/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: 2007-04-16 20:42:59 UTC
  • mto: This revision was merged to the branch mainline in revision 2422.
  • Revision ID: bialix@ukr.net-20070416204259-tyi3ptqe80gpkwcl
forget to return tree

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
    delta,
 
47
    dirstate,
 
48
    errors,
 
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revisiontree,
 
56
    textui,
 
57
    transform,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
 
61
    )
 
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
 
64
import bzrlib.ui
 
65
""")
 
66
 
 
67
from bzrlib import symbol_versioning
 
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
 
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
 
73
from bzrlib.mutabletree import needs_tree_write_lock
 
74
from bzrlib.osutils import (
 
75
    file_kind,
 
76
    isdir,
 
77
    normpath,
 
78
    pathjoin,
 
79
    rand_chars,
 
80
    realpath,
 
81
    safe_unicode,
 
82
    splitpath,
 
83
    )
 
84
from bzrlib.trace import mutter, note
 
85
from bzrlib.transport.local import LocalTransport
 
86
from bzrlib.tree import InterTree
 
87
from bzrlib.progress import DummyProgress, ProgressPhase
 
88
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
89
from bzrlib.rio import RioReader, rio_file, Stanza
 
90
from bzrlib.symbol_versioning import (deprecated_passed,
 
91
        deprecated_method,
 
92
        deprecated_function,
 
93
        DEPRECATED_PARAMETER,
 
94
        )
 
95
from bzrlib.tree import Tree
 
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
97
 
 
98
 
 
99
# This is the Windows equivalent of ENOTDIR
 
100
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
101
# just an error code.
 
102
ERROR_PATH_NOT_FOUND = 3
 
103
ERROR_DIRECTORY = 267
 
104
 
 
105
 
 
106
class WorkingTree4(WorkingTree3):
 
107
    """This is the Format 4 working tree.
 
108
 
 
109
    This differs from WorkingTree3 by:
 
110
     - Having a consolidated internal dirstate, stored in a
 
111
       randomly-accessible sorted file on disk.
 
112
     - Not having a regular inventory attribute.  One can be synthesized 
 
113
       on demand but this is expensive and should be avoided.
 
114
 
 
115
    This is new in bzr 0.15.
 
116
    """
 
117
 
 
118
    def __init__(self, basedir,
 
119
                 branch,
 
120
                 _control_files=None,
 
121
                 _format=None,
 
122
                 _bzrdir=None):
 
123
        """Construct a WorkingTree for basedir.
 
124
 
 
125
        If the branch is not supplied, it is opened automatically.
 
126
        If the branch is supplied, it must be the branch for this basedir.
 
127
        (branch.base is not cross checked, because for remote branches that
 
128
        would be meaningless).
 
129
        """
 
130
        self._format = _format
 
131
        self.bzrdir = _bzrdir
 
132
        from bzrlib.trace import note, mutter
 
133
        assert isinstance(basedir, basestring), \
 
134
            "base directory %r is not a string" % basedir
 
135
        basedir = safe_unicode(basedir)
 
136
        mutter("opening working tree %r", basedir)
 
137
        self._branch = branch
 
138
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
139
            "branch %r is not a Branch" % self.branch
 
140
        self.basedir = realpath(basedir)
 
141
        # if branch is at our basedir and is a format 6 or less
 
142
        # assume all other formats have their own control files.
 
143
        assert isinstance(_control_files, LockableFiles), \
 
144
            "_control_files must be a LockableFiles, not %r" % _control_files
 
145
        self._control_files = _control_files
 
146
        self._dirty = None
 
147
        #-------------
 
148
        # during a read or write lock these objects are set, and are
 
149
        # None the rest of the time.
 
150
        self._dirstate = None
 
151
        self._inventory = None
 
152
        #-------------
 
153
 
 
154
    @needs_tree_write_lock
 
155
    def _add(self, files, ids, kinds):
 
156
        """See MutableTree._add."""
 
157
        state = self.current_dirstate()
 
158
        for f, file_id, kind in zip(files, ids, kinds):
 
159
            f = f.strip('/')
 
160
            assert '//' not in f
 
161
            assert '..' not in f
 
162
            if self.path2id(f):
 
163
                # special case tree root handling.
 
164
                if f == '' and self.path2id(f) == ROOT_ID:
 
165
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
166
                continue
 
167
            if file_id is None:
 
168
                file_id = generate_ids.gen_file_id(f)
 
169
            # deliberately add the file with no cached stat or sha1
 
170
            # - on the first access it will be gathered, and we can
 
171
            # always change this once tests are all passing.
 
172
            state.add(f, file_id, kind, None, '')
 
173
        self._make_dirty(reset_inventory=True)
 
174
 
 
175
    def _make_dirty(self, reset_inventory):
 
176
        """Make the tree state dirty.
 
177
 
 
178
        :param reset_inventory: True if the cached inventory should be removed
 
179
            (presuming there is one).
 
180
        """
 
181
        self._dirty = True
 
182
        if reset_inventory and self._inventory is not None:
 
183
            self._inventory = None
 
184
 
 
185
    @needs_tree_write_lock
 
186
    def add_reference(self, sub_tree):
 
187
        # use standard implementation, which calls back to self._add
 
188
        # 
 
189
        # So we don't store the reference_revision in the working dirstate,
 
190
        # it's just recorded at the moment of commit. 
 
191
        self._add_reference(sub_tree)
 
192
 
 
193
    def break_lock(self):
 
194
        """Break a lock if one is present from another instance.
 
195
 
 
196
        Uses the ui factory to ask for confirmation if the lock may be from
 
197
        an active process.
 
198
 
 
199
        This will probe the repository for its lock as well.
 
200
        """
 
201
        # if the dirstate is locked by an active process, reject the break lock
 
202
        # call.
 
203
        try:
 
204
            if self._dirstate is None:
 
205
                clear = True
 
206
            else:
 
207
                clear = False
 
208
            state = self._current_dirstate()
 
209
            if state._lock_token is not None:
 
210
                # we already have it locked. sheese, cant break our own lock.
 
