747
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
748
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
749
            try:
 
750
                # perform the rename in the inventory next if needed: its easy
 
751
                # to rollback
 
752
                if update_inventory:
 
753
                    # rename the entry
 
754
                    from_entry = inv[from_id]
 
755
                    current_parent = from_entry.parent_id
 
756
                    inv.rename(from_id, to_dir_id, from_tail)
 
757
                    rollbacks.append(
 
758
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
759
                # finally do the rename in the dirstate, which is a little
 
760
                # tricky to rollback, but least likely to need it.
 
761
                old_block_index, old_entry_index, dir_present, file_present = \
 
762
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
763
                old_block = state._dirblocks[old_block_index][1]
 
764
                old_entry = old_block[old_entry_index]
 
765
                from_key, old_entry_details = old_entry
 
766
                cur_details = old_entry_details[0]
 
767
                # remove the old row
 
768
                to_key = ((to_block[0],) + from_key[1:3])
 
769
                minikind = cur_details[0]
 
770
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
771
                         minikind=minikind,
 
772
                         executable=cur_details[3],
 
773
                         fingerprint=cur_details[1],
 
774
                         packed_stat=cur_details[4],
 
775
                         size=cur_details[2],
 
776
                         to_block=to_block,
 
777
                         to_key=to_key,
 
778
                         to_path_utf8=to_rel_utf8)
 
779
 
 
780
                if minikind == 'd':
 
781
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
782
                        """Recursively update all entries in this dirblock."""
 
783
                        assert from_dir != '', "renaming root not supported"
 
784
                        from_key = (from_dir, '')
 
785
                        from_block_idx, present = \
 
786
                            state._find_block_index_from_key(from_key)
 
787
                        if not present:
 
788
                            # This is the old record, if it isn't present, then
 
789
                            # there is theoretically nothing to update.
 
790
                            # (Unless it isn't present because of lazy loading,
 
791
                            # but we don't do that yet)
 
792
                            return
 
793
                        from_block = state._dirblocks[from_block_idx]
 
794
                        to_block_index, to_entry_index, _, _ = \
 
795
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
796
                        to_block_index = state._ensure_block(
 
797
                            to_block_index, to_entry_index, to_dir_utf8)
 
798
                        to_block = state._dirblocks[to_block_index]
 
799
 
 
800
                        # Grab a copy since move_one may update the list.
 
801
                        for entry in from_block[1][:]:
 
802
                            assert entry[0][0] == from_dir
 
803
                            cur_details = entry[1][0]
 
804
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
805
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
806
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
807
                            minikind = cur_details[0]
 
808
                            if minikind in 'ar':
 
809
                                # Deleted children of a renamed directory
 
810
                                # Do not need to be updated.
 
811
                                # Children that have been renamed out of this
 
812
                                # directory should also not be updated
 
813
                                continue
 
814
                            move_one(entry, from_path_utf8=from_path_utf8,
 
815
                                     minikind=minikind,
 
816
                                     executable=cur_details[3],
 
817
                                     fingerprint=cur_details[1],
 
818
                                     packed_stat=cur_details[4],
 
819
                                     size=cur_details[2],
 
820
                                     to_block=to_block,
 
821
                                     to_key=to_key,
 
822
                                     to_path_utf8=to_path_utf8)
 
823
                            if minikind == 'd':
 
824
                                # We need to move all the children of this
 
825
                                # entry
 
826
                                update_dirblock(from_path_utf8, to_key,
 
827
                                                to_path_utf8)
 
828
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
829
            except:
 
830
                rollback_rename()
 
831
                raise
 
832
            result.append((from_rel, to_rel))
 
833
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
834
            self._make_dirty(reset_inventory=False)
 
835
 
 
836
        return result
 
837
 
 
838
    def _must_be_locked(self):
 
839
        if not self._control_files._lock_count:
 
840
            raise errors.ObjectNotLocked(self)
 
841
 
 
842
    def _new_tree(self):
 
843
        """Initialize the state in this tree to be a new tree."""
 
844
        self._dirty = True
 
845
 
 
846
    @needs_read_lock
 
847
    def path2id(self, path):
 
848
        """Return the id for path in this tree."""
 
849
        path = path.strip('/')
 
850
        entry = self._get_entry(path=path)
 
851
        if entry == (None, None):
 
852
            return None
 
853
        return entry[0][2]
 
854
 
 
855
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
856
        """See Tree.paths2ids().
 
857
 
 
858
        This specialisation fast-paths the case where all the trees are in the
 
859
        dirstate.
 
860
        """
 
861
        if paths is None:
 
862
            return None
 
863
        parents = self.get_parent_ids()
 
864
        for tree in trees:
 
865
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
866
                parents):
 
867
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
868
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
869
        # -- make all paths utf8 --
 
870
        paths_utf8 = set()
 
871
        for path in paths:
 
872
            paths_utf8.add(path.encode('utf8'))
 
873
        paths = paths_utf8
 
874
        # -- paths is now a utf8 path set --
 
875
        # -- get the state object and prepare it.
 
876
        state = self.current_dirstate()
 
877
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
878
            and '' not in paths):
 
879
            paths2ids = self._paths2ids_using_bisect
 
880
        else:
 
881
            paths2ids = self._paths2ids_in_memory
 
882
        return paths2ids(paths, search_indexes,
 
883
                         require_versioned=require_versioned)
 
884
 
 
885
    def _paths2ids_in_memory(self, paths, search_indexes,
 
886
                             require_versioned=True):
 
887
        state = self.current_dirstate()
 
888
        state._read_dirblocks_if_needed()
 
889
        def _entries_for_path(path):
 
890
            """Return a list with all the entries that match path for all ids.
 
891
            """
 
892
            dirname, basename = os.path.split(path)
 
893
            key = (dirname, basename, '')
 
894
            block_index, present = state._find_block_index_from_key(key)
 
895
            if not present:
 
896
                # the block which should contain path is absent.
 
897
                return []
 
898
            result = []
 
899
            block = state._dirblocks[block_index][1]
 
900
            entry_index, _ = state._find_entry_index(key, block)
 
901
            # we may need to look at multiple entries at this path: walk while the paths match.
 
902
            while (entry_index < len(block) and
 
903
                block[entry_index][0][0:2] == key[0:2]):
 
904
                result.append(block[entry_index])
 
905
                entry_index += 1
 
906
            return result
 
907
        if require_versioned:
 
908
            # -- check all supplied paths are versioned in a search tree. --
 
909
            all_versioned = True
 
910
            for path in paths:
 
911
                path_entries = _entries_for_path(path)
 
912
                if not path_entries:
 
913
                    # this specified path is not present at all: error
 
914
                    all_versioned = False
 
915
                    break
 
916
                found_versioned = False
 
917
                # for each id at this path
 
918
                for entry in path_entries:
 
919
                    # for each tree.
 
920
                    for index in search_indexes:
 
921
                        if entry[1][index][0] != 'a': # absent
 
922
                            found_versioned = True
 
923
                            # all good: found a versioned cell
 
924
                            break
 
925
                if not found_versioned:
 
926
                    # none of the indexes was not 'absent' at all ids for this
 
927
                    # path.
 
928
                    all_versioned = False
 
929
                    break
 
930
            if not all_versioned:
 
931
                raise errors.PathsNotVersionedError(paths)
 
932
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
933
        search_paths = osutils.minimum_path_selection(paths)
 
934
        # sketch: 
 
935
        # for all search_indexs in each path at or under each element of
 
936
        # search_paths, if the detail is relocated: add the id, and add the
 
937
        # relocated path as one to search if its not searched already. If the
 
938
        # detail is not relocated, add the id.
 
939
        searched_paths = set()
 
940
        found_ids = set()
 
941
        def _process_entry(entry):
 
942
            """Look at search_indexes within entry.
 
943
 
 
944
            If a specific tree's details are relocated, add the relocation
 
945
            target to search_paths if not searched already. If it is absent, do
 
946
            nothing. Otherwise add the id to found_ids.
 
947
            """
 
948
            for index in search_indexes:
 
949
                if entry[1][index][0] == 'r': # relocated
 
950
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
951
                        search_paths.add(entry[1][index][1])
 
952
                elif entry[1][index][0] != 'a': # absent
 
953
                    found_ids.add(entry[0][2])
 
954
        while search_paths:
 
955
            current_root = search_paths.pop()
 
956
            searched_paths.add(current_root)
 
957
            # process the entries for this containing directory: the rest will be
 
958
            # found by their parents recursively.
 
959
            root_entries = _entries_for_path(current_root)
 
960
            if not root_entries:
 
961
                # this specified path is not present at all, skip it.
 