211
                raise errors.LockActive(self.basedir)
 
212
            else:
 
213
                try:
 
214
                    # try for a write lock - need permission to get one anyhow
 
215
                    # to break locks.
 
216
                    state.lock_write()
 
217
                except errors.LockContention:
 
218
                    # oslocks fail when a process is still live: fail.
 
219
                    # TODO: get the locked lockdir info and give to the user to
 
220
                    # assist in debugging.
 
221
                    raise errors.LockActive(self.basedir)
 
222
                else:
 
223
                    state.unlock()
 
224
        finally:
 
225
            if clear:
 
226
                self._dirstate = None
 
227
        self._control_files.break_lock()
 
228
        self.branch.break_lock()
 
229
 
 
230
    def _comparison_data(self, entry, path):
 
231
        kind, executable, stat_value = \
 
232
            WorkingTree3._comparison_data(self, entry, path)
 
233
        # it looks like a plain directory, but it's really a reference -- see
 
234
        # also kind()
 
235
        if (self._repo_supports_tree_reference and
 
236
            kind == 'directory' and
 
237
            self._directory_is_tree_reference(path)):
 
238
            kind = 'tree-reference'
 
239
        return kind, executable, stat_value
 
240
 
 
241
    @needs_write_lock
 
242
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
243
        # mark the tree as dirty post commit - commit
 
244
        # can change the current versioned list by doing deletes.
 
245
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
246
        self._make_dirty(reset_inventory=True)
 
247
        return result
 
248
 
 
249
    def current_dirstate(self):
 
250
        """Return the current dirstate object.
 
251
 
 
252
        This is not part of the tree interface and only exposed for ease of
 
253
        testing.
 
254
 
 
255
        :raises errors.NotWriteLocked: when not in a lock.
 
256
        """
 
257
        self._must_be_locked()
 
258
        return self._current_dirstate()
 
259
 
 
260
    def _current_dirstate(self):
 
261
        """Internal function that does not check lock status.
 
262
 
 
263
        This is needed for break_lock which also needs the dirstate.
 
264
        """
 
265
        if self._dirstate is not None:
 
266
            return self._dirstate
 
267
        local_path = self.bzrdir.get_workingtree_transport(None
 
268
            ).local_abspath('dirstate')
 
269
        self._dirstate = dirstate.DirState.on_file(local_path)
 
270
        return self._dirstate
 
271
 
 
272
    def _directory_is_tree_reference(self, relpath):
 
273
        # as a special case, if a directory contains control files then 
 
274
        # it's a tree reference, except that the root of the tree is not
 
275
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
276
        # TODO: We could ask all the control formats whether they
 
277
        # recognize this directory, but at the moment there's no cheap api
 
278
        # to do that.  Since we probably can only nest bzr checkouts and
 
279
        # they always use this name it's ok for now.  -- mbp 20060306
 
280
        #
 
281
        # FIXME: There is an unhandled case here of a subdirectory
 
282
        # containing .bzr but not a branch; that will probably blow up
 
283
        # when you try to commit it.  It might happen if there is a
 
284
        # checkout in a subdirectory.  This can be avoided by not adding
 
285
        # it.  mbp 20070306
 
286
 
 
287
    def filter_unversioned_files(self, paths):
 
288
        """Filter out paths that are versioned.
 
289
 
 
290
        :return: set of paths.
 
291
        """
 
292
        # TODO: make a generic multi-bisect routine roughly that should list
 
293
        # the paths, then process one half at a time recursively, and feed the
 
294
        # results of each bisect in further still
 
295
        paths = sorted(paths)
 
296
        result = set()
 
297
        state = self.current_dirstate()
 
298
        # TODO we want a paths_to_dirblocks helper I think
 
299
        for path in paths:
 
300
            dirname, basename = os.path.split(path.encode('utf8'))
 
301
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
302
                dirname, basename, 0)
 
303
            if not path_is_versioned:
 
304
                result.add(path)
 
305
        return result
 
306
 
 
307
    def flush(self):
 
308
        """Write all cached data to disk."""
 
309
        if self._control_files._lock_mode != 'w':
 
310
            raise errors.NotWriteLocked(self)
 
311
        self.current_dirstate().save()
 
312
        self._inventory = None
 
313
        self._dirty = False
 
314
 
 
315
    @needs_tree_write_lock
 
316
    def _gather_kinds(self, files, kinds):
 
317
        """See MutableTree._gather_kinds."""
 
318
        for pos, f in enumerate(files):
 
319
            if kinds[pos] is None:
 
320
                kinds[pos] = self._kind(f)
 
321
 
 
322
    def _generate_inventory(self):
 
323
        """Create and set self.inventory from the dirstate object.
 
324
        
 
325
        This is relatively expensive: we have to walk the entire dirstate.
 
326
        Ideally we would not, and can deprecate this function.
 
327
        """
 
328
        #: uncomment to trap on inventory requests.
 
329
        # import pdb;pdb.set_trace()
 
330
        state = self.current_dirstate()
 
331
        state._read_dirblocks_if_needed()
 
332
        root_key, current_entry = self._get_entry(path='')
 
333
        current_id = root_key[2]
 
334
        assert current_entry[0][0] == 'd' # directory
 
335
        inv = Inventory(root_id=current_id)
 
336
        # Turn some things into local variables
 
337
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
338
        factory = entry_factory
 
339
        utf8_decode = cache_utf8._utf8_decode
 
340
        inv_byid = inv._byid
 
341
        # we could do this straight out of the dirstate; it might be fast
 
342
        # and should be profiled - RBC 20070216
 
343
        parent_ies = {'' : inv.root}
 
344
        for block in state._dirblocks[1:]: # skip the root
 
345
            dirname = block[0]
 
346
            try:
 
347
                parent_ie = parent_ies[dirname]
 
348
            except KeyError:
 
349
                # all the paths in this block are not versioned in this tree
 
350
                continue
 
351
            for key, entry in block[1]:
 
352
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
353
                if minikind in ('a', 'r'): # absent, relocated
 
354
                    # a parent tree only entry
 
355
                    continue
 
356
                name = key[1]
 
357
                name_unicode = utf8_decode(name)[0]
 
358
                file_id = key[2]
 
359
                kind = minikind_to_kind[minikind]
 
360
                inv_entry = factory[kind](file_id, name_unicode,
 
361
                                          parent_ie.file_id)
 
362
                if kind == 'file':
 
363
                    # This is only needed on win32, where this is the only way
 
364
                    # we know the executable bit.
 