962
                continue
 
963
            for entry in root_entries:
 
964
                _process_entry(entry)
 
965
            initial_key = (current_root, '', '')
 
966
            block_index, _ = state._find_block_index_from_key(initial_key)
 
967
            while (block_index < len(state._dirblocks) and
 
968
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
969
                for entry in state._dirblocks[block_index][1]:
 
970
                    _process_entry(entry)
 
971
                block_index += 1
 
972
        return found_ids
 
973
 
 
974
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
975
                                require_versioned=True):
 
976
        state = self.current_dirstate()
 
977
        found_ids = set()
 
978
 
 
979
        split_paths = sorted(osutils.split(p) for p in paths)
 
980
        found = state._bisect_recursive(split_paths)
 
981
 
 
982
        if require_versioned:
 
983
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
984
            for dir_name in split_paths:
 
985
                if dir_name not in found_dir_names:
 
986
                    raise errors.PathsNotVersionedError(paths)
 
987
 
 
988
        for dir_name_id, trees_info in found.iteritems():
 
989
            for index in search_indexes:
 
990
                if trees_info[index][0] not in ('r', 'a'):
 
991
                    found_ids.add(dir_name_id[2])
 
992
        return found_ids
 
993
 
 
994
    def read_working_inventory(self):
 
995
        """Read the working inventory.
 
996
        
 
997
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
998
        """
 
999
        return self.inventory
 
1000
 
 
1001
    @needs_read_lock
 
1002
    def revision_tree(self, revision_id):
 
1003
        """See Tree.revision_tree.
 
1004
 
 
1005
        WorkingTree4 supplies revision_trees for any basis tree.
 
1006
        """
 
1007
        revision_id = osutils.safe_revision_id(revision_id)
 
1008
        dirstate = self.current_dirstate()
 
1009
        parent_ids = dirstate.get_parent_ids()
 
1010
        if revision_id not in parent_ids:
 
1011
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1012
        if revision_id in dirstate.get_ghosts():
 
1013
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1014
        return DirStateRevisionTree(dirstate, revision_id,
 
1015
            self.branch.repository)
 
1016
 
 
1017
    @needs_tree_write_lock
 
1018
    def set_last_revision(self, new_revision):
 
1019
        """Change the last revision in the working tree."""
 
1020
        new_revision = osutils.safe_revision_id(new_revision)
 
1021
        parents = self.get_parent_ids()
 
1022
        if new_revision in (NULL_REVISION, None):
 
1023
            assert len(parents) < 2, (
 
1024
                "setting the last parent to none with a pending merge is "
 
1025
                "unsupported.")
 
1026
            self.set_parent_ids([])
 
1027
        else:
 
1028
            self.set_parent_ids([new_revision] + parents[1:],
 
1029
                allow_leftmost_as_ghost=True)
 
1030
 
 
1031
    @needs_tree_write_lock
 
1032
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1033
        """Set the parent ids to revision_ids.
 
1034
        
 
1035
        See also set_parent_trees. This api will try to retrieve the tree data
 
1036
        for each element of revision_ids from the trees repository. If you have
 
1037
        tree data already available, it is more efficient to use
 
1038
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1039
        an easier API to use.
 
1040
 
 
1041
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1042
            working tree. Any of these may be ghosts.
 
1043
        """
 
1044
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1045
        trees = []
 
1046
        for revision_id in revision_ids:
 
1047
            try:
 
1048
                revtree = self.branch.repository.revision_tree(revision_id)
 
1049
                # TODO: jam 20070213 KnitVersionedFile raises
 
1050
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1051
                #       given revision_id is not present. Should Repository be
 
1052
                #       catching it and re-raising NoSuchRevision?
 
1053
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1054
                revtree = None
 
1055
            trees.append((revision_id, revtree))
 
1056
        self.set_parent_trees(trees,
 
1057
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1058
 
 
1059
    @needs_tree_write_lock
 
1060
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1061
        """Set the parents of the working tree.
 
1062
 
 
1063
        :param parents_list: A list of (revision_id, tree) tuples.
 
1064
            If tree is None, then that element is treated as an unreachable
 
1065
            parent tree - i.e. a ghost.
 
1066
        """
 
1067
        dirstate = self.current_dirstate()
 
1068
        if len(parents_list) > 0:
 
1069
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1070
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1071
        real_trees = []
 
1072
        ghosts = []
 
1073
        # convert absent trees to the null tree, which we convert back to
 
1074
        # missing on access.
 
1075
        for rev_id, tree in parents_list:
 
1076
            rev_id = osutils.safe_revision_id(rev_id)
 
1077
            _mod_revision.check_not_reserved_id(rev_id)
 
1078
            if tree is not None:
 
1079
                real_trees.append((rev_id, tree))
 
1080
            else:
 
1081
                real_trees.append((rev_id,
 
1082
                    self.branch.repository.revision_tree(None)))
 
1083
                ghosts.append(rev_id)
 
1084
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1085
        self._make_dirty(reset_inventory=False)
 
1086
 
 
1087
    def _set_root_id(self, file_id):
 
1088
        """See WorkingTree.set_root_id."""
 
1089
        state = self.current_dirstate()
 
1090
        state.set_path_id('', file_id)
 
1091
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1092
            self._make_dirty(reset_inventory=True)
 
1093
 
 
1094
    def _sha_from_stat(self, path, stat_result):
 
1095
        """Get a sha digest from the tree's stat cache.
 
1096
 
 
1097
        The default implementation assumes no stat cache is present.
 
1098
 
 
1099
        :param path: The path.
 
1100
        :param stat_result: The stat result being looked up.
 
1101
        """
 
1102
        state = self.current_dirstate()
 
1103
        # XXX: should we make the path be passed in as utf8 ?
 
1104
        entry = state._get_entry(0, path_utf8=cache_utf8.encode(path))
 
1105
        tree_details = entry[1][0]
 
1106
        packed_stat = dirstate.pack_stat(stat_result)
 
1107
        if tree_details[4] == packed_stat:
 
1108
            return tree_details[1]
 
1109
        else:
 
1110
            return None
 
1111
 
 
1112
    @needs_read_lock
 
1113
    def supports_tree_reference(self):
 
1114
        return self._repo_supports_tree_reference
 
1115
 
 
1116
    def unlock(self):
 
1117
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1118
        # do non-implementation specific cleanup
 
1119
        self._cleanup()
 
1120
 
 
1121
        if self._control_files._lock_count == 1:
 
1122
            # eventually we should do signature checking during read locks for
 
1123
            # dirstate updates.
 
1124
            if self._control_files._lock_mode == 'w':
 
1125
                if self._dirty:
 
1126
                    self.flush()
 
1127
            if self._dirstate is not None:
 
1128
                # This is a no-op if there are no modifications.
 
1129
                self._dirstate.save()
 
1130
                self._dirstate.unlock()
 
1131
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1132
            #       point. Instead, it could check if the header has been
 
1133
            #       modified when it is locked, and if not, it can hang on to
 
1134
            #       the data it has in memory.
 
1135
            self._dirstate = None
 
1136
            self._inventory = None
 
1137
        # reverse order of locking.
 
1138
        try:
 
1139
            return self._control_files.unlock()
 
1140
        finally:
 
1141
            self.branch.unlock()
 
1142
 
 
1143
    @needs_tree_write_lock
 
1144
    def unversion(self, file_ids):
 
1145
        """Remove the file ids in file_ids from the current versioned set.
 
1146
 
 
1147
        When a file_id is unversioned, all of its children are automatically
 
1148
        unversioned.
 
1149
 
 
1150
        :param file_ids: The file ids to stop versioning.
 
1151
        :raises: NoSuchId if any fileid is not currently versioned.
 
1152
        """
 
1153
        if not file_ids:
 
1154
            return
 
1155
        state = self.current_dirstate()
 
1156
        state._read_dirblocks_if_needed()
 
1157
        ids_to_unversion = set()
 
1158
        for file_id in file_ids:
 
1159
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1160
        paths_to_unversion = set()
 
1161
        # sketch:
 
1162
        # check if the root is to be unversioned, if so, assert for now.
 
1163
        # walk the state marking unversioned things as absent.
 
1164
        # if there are any un-unversioned ids at the end, raise
 
1165
        for key, details in state._dirblocks[0][1]:
 
1166
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1167
                key[2] in ids_to_unversion):
 
1168
                # I haven't written the code to unversion / yet - it should be
 
1169
                # supported.
 
1170
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1171
        block_index = 0
 
1172
        while block_index < len(state._dirblocks):
 
1173
            # process one directory at a time.
 
1174
            block = state._dirblocks[block_index]
 
1175
            # first check: is the path one to remove - it or its children
 
1176
            delete_block = False
 
1177
            for path in paths_to_unversion:
 
1178
                if (block[0].startswith(path) and
 
1179
                    (len(block[0]) == len(path) or
 
1180
                     block[0][len(path)] == '/')):
 
1181
                    # this entire block should be deleted - its the block for a
 
1182
                    # path to unversion; or the child of one
 
1183
                    delete_block = True
 
1184
                    break
 
1185
            # TODO: trim paths_to_unversion as we pass by paths
 
1186
            if delete_block:
 
1187
                # this block is to be deleted: process it.
 