365
                    inv_entry.executable = executable
 
366
                    # not strictly needed: working tree
 
367
                    #inv_entry.text_size = size
 
368
                    #inv_entry.text_sha1 = sha1
 
369
                elif kind == 'directory':
 
370
                    # add this entry to the parent map.
 
371
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
372
                elif kind == 'tree-reference':
 
373
                    assert self._repo_supports_tree_reference, \
 
374
                        "repository of %r " \
 
375
                        "doesn't support tree references " \
 
376
                        "required by entry %r" \
 
377
                        % (self, name)
 
378
                    inv_entry.reference_revision = link_or_sha1 or None
 
379
                elif kind != 'symlink':
 
380
                    raise AssertionError("unknown kind %r" % kind)
 
381
                # These checks cost us around 40ms on a 55k entry tree
 
382
                assert file_id not in inv_byid, ('file_id %s already in'
 
383
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
384
                assert name_unicode not in parent_ie.children
 
385
                inv_byid[file_id] = inv_entry
 
386
                parent_ie.children[name_unicode] = inv_entry
 
387
        self._inventory = inv
 
388
 
 
389
    def _get_entry(self, file_id=None, path=None):
 
390
        """Get the dirstate row for file_id or path.
 
391
 
 
392
        If either file_id or path is supplied, it is used as the key to lookup.
 
393
        If both are supplied, the fastest lookup is used, and an error is
 
394
        raised if they do not both point at the same row.
 
395
        
 
396
        :param file_id: An optional unicode file_id to be looked up.
 
397
        :param path: An optional unicode path to be looked up.
 
398
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
399
        """
 
400
        if file_id is None and path is None:
 
401
            raise errors.BzrError('must supply file_id or path')
 
402
        state = self.current_dirstate()
 
403
        if path is not None:
 
404
            path = path.encode('utf8')
 
405
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
406
 
 
407
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
408
        # check file id is valid unconditionally.
 
409
        entry = self._get_entry(file_id=file_id, path=path)
 
410
        assert entry[0] is not None, 'what error should this raise'
 
411
        if path is None:
 
412
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
413
 
 
414
        file_abspath = self.abspath(path)
 
415
        state = self.current_dirstate()
 
416
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
417
                                          stat_value=stat_value)
 
418
        if entry[1][0][0] == 'f':
 
419
            return link_or_sha1
 
420
        return None
 
421
 
 
422
    def _get_inventory(self):
 
423
        """Get the inventory for the tree. This is only valid within a lock."""
 
424
        if self._inventory is not None:
 
425
            return self._inventory
 
426
        self._must_be_locked()
 
427
        self._generate_inventory()
 
428
        return self._inventory
 
429
 
 
430
    inventory = property(_get_inventory,
 
431
                         doc="Inventory of this Tree")
 
432
 
 
433
    @needs_read_lock
 
434
    def get_parent_ids(self):
 
435
        """See Tree.get_parent_ids.
 
436
        
 
437
        This implementation requests the ids list from the dirstate file.
 
438
        """
 
439
        return self.current_dirstate().get_parent_ids()
 
440
 
 
441
    def get_reference_revision(self, file_id, path=None):
 
442
        # referenced tree's revision is whatever's currently there
 
443
        return self.get_nested_tree(file_id, path).last_revision()
 
444
 
 
445
    def get_nested_tree(self, file_id, path=None):
 
446
        if path is None:
 
447
            path = self.id2path(file_id)
 
448
        # else: check file_id is at path?
 
449
        return WorkingTree.open(self.abspath(path))
 
450
 
 
451
    @needs_read_lock
 
452
    def get_root_id(self):
 
453
        """Return the id of this trees root"""
 
454
        return self._get_entry(path='')[0][2]
 
455
 
 
456
    def has_id(self, file_id):
 
457
        state = self.current_dirstate()
 
458
        file_id = osutils.safe_file_id(file_id)
 
459
        row, parents = self._get_entry(file_id=file_id)
 
460
        if row is None:
 
461
            return False
 
462
        return osutils.lexists(pathjoin(
 
463
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
464
 
 
465
    @needs_read_lock
 
466
    def id2path(self, file_id):
 
467
        file_id = osutils.safe_file_id(file_id)
 
468
        state = self.current_dirstate()
 
469
        entry = self._get_entry(file_id=file_id)
 
470
        if entry == (None, None):
 
471
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
472
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
473
        return path_utf8.decode('utf8')
 
474
 
 
475
    if not osutils.supports_executable():
 
476
        @needs_read_lock
 
477
        def is_executable(self, file_id, path=None):
 
478
            file_id = osutils.safe_file_id(file_id)
 
479
            entry = self._get_entry(file_id=file_id, path=path)
 
480
            if entry == (None, None):
 
481
                return False
 
482
            return entry[1][0][3]
 
483
    else:
 
484
        @needs_read_lock
 
485
        def is_executable(self, file_id, path=None):
 
486
            if not path:
 
487
                file_id = osutils.safe_file_id(file_id)
 
488
                path = self.id2path(file_id)
 
489
            mode = os.lstat(self.abspath(path)).st_mode
 
490
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
491
 
 
492
    @needs_read_lock
 
493
    def __iter__(self):
 
494
        """Iterate through file_ids for this tree.
 
495
 
 
496
        file_ids are in a WorkingTree if they are in the working inventory
 
497
        and the working file exists.
 
498
        """
 
499
        result = []
 
500
        for key, tree_details in self.current_dirstate()._iter_entries():
 
501
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
502
                # not relevant to the working tree
 
503
                continue
 
504
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
505
            if osutils.lexists(path):
 
506
                result.append(key[2])
 
507
        return iter(result)
 
508
 
 
509
    def iter_references(self):
 
510
        for key, tree_details in self.current_dirstate()._iter_entries():
 
511
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
512
                # not relevant to the working tree
 
513
                continue
 
514
            if not key[1]:
 
515
                # the root is not a reference.
 