1188
                # TODO: we can special case the no-parents case and
 
1189
                # just forget the whole block.
 
1190
                entry_index = 0
 
1191
                while entry_index < len(block[1]):
 
1192
                    # Mark this file id as having been removed
 
1193
                    entry = block[1][entry_index]
 
1194
                    ids_to_unversion.discard(entry[0][2])
 
1195
                    if (entry[1][0][0] == 'a'
 
1196
                        or not state._make_absent(entry)):
 
1197
                        entry_index += 1
 
1198
                # go to the next block. (At the moment we dont delete empty
 
1199
                # dirblocks)
 
1200
                block_index += 1
 
1201
                continue
 
1202
            entry_index = 0
 
1203
            while entry_index < len(block[1]):
 
1204
                entry = block[1][entry_index]
 
1205
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1206
                    # ^ some parent row.
 
1207
                    entry[0][2] not in ids_to_unversion):
 
1208
                    # ^ not an id to unversion
 
1209
                    entry_index += 1
 
1210
                    continue
 
1211
                if entry[1][0][0] == 'd':
 
1212
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1213
                if not state._make_absent(entry):
 
1214
                    entry_index += 1
 
1215
                # we have unversioned this id
 
1216
                ids_to_unversion.remove(entry[0][2])
 
1217
            block_index += 1
 
1218
        if ids_to_unversion:
 
1219
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1220
        self._make_dirty(reset_inventory=False)
 
1221
        # have to change the legacy inventory too.
 
1222
        if self._inventory is not None:
 
1223
            for file_id in file_ids:
 
1224
                self._inventory.remove_recursive_id(file_id)
 
1225
 
 
1226
    @needs_read_lock
 
1227
    def _validate(self):
 
1228
        self._dirstate._validate()
 
1229
 
 
1230
    @needs_tree_write_lock
 
1231
    def _write_inventory(self, inv):
 
1232
        """Write inventory as the current inventory."""
 
1233
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1234
        self.current_dirstate().set_state_from_inventory(inv)
 
1235
        self._make_dirty(reset_inventory=False)
 
1236
        if self._inventory is not None:
 
1237
            self._inventory = inv
 
1238
        self.flush()
 
1239
 
 
1240
 
 
1241
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1242
    """The first consolidated dirstate working tree format.
 
1243
 
 
1244
    This format:
 
1245
        - exists within a metadir controlling .bzr
 
1246
        - includes an explicit version marker for the workingtree control
 
1247
          files, separate from the BzrDir format
 
1248
        - modifies the hash cache format
 
1249
        - is new in bzr 0.15
 
1250
        - uses a LockDir to guard access to it.
 
1251
    """
 
1252
 
 
1253
    upgrade_recommended = False
 
1254
 
 
1255
    def get_format_string(self):
 
1256
        """See WorkingTreeFormat.get_format_string()."""
 
1257
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1258
 
 
1259
    def get_format_description(self):
 
1260
        """See WorkingTreeFormat.get_format_description()."""
 
1261
        return "Working tree format 4"
 
1262
 
 
1263
    def initialize(self, a_bzrdir, revision_id=None):
 
1264
        """See WorkingTreeFormat.initialize().
 
1265
 
 
1266
        :param revision_id: allows creating a working tree at a different
 
1267
        revision than the branch is at.
 
1268
 
 
1269
        These trees get an initial random root id, if their repository supports
 
1270
        rich root data, TREE_ROOT otherwise.
 
1271
        """
 
1272
        revision_id = osutils.safe_revision_id(revision_id)
 
1273
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1274
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1275
        transport = a_bzrdir.get_workingtree_transport(self)
 
1276
        control_files = self._open_control_files(a_bzrdir)
 
1277
        control_files.create_lock()
 
1278
        control_files.lock_write()
 
1279
        control_files.put_utf8('format', self.get_format_string())
 
1280
        branch = a_bzrdir.open_branch()
 
1281
        if revision_id is None:
 
1282
            revision_id = branch.last_revision()
 
1283
        local_path = transport.local_abspath('dirstate')
 
1284
        # write out new dirstate (must exist when we create the tree)
 
1285
        state = dirstate.DirState.initialize(local_path)
 
1286
        state.unlock()
 
1287
        del state
 
1288
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1289
                         branch,
 
1290
                         _format=self,
 
1291
                         _bzrdir=a_bzrdir,
 
1292
                         _control_files=control_files)
 
1293
        wt._new_tree()
 
1294
        wt.lock_tree_write()
 
1295
        try:
 
1296
            if revision_id in (None, NULL_REVISION):
 
1297
                if branch.repository.supports_rich_root():
 
1298
                    wt._set_root_id(generate_ids.gen_root_id())
 
1299
                else:
 
1300
                    wt._set_root_id(ROOT_ID)
 
1301
                wt.flush()
 
1302
            wt.set_last_revision(revision_id)
 
1303
            wt.flush()
 
1304
            basis = wt.basis_tree()
 
1305
            basis.lock_read()
 
1306
            # if the basis has a root id we have to use that; otherwise we use
 
1307
            # a new random one
 
1308
            basis_root_id = basis.get_root_id()
 
1309
            if basis_root_id is not None:
 
1310
                wt._set_root_id(basis_root_id)
 
1311
                wt.flush()
 
1312
            transform.build_tree(basis, wt)
 
1313
            basis.unlock()
 
1314
        finally:
 
1315
            control_files.unlock()
 
1316
            wt.unlock()
 
1317
        return wt
 
1318
 
 
1319
    def _open(self, a_bzrdir, control_files):
 
1320
        """Open the tree itself.
 
1321
 
 
1322
        :param a_bzrdir: the dir for the tree.
 
1323
        :param control_files: the control files for the tree.
 
1324
        """
 
1325
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1326
                           branch=a_bzrdir.open_branch(),
 
1327
                           _format=self,
 
1328
                           _bzrdir=a_bzrdir,
 
1329
                           _control_files=control_files)
 
1330
 
 
1331
    def __get_matchingbzrdir(self):
 
1332
        # please test against something that will let us do tree references
 
1333
        return bzrdir.format_registry.make_bzrdir(
 
1334
            'dirstate-with-subtree')
 
1335
 
 
1336
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1337
 
 
1338
 
 
1339
class DirStateRevisionTree(Tree):
 
1340
    """A revision tree pulling the inventory from a dirstate."""
 
1341
 
 
1342
    def __init__(self, dirstate, revision_id, repository):
 
1343
        self._dirstate = dirstate
 
1344
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1345
        self._repository = repository
 
1346
        self._inventory = None
 
1347
        self._locked = 0
 
1348
        self._dirstate_locked = False
 
1349
 
 
1350
    def __repr__(self):
 
1351
        return "<%s of %s in %s>" % \
 
1352
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1353
 
 
1354
    def annotate_iter(self, file_id,
 
1355
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1356
        """See Tree.annotate_iter"""
 
1357
        w = self._get_weave(file_id)
 
1358
        return w.annotate_iter(self.inventory[file_id].revision)
 
1359
 
 
1360
    def _get_ancestors(self, default_revision):
 
1361
        return set(self._repository.get_ancestry(self._revision_id,
 
1362
                                                 topo_sorted=False))
 
1363
    def _comparison_data(self, entry, path):
 
1364
        """See Tree._comparison_data."""
 
1365
        if entry is None:
 
1366
            return None, False, None
 
1367
        # trust the entry as RevisionTree does, but this may not be
 
1368
        # sensible: the entry might not have come from us?
 
1369
        return entry.kind, entry.executable, None
 
1370
 
 
1371
    def _file_size(self, entry, stat_value):
 
1372
        return entry.text_size
 
1373
 
 
1374
    def filter_unversioned_files(self, paths):
 
1375
        """Filter out paths that are not versioned.
 
1376
 
 
1377
        :return: set of paths.
 
1378
        """
 
1379
        pred = self.has_filename
 
1380
        return set((p for p in paths if not pred(p)))
 
1381
 
 
1382
    def get_root_id(self):
 
1383
        return self.path2id('')
 
1384
 
 
1385
    def _get_parent_index(self):
 
1386
        """Return the index in the dirstate referenced by this tree."""
 
1387
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1388
 
 
1389
    def _get_entry(self, file_id=None, path=None):
 
1390
        """Get the dirstate row for file_id or path.
 
1391
 
 
1392
        If either file_id or path is supplied, it is used as the key to lookup.
 
1393
        If both are supplied, the fastest lookup is used, and an error is
 
1394
        raised if they do not both point at the same row.
 
1395
        
 
1396
        :param file_id: An optional unicode file_id to be looked up.
 
1397
        :param path: An optional unicode path to be looked up.
 
1398
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1399
        """
 
1400
        if file_id is None and path is None:
 
1401
            raise errors.BzrError('must supply file_id or path')
 
1402
        file_id = osutils.safe_file_id(file_id)
 
1403
        if path is not None:
 
1404
            path = path.encode('utf8')
 
1405
        parent_index = self._get_parent_index()
 
1406
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1407
 
 
1408
    def _generate_inventory(self):
 
1409
        """Create and set self.inventory from the dirstate object.
 
1410
 
 
1411
        (So this is only called the first time the inventory is requested for
 
1412
        this tree; it then remains in memory until it's out of date.)
 
1413
 
 
1414
        This is relatively expensive: we have to walk the entire dirstate.
 
1415
        """
 
1416
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1417
            'dirstate revision tree'
 
1418
        # separate call for profiling - makes it clear where the costs are.
 
1419
        self._dirstate._read_dirblocks_if_needed()
 
1420
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1421
            'parent %s has disappeared from %s' % (
 
1422
            self._revision_id, self._dirstate.get_parent_ids())
 
1423
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1424
        # This is identical now to the WorkingTree _generate_inventory except
 
1425
        # for the tree index use.
 
1426
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1427
        current_id = root_key[2]
 
1428
        assert current_entry[parent_index][0] == 'd'
 
1429
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1430
        inv.root.revision = current_entry[parent_index][4]
 
1431
        # Turn some things into local variables
 
1432
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1433
        factory = entry_factory
 
1434
        utf8_decode = cache_utf8._utf8_decode
 
1435
        inv_byid = inv._byid
 
1436
        # we could do this straight out of the dirstate; it might be fast
 
1437
        # and should be profiled - RBC 20070216
 
1438
        parent_ies = {'' : inv.root}
 
1439
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1440
            dirname = block[0]
 
1441
            try:
 
1442
                parent_ie = parent_ies[dirname]
 
1443
            except KeyError:
 
1444
                # all the paths in this block are not versioned in this tree
 
1445
                continue
 
1446
            for key, entry in block[1]:
 
1447
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1448
                if minikind in ('a', 'r'): # absent, relocated
 
1449
                    # not this tree
 
1450
                    continue
 
1451
                name = key[1]
 
1452
                name_unicode = utf8_decode(name)[0]
 
1453
                file_id = key[2]
 
1454
                kind = minikind_to_kind[minikind]
 
1455
                inv_entry = factory[kind](file_id, name_unicode,
 
1456
                                          parent_ie.file_id)
 
1457
                inv_entry.revision = revid
 
1458
                if kind == 'file':
 
1459
                    inv_entry.executable = executable
 
1460
                    inv_entry.text_size = size
 
1461
                    inv_entry.text_sha1 = fingerprint
 
1462
                elif kind == 'directory':
 
1463
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1464
                elif kind == 'symlink':
 
1465
                    inv_entry.executable = False
 
1466
                    inv_entry.text_size = None
 
1467
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1468
                elif kind == 'tree-reference':
 
1469
                    inv_entry.reference_revision = fingerprint or None
 
1470
                else:
 
1471
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1472
                            % entry)
 
1473
                # These checks cost us around 40ms on a 55k entry tree
 
1474
                assert file_id not in inv_byid
 
1475
                assert name_unicode not in parent_ie.children
 
1476
                inv_byid[file_id] = inv_entry
 
1477
                parent_ie.children[name_unicode] = inv_entry
 
1478
        self._inventory = inv
 
1479
 
 
1480
    def get_file_mtime(self, file_id, path=None):
 
1481
        """Return the modification time for this record.
 
1482
 
 
1483
        We return the timestamp of the last-changed revision.
 
1484
        """
 
1485
        # Make sure the file exists
 
1486
        entry = self._get_entry(file_id, path=path)
 
1487
        if entry == (None, None): # do we raise?
 
1488
            return None
 
1489
        parent_index = self._get_parent_index()
 
1490
        last_changed_revision = entry[1][parent_index][4]
 
1491
        return self._repository.get_revision(last_changed_revision).timestamp
 
1492
 
 
1493
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1494
        entry = self._get_entry(file_id=file_id, path=path)
 
1495
        parent_index = self._get_parent_index()
 
1496
        parent_details = entry[1][parent_index]
 
1497
        if parent_details[0] == 'f':
 
1498
            return parent_details[1]
 
1499
        return None
 
1500
 
 
1501
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1502
    def get_weave(self, file_id):
 
1503
        return self._get_weave(file_id)
 
1504
 
 
1505
    def _get_weave(self, file_id):
 
1506
        return self._repository.weave_store.get_weave(file_id,
 
1507
                self._repository.get_transaction())
 
1508
 
 
1509
    def get_file(self, file_id, path=None):
 
1510
        return StringIO(self.get_file_text(file_id))
 
1511
 
 
1512
    def get_file_lines(self, file_id):
 
1513
        ie = self.inventory[file_id]
 
1514
        return self._get_weave(file_id).get_lines(ie.revision)
 
1515
 
 
1516
    def get_file_size(self, file_id):
 
1517
        return self.inventory[file_id].text_size
 
1518
 
 
1519
    def get_file_text(self, file_id):
 
1520
        return ''.join(self.get_file_lines(file_id))
 
1521
 
 
1522
    def get_reference_revision(self, file_id, path=None):
 
1523
        return self.inventory[file_id].reference_revision
 
1524
 
 
1525
    def iter_files_bytes(self, desired_files):
 
1526
        """See Tree.iter_files_bytes.
 
1527
 
 
1528
        This version is implemented on top of Repository.iter_files_bytes"""
 
1529
        parent_index = self._get_parent_index()
 
1530
        repo_desired_files = []
 
1531
        for file_id, identifier in desired_files:
 
1532
            entry = self._get_entry(file_id)
 
1533
            if entry == (None, None):
 
1534
                raise errors.NoSuchId(self, file_id)
 
1535
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1536
                                       identifier))
 
1537
        return self._repository.iter_files_bytes(repo_desired_files)
 
1538
 
 
1539
    def get_symlink_target(self, file_id):
 
1540
        entry = self._get_entry(file_id=file_id)
 
1541
        parent_index = self._get_parent_index()
 
1542
        if entry[1][parent_index][0] != 'l':
 
1543
            return None
 
1544
        else:
 
1545
            # At present, none of the tree implementations supports non-ascii
 
1546
            # symlink targets. So we will just assume that the dirstate path is
 
1547
            # correct.
 
1548
            return entry[1][parent_index][1]
 
1549
 
 
1550
    def get_revision_id(self):
 
1551
        """Return the revision id for this tree."""
 
1552
        return self._revision_id
 
1553
 
 
1554
    def _get_inventory(self):
 
1555
        if self._inventory is not None:
 
1556
            return self._inventory
 
1557
        self._must_be_locked()
 
1558
        self._generate_inventory()
 
1559
        return self._inventory
 
1560
 
 
1561
    inventory = property(_get_inventory,
 
1562
                         doc="Inventory of this Tree")
 
1563
 
 
1564
    def get_parent_ids(self):
 
1565
        """The parents of a tree in the dirstate are not cached."""
 
1566
        return self._repository.get_revision(self._revision_id).parent_ids
 
1567
 
 
1568
    def has_filename(self, filename):
 
1569
        return bool(self.path2id(filename))
 
1570
 
 
1571
    def kind(self, file_id):
 
1572
        return self.inventory[file_id].kind
 
1573
 
 
1574
    def path_content_summary(self, path):
 
1575
        """See Tree.path_content_summary."""
 
1576
        id = self.inventory.path2id(path)
 
1577
        if id is None:
 
1578
            return ('missing', None, None, None)
 
1579
        entry = self._inventory[id]
 
1580
        kind = entry.kind
 
1581
        if kind == 'file':
 
1582
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1583
        elif kind == 'symlink':
 
1584
            return (kind, None, None, entry.symlink_target)
 
1585
        else:
 
1586
            return (kind, None, None, None)
 
1587
 
 
1588
    def is_executable(self, file_id, path=None):
 
1589
        ie = self.inventory[file_id]
 
1590
        if ie.kind != "file":
 
1591
            return None
 
1592
        return ie.executable
 
1593
 
 
1594
    def list_files(self, include_root=False):
 
1595
        # We use a standard implementation, because DirStateRevisionTree is
 
1596
        # dealing with one of the parents of the current state
 
1597
        inv = self._get_inventory()
 
1598
        entries = inv.iter_entries()
 
1599
        if self.inventory.root is not None and not include_root:
 
1600
            entries.next()
 
1601
        for path, entry in entries:
 
1602
            yield path, 'V', entry.kind, entry.file_id, entry
 
1603
 
 
1604
    def lock_read(self):
 
1605
        """Lock the tree for a set of operations."""
 
1606
        if not self._locked:
 
1607
            self._repository.lock_read()
 
1608
            if self._dirstate._lock_token is None:
 
1609
                self._dirstate.lock_read()
 
1610
                self._dirstate_locked = True
 
1611
        self._locked += 1
 
1612
 
 
1613
    def _must_be_locked(self):
 
1614
        if not self._locked:
 
1615
            raise errors.ObjectNotLocked(self)
 
1616
 
 
1617
    @needs_read_lock
 
1618
    def path2id(self, path):
 
1619
        """Return the id for path in this tree."""
 