516
                continue
 
517
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
518
            try:
 
519
                if self._kind(path) == 'tree-reference':
 
520
                    yield path, key[2]
 
521
            except errors.NoSuchFile:
 
522
                # path is missing on disk.
 
523
                continue
 
524
 
 
525
    @needs_read_lock
 
526
    def kind(self, file_id):
 
527
        """Return the kind of a file.
 
528
 
 
529
        This is always the actual kind that's on disk, regardless of what it
 
530
        was added as.
 
531
        """
 
532
        relpath = self.id2path(file_id)
 
533
        assert relpath != None, \
 
534
            "path for id {%s} is None!" % file_id
 
535
        return self._kind(relpath)
 
536
 
 
537
    def _kind(self, relpath):
 
538
        abspath = self.abspath(relpath)
 
539
        kind = file_kind(abspath)
 
540
        if (self._repo_supports_tree_reference and
 
541
            kind == 'directory' and
 
542
            self._directory_is_tree_reference(relpath)):
 
543
            kind = 'tree-reference'
 
544
        return kind
 
545
 
 
546
    @needs_read_lock
 
547
    def _last_revision(self):
 
548
        """See Mutable.last_revision."""
 
549
        parent_ids = self.current_dirstate().get_parent_ids()
 
550
        if parent_ids:
 
551
            return parent_ids[0]
 
552
        else:
 
553
            return None
 
554
 
 
555
    def lock_read(self):
 
556
        """See Branch.lock_read, and WorkingTree.unlock."""
 
557
        self.branch.lock_read()
 
558
        try:
 
559
            self._control_files.lock_read()
 
560
            try:
 
561
                state = self.current_dirstate()
 
562
                if not state._lock_token:
 
563
                    state.lock_read()
 
564
                # set our support for tree references from the repository in
 
565
                # use.
 
566
                self._repo_supports_tree_reference = getattr(
 
567
                    self.branch.repository._format, "supports_tree_reference",
 
568
                    False)
 
569
            except:
 
570
                self._control_files.unlock()
 
571
                raise
 
572
        except:
 
573
            self.branch.unlock()
 
574
            raise
 
575
 
 
576
    def _lock_self_write(self):
 
577
        """This should be called after the branch is locked."""
 
578
        try:
 
579
            self._control_files.lock_write()
 
580
            try:
 
581
                state = self.current_dirstate()
 
582
                if not state._lock_token:
 
583
                    state.lock_write()
 
584
                # set our support for tree references from the repository in
 
585
                # use.
 
586
                self._repo_supports_tree_reference = getattr(
 
587
                    self.branch.repository._format, "supports_tree_reference",
 
588
                    False)
 
589
            except:
 
590
                self._control_files.unlock()
 
591
                raise
 
592
        except:
 
593
            self.branch.unlock()
 
594
            raise
 
595
 
 
596
    def lock_tree_write(self):
 
597
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
598
        self.branch.lock_read()
 
599
        self._lock_self_write()
 
600
 
 
601
    def lock_write(self):
 
602
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
603
        self.branch.lock_write()
 
604
        self._lock_self_write()
 
605
 
 
606
    @needs_tree_write_lock
 
607
    def move(self, from_paths, to_dir, after=False):
 
608
        """See WorkingTree.move()."""
 
609
        result = []
 
610
        if not from_paths:
 
611
            return result
 
612
 
 
613
        state = self.current_dirstate()
 
614
 
 
615
        assert not isinstance(from_paths, basestring)
 
616
        to_dir_utf8 = to_dir.encode('utf8')
 
617
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
618
        id_index = state._get_id_index()
 
619
        # check destination directory
 
620
        # get the details for it
 
621
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
622
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
623
        if not entry_present:
 
624
            raise errors.BzrMoveFailedError('', to_dir,
 
625
                errors.NotVersionedError(to_dir))
 
626
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
627
        # get a handle on the block itself.
 
628
        to_block_index = state._ensure_block(
 
629
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
630
        to_block = state._dirblocks[to_block_index]
 
631
        to_abs = self.abspath(to_dir)
 
632
        if not isdir(to_abs):
 
633
            raise errors.BzrMoveFailedError('',to_dir,
 
634
                errors.NotADirectory(to_abs))
 
635
 
 
636
        if to_entry[1][0][0] != 'd':
 
637
            raise errors.BzrMoveFailedError('',to_dir,
 
638
                errors.NotADirectory(to_abs))
 
639
 
 
640
        if self._inventory is not None:
 
641
            update_inventory = True
 
642
            inv = self.inventory
 
643
            to_dir_ie = inv[to_dir_id]
 
644
            to_dir_id = to_entry[0][2]
 
645
        else:
 
646
            update_inventory = False
 
647
 
 
648
        rollbacks = []
 
649
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
650
                     fingerprint, packed_stat, size,
 
651
                     to_block, to_key, to_path_utf8):
 
652
            state._make_absent(old_entry)
 
653
            from_key = old_entry[0]
 
654
            rollbacks.append(
 
655
                lambda:state.update_minimal(from_key,
 
656
                    minikind,
 
657
                    executable=executable,
 
658
                    fingerprint=fingerprint,
 
659
                    packed_stat=packed_stat,
 
660
                    size=size,
 
661
                    path_utf8=from_path_utf8))
 
662
            state.update_minimal(to_key,
 
663
                    minikind,
 
664
                    executable=executable,
 
665
                    fingerprint=fingerprint,
 
666
                    packed_stat=packed_stat,
 
667
                    size=size,
 
668
                    path_utf8=to_path_utf8)
 
669
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
670
            new_entry = to_block[1][added_entry_index]
 
671
            rollbacks.append(lambda:state._make_absent(new_entry))
 
672
 
 
673
        # create rename entries and tuples
 
674
        for from_rel in from_paths:
 
675
            # from_rel is 'pathinroot/foo/bar'
 
676
            from_rel_utf8 = from_rel.encode('utf8')
 
677
            from_dirname, from_tail = osutils.split(from_rel)
 
678
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
679
            from_entry = self._get_entry(path=from_rel)
 
680
            if from_entry == (None, None):
 
681
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
682
                    errors.NotVersionedError(path=str(from_rel)))
 
683
 
 
684
            from_id = from_entry[0][2]
 
685
            to_rel = pathjoin(to_dir, from_tail)
 
686
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
687
            item_to_entry = self._get_entry(path=to_rel)
 
688
            if item_to_entry != (None, None):
 
689
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
690
                    "Target is already versioned.")
 