1620
        # lookup by path: faster than splitting and walking the ivnentory.
 
1621
        entry = self._get_entry(path=path)
 
1622
        if entry == (None, None):
 
1623
            return None
 
1624
        return entry[0][2]
 
1625
 
 
1626
    def unlock(self):
 
1627
        """Unlock, freeing any cache memory used during the lock."""
 
1628
        # outside of a lock, the inventory is suspect: release it.
 
1629
        self._locked -=1
 
1630
        if not self._locked:
 
1631
            self._inventory = None
 
1632
            self._locked = 0
 
1633
            if self._dirstate_locked:
 
1634
                self._dirstate.unlock()
 
1635
                self._dirstate_locked = False
 
1636
            self._repository.unlock()
 
1637
 
 
1638
    def walkdirs(self, prefix=""):
 
1639
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1640
        # implementation based on an inventory.  
 
1641
        # This should be cleaned up to use the much faster Dirstate code
 
1642
        # So for now, we just build up the parent inventory, and extract
 
1643
        # it the same way RevisionTree does.
 
1644
        _directory = 'directory'
 
1645
        inv = self._get_inventory()
 
1646
        top_id = inv.path2id(prefix)
 
1647
        if top_id is None:
 
1648
            pending = []
 
1649
        else:
 
1650
            pending = [(prefix, top_id)]
 
1651
        while pending:
 
1652
            dirblock = []
 
1653
            relpath, file_id = pending.pop()
 
1654
            # 0 - relpath, 1- file-id
 
1655
            if relpath:
 
1656
                relroot = relpath + '/'
 
1657
            else:
 
1658
                relroot = ""
 
1659
            # FIXME: stash the node in pending
 
1660
            entry = inv[file_id]
 
1661
            for name, child in entry.sorted_children():
 
1662
                toppath = relroot + name
 
1663
                dirblock.append((toppath, name, child.kind, None,
 
1664
                    child.file_id, child.kind
 
1665
                    ))
 
1666
            yield (relpath, entry.file_id), dirblock
 
1667
            # push the user specified dirs from dirblock
 
1668
            for dir in reversed(dirblock):
 
1669
                if dir[2] == _directory:
 
1670
                    pending.append((dir[0], dir[4]))
 
1671
 
 
1672
 
 
1673
class InterDirStateTree(InterTree):
 
1674
    """Fast path optimiser for changes_from with dirstate trees.
 
1675
    
 
1676
    This is used only when both trees are in the dirstate working file, and 
 
1677
    the source is any parent within the dirstate, and the destination is 
 
1678
    the current working tree of the same dirstate.
 
1679
    """
 
1680
    # this could be generalized to allow comparisons between any trees in the
 
1681
    # dirstate, and possibly between trees stored in different dirstates.
 
1682
 
 
1683
    def __init__(self, source, target):
 
1684
        super(InterDirStateTree, self).__init__(source, target)
 
1685
        if not InterDirStateTree.is_compatible(source, target):
 
1686
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1687
 
 
1688
    @staticmethod
 
1689
    def make_source_parent_tree(source, target):
 
1690
        """Change the source tree into a parent of the target."""
 
1691
        revid = source.commit('record tree')
 
1692
        target.branch.repository.fetch(source.branch.repository, revid)
 
1693
        target.set_parent_ids([revid])
 
1694
        return target.basis_tree(), target
 
1695
 
 
1696
    _matching_from_tree_format = WorkingTreeFormat4()
 
1697
    _matching_to_tree_format = WorkingTreeFormat4()
 
1698
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1699
 
 
1700
    def _iter_changes(self, include_unchanged=False,
 
1701
                      specific_files=None, pb=None, extra_trees=[],
 
1702
                      require_versioned=True, want_unversioned=False):
 
1703
        """Return the changes from source to target.
 
1704
 
 
1705
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1706
            for details.
 
1707
        :param specific_files: An optional list of file paths to restrict the
 
1708
            comparison to. When mapping filenames to ids, all matches in all
 
1709
            trees (including optional extra_trees) are used, and all children of
 
1710
            matched directories are included.
 
1711
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1712
            unchanged entries in the result.
 
1713
        :param extra_trees: An optional list of additional trees to use when
 
1714
            mapping the contents of specific_files (paths) to file_ids.
 
1715
        :param require_versioned: If True, all files in specific_files must be
 
1716
            versioned in one of source, target, extra_trees or
 
1717
            PathsNotVersionedError is raised.
 
1718
        :param want_unversioned: Should unversioned files be returned in the
 
1719
            output. An unversioned file is defined as one with (False, False)
 
1720
            for the versioned pair.
 
1721
        """
 
1722
        utf8_decode = cache_utf8._utf8_decode
 
1723
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1724
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1725
        # NB: show_status depends on being able to pass in non-versioned files
 
1726
        # and report them as unknown
 
1727
        # TODO: handle extra trees in the dirstate.
 
1728
        # TODO: handle comparisons as an empty tree as a different special
 
1729
        # case? mbp 20070226
 
1730
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1731
            or specific_files == []):
 
1732
            # we can't fast-path these cases (yet)
 
1733
            for f in super(InterDirStateTree, self)._iter_changes(
 
1734
                include_unchanged, specific_files, pb, extra_trees,
 
1735
                require_versioned, want_unversioned=want_unversioned):
 
1736
                yield f
 
1737
            return
 
1738
        parent_ids = self.target.get_parent_ids()
 
1739
        assert (self.source._revision_id in parent_ids), \
 
1740
                "revision {%s} is not stored in {%s}, but %s " \
 
1741
                "can only be used for trees stored in the dirstate" \
 
1742
                % (self.source._revision_id, self.target, self._iter_changes)
 
1743
        target_index = 0
 
1744
        if self.source._revision_id == NULL_REVISION:
 
1745
            source_index = None
 
1746
            indices = (target_index,)
 
1747
        else:
 
1748
            assert (self.source._revision_id in parent_ids), \
 
1749
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1750
                self.source._revision_id, parent_ids)
 
1751
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1752
            indices = (source_index,target_index)
 
1753
        # -- make all specific_files utf8 --
 
1754
        if specific_files:
 
1755
            specific_files_utf8 = set()
 
1756
            for path in specific_files:
 
1757
                specific_files_utf8.add(path.encode('utf8'))
 
1758
            specific_files = specific_files_utf8
 
1759
        else:
 
1760
            specific_files = set([''])
 
1761
        # -- specific_files is now a utf8 path set --
 
1762
        # -- get the state object and prepare it.
 
1763
        state = self.target.current_dirstate()
 
1764
        state._read_dirblocks_if_needed()
 
1765
        def _entries_for_path(path):
 
1766
            """Return a list with all the entries that match path for all ids.
 
1767
            """
 
1768
            dirname, basename = os.path.split(path)
 
1769
            key = (dirname, basename, '')
 
1770
            block_index, present = state._find_block_index_from_key(key)
 
1771
            if not present:
 
1772
                # the block which should contain path is absent.
 
1773
                return []
 
1774
            result = []
 
1775
            block = state._dirblocks[block_index][1]
 
1776
            entry_index, _ = state._find_entry_index(key, block)
 
1777
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1778
            while (entry_index < len(block) and
 
1779
                block[entry_index][0][0:2] == key[0:2]):
 
1780
                result.append(block[entry_index])
 
1781
                entry_index += 1
 
1782
            return result
 
1783
        if require_versioned:
 
1784
            # -- check all supplied paths are versioned in a search tree. --
 
1785
            all_versioned = True
 
1786
            for path in specific_files:
 
1787
                path_entries = _entries_for_path(path)
 
1788
                if not path_entries:
 
1789
                    # this specified path is not present at all: error
 
1790
                    all_versioned = False
 
1791
                    break
 
1792
                found_versioned = False
 
1793
                # for each id at this path
 
1794
                for entry in path_entries:
 
1795
                    # for each tree.
 
1796
                    for index in indices:
 
1797
                        if entry[1][index][0] != 'a': # absent
 
1798
                            found_versioned = True
 
1799
                            # all good: found a versioned cell
 
1800
                            break
 
1801
                if not found_versioned:
 
1802
                    # none of the indexes was not 'absent' at all ids for this
 
1803
                    # path.
 
1804
                    all_versioned = False
 
1805
                    break
 
1806
            if not all_versioned:
 
1807
                raise errors.PathsNotVersionedError(specific_files)
 
1808
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1809
        search_specific_files = set()
 
1810
        for path in specific_files:
 
1811
            other_specific_files = specific_files.difference(set([path]))
 
1812
            if not osutils.is_inside_any(other_specific_files, path):
 
1813
                # this is a top level path, we must check it.
 
1814
                search_specific_files.add(path)
 
1815
        # sketch: 
 
1816
        # compare source_index and target_index at or under each element of search_specific_files.
 
1817
        # follow the following comparison table. Note that we only want to do diff operations when
 
1818
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1819
        # for the target.
 
1820
        # cases:
 
1821
        # 
 
1822
        # Source | Target | disk | action
 
1823
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1824
        #        |        |      | diff check on source-target
 
1825
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1826
        #        |        |      | ???
 
1827
        #   r    |  a     |      | add source to search
 
1828
        #   r    |  a     |  a   | 
 
1829
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1830
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1831
        #   a    | fdlt   |      | add new id
 
1832
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1833
        #   a    |  a     |      | not present in either tree, skip
 
1834
        #   a    |  a     |  a   | not present in any tree, skip
 
1835
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1836
        #        |        |      | may not be selected by the users list of paths.
 
1837
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1838
        #        |        |      | may not be selected by the users list of paths.
 
1839
        #  fdlt  | fdlt   |      | content in both: diff them
 
1840
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1841
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1842
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1843
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1844
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1845
        #        |        |      | this id at the other path.
 
1846
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1847
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1848
        #        |        |      | this id at the other path.
 
1849
 
 
1850
        # for all search_indexs in each path at or under each element of
 
1851
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1852
        # relocated path as one to search if its not searched already. If the
 
1853
        # detail is not relocated, add the id.
 
1854
        searched_specific_files = set()
 
1855
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1856
        # Using a list so that we can access the values and change them in
 
1857
        # nested scope. Each one is [path, file_id, entry]
 
1858
        last_source_parent = [None, None]
 
1859
        last_target_parent = [None, None]
 
1860
 
 
1861
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1862
 
 
1863
        # Just a sentry, so that _process_entry can say that this
 
1864
        # record is handled, but isn't interesting to process (unchanged)
 
1865
        uninteresting = object()
 
1866
 
 
1867
 
 
1868
        old_dirname_to_file_id = {}
 
1869
        new_dirname_to_file_id = {}
 
1870
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1871
        #       keeping a cache of directories that we have seen.
 
1872
 
 
1873
        def _process_entry(entry, path_info):
 
1874
            """Compare an entry and real disk to generate delta information.
 
1875
 
 
1876
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1877
                the path of entry. If None, then the path is considered absent.
 
1878
                (Perhaps we should pass in a concrete entry for this ?)
 
1879
                Basename is returned as a utf8 string because we expect this
 
1880
                tuple will be ignored, and don't want to take the time to
 
1881
                decode.
 
1882
            :return: None if these don't match
 
1883
                     A tuple of information about the change, or
 
1884
                     the object 'uninteresting' if these match, but are
 
1885
                     basically identical.
 
1886
            """
 
1887
            if source_index is None:
 
1888
                source_details = NULL_PARENT_DETAILS
 
1889
            else:
 
1890
                source_details = entry[1][source_index]
 
1891
            target_details = entry[1][target_index]
 
1892
            target_minikind = target_details[0]
 
1893
            if path_info is not None and target_minikind in 'fdlt':
 
1894
                assert target_index == 0
 
1895
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1896
                                                  stat_value=path_info[3])
 
1897
                # The entry may have been modified by update_entry
 
1898
                target_details = entry[1][target_index]
 
1899
                target_minikind = target_details[0]
 
1900
            else:
 
1901
                link_or_sha1 = None
 
1902
            file_id = entry[0][2]
 
1903
            source_minikind = source_details[0]
 
1904
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1905
                # claimed content in both: diff
 
1906
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1907
                #        |        |      | diff check on source-target
 
1908
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1909
                #        |        |      | ???
 
1910
                if source_minikind in 'r':
 
1911
                    # add the source to the search path to find any children it
 
1912
                    # has.  TODO ? : only add if it is a container ?
 
1913
                    if not osutils.is_inside_any(searched_specific_files,
 
1914
                                                 source_details[1]):
 
1915
                        search_specific_files.add(source_details[1])
 
1916
                    # generate the old path; this is needed for stating later
 
1917
                    # as well.
 
1918
                    old_path = source_details[1]
 
1919
                    old_dirname, old_basename = os.path.split(old_path)
 
1920
                    path = pathjoin(entry[0][0], entry[0][1])
 
1921
                    old_entry = state._get_entry(source_index,
 
1922
                                                 path_utf8=old_path)
 
1923
                    # update the source details variable to be the real
 
1924
                    # location.
 
1925
                    source_details = old_entry[1][source_index]
 
1926
                    source_minikind = source_details[0]
 
1927
                else:
 
1928
                    old_dirname = entry[0][0]
 
1929
                    old_basename = entry[0][1]
 
1930
                    old_path = path = None
 
1931
                if path_info is None:
 
1932
                    # the file is missing on disk, show as removed.
 
1933
                    content_change = True
 
1934
                    target_kind = None
 
1935
                    target_exec = False
 
1936
                else:
 
1937
                    # source and target are both versioned and disk file is present.
 
1938
                    target_kind = path_info[2]
 
1939
                    if target_kind == 'directory':
 
1940
                        if path is None:
 
1941
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1942
                        new_dirname_to_file_id[path] = file_id
 
1943
                        if source_minikind != 'd':
 
1944
                            content_change = True
 
1945
                        else:
 
1946
                            # directories have no fingerprint
 
1947
                            content_change = False
 
1948
                        target_exec = False
 
1949
                    elif target_kind == 'file':
 
1950
                        if source_minikind != 'f':
 
1951
                            content_change = True
 
1952
                        else:
 
1953
                            # We could check the size, but we already have the
 
1954
                            # sha1 hash.
 
1955
                            content_change = (link_or_sha1 != source_details[1])
 
1956
                        # Target details is updated at update_entry time
 
1957
                        if use_filesystem_for_exec:
 
1958
                            # We don't need S_ISREG here, because we are sure
 
1959
                            # we are dealing with a file.
 
1960
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1961
                        else:
 
1962
                            target_exec = target_details[3]
 
1963
                    elif target_kind == 'symlink':
 
1964
                        if source_minikind != 'l':
 
1965
                            content_change = True
 
1966
                        else:
 
1967
                            content_change = (link_or_sha1 != source_details[1])
 
1968
                        target_exec = False
 
1969
                    elif target_kind == 'tree-reference':
 
1970
                        if source_minikind != 't':
 
1971
                            content_change = True
 
1972
                        else:
 
1973
                            content_change = False
 
1974
                        target_exec = False
 
1975
                    else:
 
1976
                        raise Exception, "unknown kind %s" % path_info[2]
 
1977
                if source_minikind == 'd':
 
1978
                    if path is None:
 
1979
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1980
                    old_dirname_to_file_id[old_path] = file_id
 
1981
                # parent id is the entry for the path in the target tree
 
1982
                if old_dirname == last_source_parent[0]:
 
1983
                    source_parent_id = last_source_parent[1]
 
1984
                else:
 
1985
                    try:
 
1986
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1987
                    except KeyError:
 
1988
                        source_parent_entry = state._get_entry(source_index,
 
1989
                                                               path_utf8=old_dirname)
 
1990
                        source_parent_id = source_parent_entry[0][2]
 
1991
                    if source_parent_id == entry[0][2]:
 
1992
                        # This is the root, so the parent is None
 
1993
                        source_parent_id = None
 
1994
                    else:
 
1995
                        last_source_parent[0] = old_dirname
 
1996
                        last_source_parent[1] = source_parent_id
 
1997
                new_dirname = entry[0][0]
 
1998
                if new_dirname == last_target_parent[0]:
 
1999
                    target_parent_id = last_target_parent[1]
 
2000
                else:
 
2001
                    try:
 
2002
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2003
                    except KeyError:
 
2004
                        # TODO: We don't always need to do the lookup, because the
 
2005
                        #       parent entry will be the same as the source entry.
 
2006
                        target_parent_entry = state._get_entry(target_index,
 
2007
                                                               path_utf8=new_dirname)
 
2008
                        assert target_parent_entry != (None, None), (
 
2009
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2010
                            % (new_dirname, entry))
 
2011
                        target_parent_id = target_parent_entry[0][2]
 
2012
                    if target_parent_id == entry[0][2]:
 
2013
                        # This is the root, so the parent is None
 
2014
                        target_parent_id = None
 
2015
                    else:
 
2016
                        last_target_parent[0] = new_dirname
 
2017
                        last_target_parent[1] = target_parent_id
 
2018
 
 
2019
                source_exec = source_details[3]
 
2020
                if (include_unchanged
 
2021
                    or content_change
 
2022
                    or source_parent_id != target_parent_id
 
2023
                    or old_basename != entry[0][1]
 
2024
                    or source_exec != target_exec
 
2025
                    ):
 
2026
                    if old_path is None:
 
2027
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2028
                        old_path_u = utf8_decode(old_path)[0]
 
2029
                        path_u = old_path_u
 
2030
                    else:
 
2031
                        old_path_u = utf8_decode(old_path)[0]
 
2032
                        if old_path == path:
 
2033
                            path_u = old_path_u
 
2034
                        else:
 
2035
                            path_u = utf8_decode(path)[0]
 
2036
                    source_kind = _minikind_to_kind[source_minikind]
 
2037
                    return (entry[0][2],
 
2038
                           (old_path_u, path_u),
 
2039
                           content_change,
 
2040
                           (True, True),
 
2041
                           (source_parent_id, target_parent_id),
 
2042
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2043
                           (source_kind, target_kind),
 
2044
                           (source_exec, target_exec))
 
2045
                else:
 
2046
                    return uninteresting
 
2047
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2048
                # looks like a new file
 
2049
                if path_info is not None:
 
2050
                    path = pathjoin(entry[0][0], entry[0][1])
 
2051
                    # parent id is the entry for the path in the target tree
 
2052
                    # TODO: these are the same for an entire directory: cache em.
 