691
 
 
692
            if from_rel == to_rel:
 
693
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
694
                    "Source and target are identical.")
 
695
 
 
696
            from_missing = not self.has_filename(from_rel)
 
697
            to_missing = not self.has_filename(to_rel)
 
698
            if after:
 
699
                move_file = False
 
700
            else:
 
701
                move_file = True
 
702
            if to_missing:
 
703
                if not move_file:
 
704
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
705
                        errors.NoSuchFile(path=to_rel,
 
706
                        extra="New file has not been created yet"))
 
707
                elif from_missing:
 
708
                    # neither path exists
 
709
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
710
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
711
            else:
 
712
                if from_missing: # implicitly just update our path mapping
 
713
                    move_file = False
 
714
                elif not after:
 
715
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
716
                        extra="(Use --after to update the Bazaar id)")
 
717
 
 
718
            rollbacks = []
 
719
            def rollback_rename():
 
720
                """A single rename has failed, roll it back."""
 
721
                # roll back everything, even if we encounter trouble doing one
 
722
                # of them.
 
723
                #
 
724
                # TODO: at least log the other exceptions rather than just
 
725
                # losing them mbp 20070307
 
726
                exc_info = None
 
727
                for rollback in reversed(rollbacks):
 
728
                    try:
 
729
                        rollback()
 
730
                    except Exception, e:
 
731
                        exc_info = sys.exc_info()
 
732
                if exc_info:
 
733
                    raise exc_info[0], exc_info[1], exc_info[2]
 
734
 
 
735
            # perform the disk move first - its the most likely failure point.
 
736
            if move_file:
 
737
                from_rel_abs = self.abspath(from_rel)
 
738
                to_rel_abs = self.abspath(to_rel)
 
739
                try:
 
740
                    osutils.rename(from_rel_abs, to_rel_abs)
 
741
                except OSError, e:
 
742
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
743
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
744
            try:
 
745
                # perform the rename in the inventory next if needed: its easy
 
746
                # to rollback
 
747
                if update_inventory:
 
748
                    # rename the entry
 
749
                    from_entry = inv[from_id]
 
750
                    current_parent = from_entry.parent_id
 
751
                    inv.rename(from_id, to_dir_id, from_tail)
 
752
                    rollbacks.append(
 
753
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
754
                # finally do the rename in the dirstate, which is a little
 
755
                # tricky to rollback, but least likely to need it.
 
756
                old_block_index, old_entry_index, dir_present, file_present = \
 
757
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
758
                old_block = state._dirblocks[old_block_index][1]
 
759
                old_entry = old_block[old_entry_index]
 
760
                from_key, old_entry_details = old_entry
 
761
                cur_details = old_entry_details[0]
 
762
                # remove the old row
 
763
                to_key = ((to_block[0],) + from_key[1:3])
 
764
                minikind = cur_details[0]
 
765
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
766
                         minikind=minikind,
 
767
                         executable=cur_details[3],
 
768
                         fingerprint=cur_details[1],
 
769
                         packed_stat=cur_details[4],
 
770
                         size=cur_details[2],
 
771
                         to_block=to_block,
 
772
                         to_key=to_key,
 
773
                         to_path_utf8=to_rel_utf8)
 
774
 
 
775
                if minikind == 'd':
 
776
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
777
                        """all entries in this block need updating.
 
778
 
 
779
                        TODO: This is pretty ugly, and doesn't support
 
780
                        reverting, but it works.
 
781
                        """
 
782
                        assert from_dir != '', "renaming root not supported"
 
783
                        from_key = (from_dir, '')
 
784
                        from_block_idx, present = \
 
785
                            state._find_block_index_from_key(from_key)
 
786
                        if not present:
 
787
                            # This is the old record, if it isn't present, then
 
788
                            # there is theoretically nothing to update.
 
789
                            # (Unless it isn't present because of lazy loading,
 
790
                            # but we don't do that yet)
 
791
                            return
 
792
                        from_block = state._dirblocks[from_block_idx]
 
793
                        to_block_index, to_entry_index, _, _ = \
 
794
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
795
                        to_block_index = state._ensure_block(
 
796
                            to_block_index, to_entry_index, to_dir_utf8)
 
797
                        to_block = state._dirblocks[to_block_index]
 
798
                        for entry in from_block[1]:
 
799
                            assert entry[0][0] == from_dir
 
800
                            cur_details = entry[1][0]
 
801
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
802
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
803
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
804
                            minikind = cur_details[0]
 
805
                            move_one(entry, from_path_utf8=from_path_utf8,
 
806
                                     minikind=minikind,
 
807
                                     executable=cur_details[3],
 
808
                                     fingerprint=cur_details[1],
 
809
                                     packed_stat=cur_details[4],
 
810
                                     size=cur_details[2],
 
811
                                     to_block=to_block,
 
812
                                     to_key=to_key,
 
813
                                     to_path_utf8=to_path_utf8)
 
814
                            if minikind == 'd':
 
815
                                # We need to move all the children of this
 
816
                                # entry
 
817
                                update_dirblock(from_path_utf8, to_key,
 
818
                                                to_path_utf8)
 
819
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
820
            except:
 
821
                rollback_rename()
 
822
                raise
 
823
            result.append((from_rel, to_rel))
 
824
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
825
            self._make_dirty(reset_inventory=False)
 
826
 
 
827
        return result
 
828
 
 
829
    def _must_be_locked(self):
 
830
        if not self._control_files._lock_count:
 
831
            raise errors.ObjectNotLocked(self)
 
832
 
 
833
    def _new_tree(self):
 
834
        """Initialize the state in this tree to be a new tree."""
 