2053
                    parent_id = state._get_entry(target_index,
 
2054
                                                 path_utf8=entry[0][0])[0][2]
 
2055
                    if parent_id == entry[0][2]:
 
2056
                        parent_id = None
 
2057
                    if use_filesystem_for_exec:
 
2058
                        # We need S_ISREG here, because we aren't sure if this
 
2059
                        # is a file or not.
 
2060
                        target_exec = bool(
 
2061
                            stat.S_ISREG(path_info[3].st_mode)
 
2062
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2063
                    else:
 
2064
                        target_exec = target_details[3]
 
2065
                    return (entry[0][2],
 
2066
                           (None, utf8_decode(path)[0]),
 
2067
                           True,
 
2068
                           (False, True),
 
2069
                           (None, parent_id),
 
2070
                           (None, utf8_decode(entry[0][1])[0]),
 
2071
                           (None, path_info[2]),
 
2072
                           (None, target_exec))
 
2073
                else:
 
2074
                    # but its not on disk: we deliberately treat this as just
 
2075
                    # never-present. (Why ?! - RBC 20070224)
 
2076
                    pass
 
2077
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2078
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2079
                # if its still on disk, *and* theres no other entry at this
 
2080
                # path [we dont know this in this routine at the moment -
 
2081
                # perhaps we should change this - then it would be an unknown.
 
2082
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2083
                # parent id is the entry for the path in the target tree
 
2084
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2085
                if parent_id == entry[0][2]:
 
2086
                    parent_id = None
 
2087
                return (entry[0][2],
 
2088
                       (utf8_decode(old_path)[0], None),
 
2089
                       True,
 
2090
                       (True, False),
 
2091
                       (parent_id, None),
 
2092
                       (utf8_decode(entry[0][1])[0], None),
 
2093
                       (_minikind_to_kind[source_minikind], None),
 
2094
                       (source_details[3], None))
 
2095
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2096
                # a rename; could be a true rename, or a rename inherited from
 
2097
                # a renamed parent. TODO: handle this efficiently. Its not
 
2098
                # common case to rename dirs though, so a correct but slow
 
2099
                # implementation will do.
 
2100
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2101
                    search_specific_files.add(target_details[1])
 
2102
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2103
                # neither of the selected trees contain this file,
 
2104
                # so skip over it. This is not currently directly tested, but
 
2105
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2106
                pass
 
2107
            else:
 
2108
                raise AssertionError("don't know how to compare "
 
2109
                    "source_minikind=%r, target_minikind=%r"
 
2110
                    % (source_minikind, target_minikind))
 
2111
                ## import pdb;pdb.set_trace()
 
2112
            return None
 
2113
 
 
2114
        while search_specific_files:
 
2115
            # TODO: the pending list should be lexically sorted?  the
 
2116
            # interface doesn't require it.
 
2117
            current_root = search_specific_files.pop()
 
2118
            current_root_unicode = current_root.decode('utf8')
 
2119
            searched_specific_files.add(current_root)
 
2120
            # process the entries for this containing directory: the rest will be
 
2121
            # found by their parents recursively.
 
2122
            root_entries = _entries_for_path(current_root)
 
2123
            root_abspath = self.target.abspath(current_root_unicode)
 
2124
            try:
 
2125
                root_stat = os.lstat(root_abspath)
 
2126
            except OSError, e:
 
2127
                if e.errno == errno.ENOENT:
 
2128
                    # the path does not exist: let _process_entry know that.
 
2129
                    root_dir_info = None
 
2130
                else:
 
2131
                    # some other random error: hand it up.
 
2132
                    raise
 
2133
            else:
 
2134
                root_dir_info = ('', current_root,
 
2135
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2136
                    root_abspath)
 
2137
                if root_dir_info[2] == 'directory':
 
2138
                    if self.target._directory_is_tree_reference(
 
2139
                        current_root.decode('utf8')):
 
2140
                        root_dir_info = root_dir_info[:2] + \
 
2141
                            ('tree-reference',) + root_dir_info[3:]
 
2142
 
 
2143
            if not root_entries and not root_dir_info:
 
2144
                # this specified path is not present at all, skip it.
 
2145
                continue
 
2146
            path_handled = False
 
2147
            for entry in root_entries:
 
2148
                result = _process_entry(entry, root_dir_info)
 
2149
                if result is not None:
 
2150
                    path_handled = True
 
2151
                    if result is not uninteresting:
 
2152
                        yield result
 
2153
            if want_unversioned and not path_handled and root_dir_info:
 
2154
                new_executable = bool(
 
2155
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2156
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2157
                yield (None,
 
2158
                       (None, current_root_unicode),
 
2159
                       True,
 
2160
                       (False, False),
 
2161
                       (None, None),
 
2162
                       (None, splitpath(current_root_unicode)[-1]),
 
2163
                       (None, root_dir_info[2]),
 
2164
                       (None, new_executable)
 
2165
                      )
 
2166
            initial_key = (current_root, '', '')
 
2167
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2168
            if block_index == 0:
 
2169
                # we have processed the total root already, but because the
 
2170
                # initial key matched it we should skip it here.
 
2171
                block_index +=1
 
2172
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2173
                current_dir_info = None
 
2174
            else:
 
2175
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2176
                try:
 
2177
                    current_dir_info = dir_iterator.next()
 
2178
                except OSError, e:
 
2179
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2180
                    # python 2.5 has e.errno == EINVAL,
 
2181
                    #            and e.winerror == ERROR_DIRECTORY
 
2182
                    e_winerror = getattr(e, 'winerror', None)
 
2183
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2184
                    # there may be directories in the inventory even though
 
2185
                    # this path is not a file on disk: so mark it as end of
 
2186
                    # iterator
 
2187
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2188
                        current_dir_info = None
 
2189
                    elif (sys.platform == 'win32'
 
2190
                          and (e.errno in win_errors
 
2191
                               or e_winerror in win_errors)):
 
2192
                        current_dir_info = None
 
2193
                    else:
 
2194
                        raise
 
2195
                else:
 
2196
                    if current_dir_info[0][0] == '':
 
2197
                        # remove .bzr from iteration
 
2198
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2199
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2200
                        del current_dir_info[1][bzr_index]
 
2201
            # walk until both the directory listing and the versioned metadata
 
2202
            # are exhausted. 
 
2203
            if (block_index < len(state._dirblocks) and
 
2204
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2205
                current_block = state._dirblocks[block_index]
 
2206
            else:
 
2207
                current_block = None
 
2208
            while (current_dir_info is not None or
 
2209
                   current_block is not None):
 
2210
                if (current_dir_info and current_block
 
2211
                    and current_dir_info[0][0] != current_block[0]):
 
2212
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2213
                        # filesystem data refers to paths not covered by the dirblock.
 
2214
                        # this has two possibilities:
 
2215
                        # A) it is versioned but empty, so there is no block for it
 
2216
                        # B) it is not versioned.
 
2217
 
 
2218
                        # if (A) then we need to recurse into it to check for
 
2219
                        # new unknown files or directories.
 
2220
                        # if (B) then we should ignore it, because we don't
 
2221
                        # recurse into unknown directories.
 
2222
                        path_index = 0
 
2223
                        while path_index < len(current_dir_info[1]):
 
2224
                                current_path_info = current_dir_info[1][path_index]
 
2225
                                if want_unversioned:
 
2226
                                    if current_path_info[2] == 'directory':
 
2227
                                        if self.target._directory_is_tree_reference(
 
2228
                                            current_path_info[0].decode('utf8')):
 
2229
                                            current_path_info = current_path_info[:2] + \
 
2230
                                                ('tree-reference',) + current_path_info[3:]
 
2231
                                    new_executable = bool(
 
2232
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2233
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2234
                                    yield (None,
 
2235
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2236
                                        True,
 
2237
                                        (False, False),
 
2238
                                        (None, None),
 
2239
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2240
                                        (None, current_path_info[2]),
 
2241
                                        (None, new_executable))
 
2242
                                # dont descend into this unversioned path if it is
 
2243
                                # a dir
 
2244
                                if current_path_info[2] in ('directory',
 
2245
                                                            'tree-reference'):
 
2246
                                    del current_dir_info[1][path_index]
 
2247
                                    path_index -= 1
 
2248
                                path_index += 1
 
2249
 
 
2250
                        # This dir info has been handled, go to the next
 
2251
                        try:
 
2252
                            current_dir_info = dir_iterator.next()
 
2253
                        except StopIteration:
 
2254
                            current_dir_info = None
 
2255
                    else:
 
2256
                        # We have a dirblock entry for this location, but there
 
2257
                        # is no filesystem path for this. This is most likely
 
2258
                        # because a directory was removed from the disk.
 
2259
                        # We don't have to report the missing directory,
 
2260
                        # because that should have already been handled, but we
 
2261
                        # need to handle all of the files that are contained
 
2262
                        # within.
 
2263
                        for current_entry in current_block[1]:
 
2264
                            # entry referring to file not present on disk.
 
2265
                            # advance the entry only, after processing.
 
2266
                            result = _process_entry(current_entry, None)
 
2267
                            if result is not None:
 
2268
                                if result is not uninteresting:
 
2269
                                    yield result
 
2270
                        block_index +=1
 
2271
                        if (block_index < len(state._dirblocks) and
 
2272
                            osutils.is_inside(current_root,
 
2273
                                              state._dirblocks[block_index][0])):
 
2274
                            current_block = state._dirblocks[block_index]
 
2275
                        else:
 
2276
                            current_block = None
 
2277
                    continue
 
2278
                entry_index = 0
 
2279
                if current_block and entry_index < len(current_block[1]):
 
2280
                    current_entry = current_block[1][entry_index]
 
2281
                else:
 
2282
                    current_entry = None
 
2283
                advance_entry = True
 
2284
                path_index = 0
 
2285
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2286
                    current_path_info = current_dir_info[1][path_index]
 
2287
                    if current_path_info[2] == 'directory':
 
2288
                        if self.target._directory_is_tree_reference(
 
2289
                            current_path_info[0].decode('utf8')):
 
2290
                            current_path_info = current_path_info[:2] + \
 
2291
                                ('tree-reference',) + current_path_info[3:]
 
2292
                else:
 
2293
                    current_path_info = None
 
2294
                advance_path = True
 
2295
                path_handled = False
 
2296
                while (current_entry is not None or
 
2297
                    current_path_info is not None):
 
2298
                    if current_entry is None:
 
2299
                        # the check for path_handled when the path is adnvaced
 
2300
                        # will yield this path if needed.
 
2301
                        pass
 
2302
                    elif current_path_info is None:
 
2303
                        # no path is fine: the per entry code will handle it.
 
2304
                        result = _process_entry(current_entry, current_path_info)
 
2305
                        if result is not None:
 
2306
                            if result is not uninteresting:
 
2307
                                yield result
 
2308
                    elif (current_entry[0][1] != current_path_info[1]
 
2309
                          or current_entry[1][target_index][0] in 'ar'):
 
2310
                        # The current path on disk doesn't match the dirblock
 
2311
                        # record. Either the dirblock is marked as absent, or
 
2312
                        # the file on disk is not present at all in the
 
2313
                        # dirblock. Either way, report about the dirblock
 
2314
                        # entry, and let other code handle the filesystem one.
 
2315
 
 
2316
                        # Compare the basename for these files to determine
 
2317
                        # which comes first
 
2318
                        if current_path_info[1] < current_entry[0][1]:
 
2319
                            # extra file on disk: pass for now, but only
 
2320
                            # increment the path, not the entry
 
2321
                            advance_entry = False
 
2322
                        else:
 
2323
                            # entry referring to file not present on disk.
 
2324
                            # advance the entry only, after processing.
 
2325
                            result = _process_entry(current_entry, None)
 
2326
                            if result is not None:
 
2327
                                if result is not uninteresting:
 
2328
                                    yield result
 
2329
                            advance_path = False
 
2330
                    else:
 
2331
                        result = _process_entry(current_entry, current_path_info)
 
2332
                        if result is not None:
 
2333
                            path_handled = True
 
2334
                            if result is not uninteresting:
 
2335
                                yield result
 
2336
                    if advance_entry and current_entry is not None:
 
2337
                        entry_index += 1
 
2338
                        if entry_index < len(current_block[1]):
 
2339
                            current_entry = current_block[1][entry_index]
 
2340
                        else:
 
2341
                            current_entry = None
 
2342
                    else:
 
2343
                        advance_entry = True # reset the advance flaga
 
2344
                    if advance_path and current_path_info is not None:
 
2345
                        if not path_handled:
 
2346
                            # unversioned in all regards
 
2347
                            if want_unversioned:
 
2348
                                new_executable = bool(
 
2349
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2350
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2351
                                yield (None,
 
2352
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2353
                                    True,
 
2354
                                    (False, False),
 
2355
                                    (None, None),
 
2356
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2357
                                    (None, current_path_info[2]),
 
2358
                                    (None, new_executable))
 
2359
                            # dont descend into this unversioned path if it is
 
2360
                            # a dir
 
2361
                            if current_path_info[2] in ('directory'):
 
2362
                                del current_dir_info[1][path_index]
 
2363
                                path_index -= 1
 
2364
                        # dont descend the disk iterator into any tree 
 
2365
                        # paths.
 
2366
                        if current_path_info[2] == 'tree-reference':
 
2367
                            del current_dir_info[1][path_index]
 
2368
                            path_index -= 1
 
2369
                        path_index += 1
 
2370
                        if path_index < len(current_dir_info[1]):
 
2371
                            current_path_info = current_dir_info[1][path_index]
 
2372
                            if current_path_info[2] == 'directory':
 
2373
                                if self.target._directory_is_tree_reference(
 
2374
                                    current_path_info[0].decode('utf8')):
 
2375
                                    current_path_info = current_path_info[:2] + \
 
2376
                                        ('tree-reference',) + current_path_info[3:]
 
2377
                        else:
 
2378
                            current_path_info = None
 
2379
                        path_handled = False
 
2380
                    else:
 
2381
                        advance_path = True # reset the advance flagg.
 
2382
                if current_block is not None:
 
2383
                    block_index += 1
 
2384
                    if (block_index < len(state._dirblocks) and
 
2385
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2386
                        current_block = state._dirblocks[block_index]
 
2387
                    else:
 
2388
                        current_block = None
 
2389
                if current_dir_info is not None:
 
2390
                    try:
 
2391
                        current_dir_info = dir_iterator.next()
 
2392
                    except StopIteration:
 
2393
                        current_dir_info = None
 
2394
 
 
2395
 
 
2396
    @staticmethod
 
2397
    def is_compatible(source, target):
 
2398
        # the target must be a dirstate working tree
 
2399
        if not isinstance(target, WorkingTree4):
 
2400
            return False
 
2401
        # the source must be a revtreee or dirstate rev tree.
 
2402
        if not isinstance(source,
 
2403
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2404
            return False
 
2405
        # the source revid must be in the target dirstate
 
2406
        if not (source._revision_id == NULL_REVISION or
 
2407
            source._revision_id in target.get_parent_ids()):
 
2408
            # TODO: what about ghosts? it may well need to 
 
2409
            # check for them explicitly.
 
2410
            return False
 
2411
        return True
 
2412
 
 
2413
InterTree.register_optimiser(InterDirStateTree)
 
2414
 
 
2415
 
 
2416
class Converter3to4(object):
 
2417
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2418
 
 
2419
    def __init__(self):
 
2420
        self.target_format = WorkingTreeFormat4()
 
2421
 
 
2422
    def convert(self, tree):
 
2423
        # lock the control files not the tree, so that we dont get tree
 
2424
        # on-unlock behaviours, and so that noone else diddles with the 
 
2425
        # tree during upgrade.
 
2426
        tree._control_files.lock_write()
 
2427
        try:
 
2428
            tree.read_working_inventory()
 
2429
            self.create_dirstate_data(tree)
 
2430
            self.update_format(tree)
 
2431
            self.remove_xml_files(tree)
 
2432
        finally:
 
2433
            tree._control_files.unlock()
 
2434
 
 
2435
    def create_dirstate_data(self, tree):
 
2436
        """Create the dirstate based data for tree."""
 
2437
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2438
            ).local_abspath('dirstate')
 
2439
        state = dirstate.DirState.from_tree(tree, local_path)
 
2440
        state.save()
 
2441
        state.unlock()
 
2442
 
 
2443
    def remove_xml_files(self, tree):
 
2444
        """Remove the oldformat 3 data."""
 
2445
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2446
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2447
            'pending-merges', 'stat-cache']:
 
2448
            try:
 
2449
                transport.delete(path)
 
2450
            except errors.NoSuchFile:
 
2451
                # some files are optional - just deal.
 
2452
                pass
 
2453
 
 
2454
    def update_format(self, tree):
 
2455
        """Change the format marker."""
 
2456
        tree._control_files.put_utf8('format',
 
2457
            self.target_format.get_format_string())