835
        self._dirty = True
 
836
 
 
837
    @needs_read_lock
 
838
    def path2id(self, path):
 
839
        """Return the id for path in this tree."""
 
840
        path = path.strip('/')
 
841
        entry = self._get_entry(path=path)
 
842
        if entry == (None, None):
 
843
            return None
 
844
        return entry[0][2]
 
845
 
 
846
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
847
        """See Tree.paths2ids().
 
848
 
 
849
        This specialisation fast-paths the case where all the trees are in the
 
850
        dirstate.
 
851
        """
 
852
        if paths is None:
 
853
            return None
 
854
        parents = self.get_parent_ids()
 
855
        for tree in trees:
 
856
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
857
                parents):
 
858
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
859
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
860
        # -- make all paths utf8 --
 
861
        paths_utf8 = set()
 
862
        for path in paths:
 
863
            paths_utf8.add(path.encode('utf8'))
 
864
        paths = paths_utf8
 
865
        # -- paths is now a utf8 path set --
 
866
        # -- get the state object and prepare it.
 
867
        state = self.current_dirstate()
 
868
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
869
            and '' not in paths):
 
870
            paths2ids = self._paths2ids_using_bisect
 
871
        else:
 
872
            paths2ids = self._paths2ids_in_memory
 
873
        return paths2ids(paths, search_indexes,
 
874
                         require_versioned=require_versioned)
 
875
 
 
876
    def _paths2ids_in_memory(self, paths, search_indexes,
 
877
                             require_versioned=True):
 
878
        state = self.current_dirstate()
 
879
        state._read_dirblocks_if_needed()
 
880
        def _entries_for_path(path):
 
881
            """Return a list with all the entries that match path for all ids.
 
882
            """
 
883
            dirname, basename = os.path.split(path)
 
884
            key = (dirname, basename, '')
 
885
            block_index, present = state._find_block_index_from_key(key)
 
886
            if not present:
 
887
                # the block which should contain path is absent.
 
888
                return []
 
889
            result = []
 
890
            block = state._dirblocks[block_index][1]
 
891
            entry_index, _ = state._find_entry_index(key, block)
 
892
            # we may need to look at multiple entries at this path: walk while the paths match.
 
893
            while (entry_index < len(block) and
 
894
                block[entry_index][0][0:2] == key[0:2]):
 
895
                result.append(block[entry_index])
 
896
                entry_index += 1
 
897
            return result
 
898
        if require_versioned:
 
899
            # -- check all supplied paths are versioned in a search tree. --
 
900
            all_versioned = True
 
901
            for path in paths:
 
902
                path_entries = _entries_for_path(path)
 
903
                if not path_entries:
 
904
                    # this specified path is not present at all: error
 
905
                    all_versioned = False
 
906
                    break
 
907
                found_versioned = False
 
908
                # for each id at this path
 
909
                for entry in path_entries:
 
910
                    # for each tree.
 
911
                    for index in search_indexes:
 
912
                        if entry[1][index][0] != 'a': # absent
 
913
                            found_versioned = True
 
914
                            # all good: found a versioned cell
 
915
                            break
 
916
                if not found_versioned:
 
917
                    # none of the indexes was not 'absent' at all ids for this
 
918
                    # path.
 
919
                    all_versioned = False
 
920
                    break
 
921
            if not all_versioned:
 
922
                raise errors.PathsNotVersionedError(paths)
 
923
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
924
        search_paths = set()
 
925
        for path in paths:
 
926
            other_paths = paths.difference(set([path]))
 
927
            if not osutils.is_inside_any(other_paths, path):
 
928
                # this is a top level path, we must check it.
 
929
                search_paths.add(path)
 
930
        # sketch: 
 
931
        # for all search_indexs in each path at or under each element of
 
932
        # search_paths, if the detail is relocated: add the id, and add the
 
933
        # relocated path as one to search if its not searched already. If the
 
934
        # detail is not relocated, add the id.
 
935
        searched_paths = set()
 
936
        found_ids = set()
 
937
        def _process_entry(entry):
 
938
            """Look at search_indexes within entry.
 
939
 
 
940
            If a specific tree's details are relocated, add the relocation
 
941
            target to search_paths if not searched already. If it is absent, do
 
942
            nothing. Otherwise add the id to found_ids.
 
943
            """
 
944
            for index in search_indexes:
 
945
                if entry[1][index][0] == 'r': # relocated
 
946
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
947
                        search_paths.add(entry[1][index][1])
 
948
                elif entry[1][index][0] != 'a': # absent
 
949
                    found_ids.add(entry[0][2])
 
950
        while search_paths:
 
951
            current_root = search_paths.pop()
 
952
            searched_paths.add(current_root)
 
953
            # process the entries for this containing directory: the rest will be
 
954
            # found by their parents recursively.
 
955
            root_entries = _entries_for_path(current_root)
 
956
            if not root_entries:
 
957
                # this specified path is not present at all, skip it.
 
958
                continue
 
959
            for entry in root_entries:
 
960
                _process_entry(entry)
 
961
            initial_key = (current_root, '', '')
 
962
            block_index, _ = state._find_block_index_from_key(initial_key)
 
963
            while (block_index < len(state._dirblocks) and
 
964
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
965
                for entry in state._dirblocks[block_index][1]:
 
966
                    _process_entry(entry)
 
967
                block_index += 1
 
968
        return found_ids
 
969
 
 
970
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
971
                                require_versioned=True):
 
972
        state = self.current_dirstate()
 
973
        found_ids = set()
 
974
 
 
975
        split_paths = sorted(osutils.split(p) for p in paths)
 
976
        found = state._bisect_recursive(split_paths)
 
977
 
 
978
        if require_versioned:
 
979
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
980
            for dir_name in split_paths:
 
981
                if dir_name not in found_dir_names:
 
982
                    raise errors.PathsNotVersionedError(paths)
 
983
 
 
984
        for dir_name_id, trees_info in found.iteritems():
 
985
            for index in search_indexes:
 
986
                if trees_info[index][0] not in ('r', 'a'):
 
987
                    found_ids.add(dir_name_id[2])
 
988
        return found_ids
 
989
 
 
990
    def read_working_inventory(self):
 
991
        """Read the working inventory.
 
992
        
 
993
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
994
        """
 
995
        return self.inventory
 
996
 
 
997
    @needs_read_lock
 
998
    def revision_tree(self, revision_id):
 
999
        """See Tree.revision_tree.
 
1000
 
 
1001
        WorkingTree4 supplies revision_trees for any basis tree.
 
1002
        """
 
1003
        revision_id = osutils.safe_revision_id(revision_id)
 
1004
        dirstate = self.current_dirstate()
 
1005
        parent_ids = dirstate.get_parent_ids()
 
1006
        if revision_id not in parent_ids:
 
1007
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1008
        if revision_id in dirstate.get_ghosts():
 
1009
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1010
        return DirStateRevisionTree(dirstate, revision_id,
 
1011
            self.branch.repository)
 
1012
 
 
1013
    @needs_tree_write_lock
 
1014
    def set_last_revision(self, new_revision):
 
1015
        """Change the last revision in the working tree."""
 
1016
        new_revision = osutils.safe_revision_id(new_revision)
 
1017
        parents = self.get_parent_ids()
 
1018
        if new_revision in (NULL_REVISION, None):
 
1019
            assert len(parents) < 2, (
 
1020
                "setting the last parent to none with a pending merge is "
 
1021
                "unsupported.")
 
1022
            self.set_parent_ids([])
 
1023
        else:
 
1024
            self.set_parent_ids([new_revision] + parents[1:],
 
1025
                allow_leftmost_as_ghost=True)
 
1026
 
 
1027
    @needs_tree_write_lock
 
1028
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1029
        """Set the parent ids to revision_ids.
 
1030
        
 
1031
        See also set_parent_trees. This api will try to retrieve the tree data
 
1032
        for each element of revision_ids from the trees repository. If you have
 
1033
        tree data already available, it is more efficient to use
 
1034
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1035
        an easier API to use.
 
1036
 
 
1037
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1038
            working tree. Any of these may be ghosts.
 
1039
        """
 
1040
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1041
        trees = []
 
1042
        for revision_id in revision_ids:
 
1043
            try:
 
1044
                revtree = self.branch.repository.revision_tree(revision_id)
 
1045
                # TODO: jam 20070213 KnitVersionedFile raises
 
1046
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1047
                #       given revision_id is not present. Should Repository be
 
1048
                #       catching it and re-raising NoSuchRevision?
 
1049
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1050
                revtree = None
 
1051
            trees.append((revision_id, revtree))
 
1052
        self.current_dirstate()._validate()
 
1053
        self.set_parent_trees(trees,
 
1054
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1055
        self.current_dirstate()._validate()
 
1056
 
 
1057
    @needs_tree_write_lock
 
1058
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1059
        """Set the parents of the working tree.
 
1060
 
 
1061
        :param parents_list: A list of (revision_id, tree) tuples.
 
1062
            If tree is None, then that element is treated as an unreachable
 
1063
            parent tree - i.e. a ghost.
 
1064
        """
 
1065
        dirstate = self.current_dirstate()
 
1066
        dirstate._validate()
 
1067
        if len(parents_list) > 0:
 
1068
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1069
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1070
        real_trees = []
 
1071
        ghosts = []
 
1072
        # convert absent trees to the null tree, which we convert back to
 
1073
        # missing on access.
 
1074
        for rev_id, tree in parents_list:
 
1075
            rev_id = osutils.safe_revision_id(rev_id)
 
1076
            if tree is not None:
 
1077
                real_trees.append((rev_id, tree))
 
1078
            else:
 
1079
                real_trees.append((rev_id,
 
1080
                    self.branch.repository.revision_tree(None)))
 
1081
                ghosts.append(rev_id)
 
1082
        dirstate._validate()
 
1083
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1084
        dirstate._validate()
 
1085
        self._make_dirty(reset_inventory=False)
 
1086
        dirstate._validate()
 
1087
 
 
1088
    def _set_root_id(self, file_id):
 
1089
        """See WorkingTree.set_root_id."""
 
1090
        state = self.current_dirstate()
 
1091
        state.set_path_id('', file_id)
 
1092
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1093
            self._make_dirty(reset_inventory=True)
 
1094
 
 
1095
    @needs_read_lock
 
1096
    def supports_tree_reference(self):
 
1097
        return self._repo_supports_tree_reference
 
1098
 
 
1099
    def unlock(self):
 
1100
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1101
        if self._control_files._lock_count == 1:
 
1102
            # eventually we should do signature checking during read locks for
 
1103
            # dirstate updates.
 
1104
            if self._control_files._lock_mode == 'w':
 
1105
                if self._dirty:
 
1106
                    self.flush()
 
1107
            if self._dirstate is not None:
 
1108
                # This is a no-op if there are no modifications.
 
1109
                self._dirstate.save()
 
1110
                self._dirstate.unlock()
 
1111
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1112
            #       point. Instead, it could check if the header has been
 
1113
            #       modified when it is locked, and if not, it can hang on to
 
1114
            #       the data it has in memory.
 
1115
            self._dirstate = None
 
1116
            self._inventory = None
 
1117
        # reverse order of locking.
 
1118
        try:
 
1119
            return self._control_files.unlock()
 
1120
        finally:
 
1121
            self.branch.unlock()
 
1122
 
 
1123
    @needs_tree_write_lock
 
1124
    def unversion(self, file_ids):
 
1125
        """Remove the file ids in file_ids from the current versioned set.
 
1126
 
 
1127
        When a file_id is unversioned, all of its children are automatically
 
1128
        unversioned.
 
1129
 
 
1130
        :param file_ids: The file ids to stop versioning.
 
1131
        :raises: NoSuchId if any fileid is not currently versioned.
 
1132
        """
 
1133
        if not file_ids:
 
1134
            return
 
1135
        state = self.current_dirstate()
 
1136
        state._read_dirblocks_if_needed()
 
1137
        ids_to_unversion = set()
 
1138
        for file_id in file_ids:
 
1139
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1140
        paths_to_unversion = set()
 
1141
        # sketch:
 
1142
        # check if the root is to be unversioned, if so, assert for now.
 
1143
        # walk the state marking unversioned things as absent.
 
1144
        # if there are any un-unversioned ids at the end, raise
 
1145
        for key, details in state._dirblocks[0][1]:
 
1146
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1147
                key[2] in ids_to_unversion):
 
1148
                # I haven't written the code to unversion / yet - it should be
 
1149
                # supported.
 
1150
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1151
        block_index = 0
 
1152
        while block_index < len(state._dirblocks):
 
1153
            # process one directory at a time.
 
1154
            block = state._dirblocks[block_index]
 
1155
            # first check: is the path one to remove - it or its children
 
1156
            delete_block = False
 
1157
            for path in paths_to_unversion:
 
1158
                if (block[0].startswith(path) and
 
1159
                    (len(block[0]) == len(path) or
 
1160
                     block[0][len(path)] == '/')):
 
1161
                    # this entire block should be deleted - its the block for a
 
1162
                    # path to unversion; or the child of one
 
1163
                    delete_block = True
 
1164
                    break
 
1165
            # TODO: trim paths_to_unversion as we pass by paths
 
1166
            if delete_block:
 
1167
                # this block is to be deleted: process it.
 
1168
                # TODO: we can special case the no-parents case and
 
1169
                # just forget the whole block.
 
1170
                entry_index = 0
 
1171
                while entry_index < len(block[1]):
 
1172
                    # Mark this file id as having been removed
 
1173
                    entry = block[1][entry_index]
 
1174
                    ids_to_unversion.discard(entry[0][2])
 
1175
                    if (entry[1][0][0] == 'a'
 
1176
                        or not state._make_absent(entry)):
 
1177
                        entry_index += 1
 
1178
                # go to the next block. (At the moment we dont delete empty
 
1179
                # dirblocks)
 
1180
                block_index += 1
 
1181
                continue
 
1182
            entry_index = 0
 
1183
            while entry_index < len(block[1]):
 
1184
                entry = block[1][entry_index]
 
1185
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1186
                    # ^ some parent row.
 
1187
                    entry[0][2] not in ids_to_unversion):
 
1188
                    # ^ not an id to unversion
 
1189
                    entry_index += 1
 
1190
                    continue
 
1191
                if entry[1][0][0] == 'd':
 
1192
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1193
                if not state._make_absent(entry):
 
1194
                    entry_index += 1
 
1195
                # we have unversioned this id
 
1196
                ids_to_unversion.remove(entry[0][2])
 
1197
            block_index += 1
 
1198
        if ids_to_unversion:
 
1199
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1200
        self._make_dirty(reset_inventory=False)
 
1201
        # have to change the legacy inventory too.
 
1202
        if self._inventory is not None:
 
1203
            for file_id in file_ids:
 
1204
                self._inventory.remove_recursive_id(file_id)
 
1205
 
 
1206
    @needs_read_lock
 
1207
    def _validate(self):
 
1208
        self._dirstate._validate()
 
1209
 
 
1210
    @needs_tree_write_lock
 
1211
    def _write_inventory(self, inv):
 
1212
        """Write inventory as the current inventory."""
 
1213
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1214
        self.current_dirstate().set_state_from_inventory(inv)
 
1215
        self._make_dirty(reset_inventory=False)
 
1216
        if self._inventory is not None:
 
1217
            self._inventory = inv
 
1218
        self.flush()
 
1219
 
 
1220
 
 
1221
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1222
    """The first consolidated dirstate working tree format.
 
1223
 
 
1224
    This format:
 
1225
        - exists within a metadir controlling .bzr
 
1226
        - includes an explicit version marker for the workingtree control
 
1227
          files, separate from the BzrDir format
 
1228
        - modifies the hash cache format
 
1229
        - is new in bzr 0.15
 
1230
        - uses a LockDir to guard access to it.
 
1231
    """
 
1232
 
 
1233
    upgrade_recommended = False
 
1234
 
 
1235
    def get_format_string(self):
 
1236
        """See WorkingTreeFormat.get_format_string()."""
 
1237
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1238
 
 
1239
    def get_format_description(self):
 
1240
        """See WorkingTreeFormat.get_format_description()."""
 
1241
        return "Working tree format 4"
 
1242
 
 
1243
    def initialize(self, a_bzrdir, revision_id=None):
 
1244
        """See WorkingTreeFormat.initialize().
 
1245
 
 
1246
        :param revision_id: allows creating a working tree at a different
 
1247
        revision than the branch is at.
 
1248
 
 
1249
        These trees get an initial random root id.
 
1250
        """
 
1251
        revision_id = osutils.safe_revision_id(revision_id)
 
1252
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1253
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1254
        transport = a_bzrdir.get_workingtree_transport(self)
 
1255
        control_files = self._open_control_files(a_bzrdir)
 
1256
        control_files.create_lock()
 
1257
        control_files.lock_write()
 
1258
        control_files.put_utf8('format', self.get_format_string())
 
1259
        branch = a_bzrdir.open_branch()
 
1260
        if revision_id is None:
 
1261
            revision_id = branch.last_revision()
 
1262
        local_path = transport.local_abspath('dirstate')
 
1263
        # write out new dirstate (must exist when we create the tree)
 
1264
        state = dirstate.DirState.initialize(local_path)
 
1265
        state.unlock()
 
1266
        del state
 
1267
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1268
                         branch,
 
1269
                         _format=self,
 
1270
                         _bzrdir=a_bzrdir,
 
1271
                         _control_files=control_files)
 
1272
        wt._new_tree()
 
1273
        wt.lock_tree_write()
 
1274
        wt.current_dirstate()._validate()
 
1275
        try:
 
1276
            if revision_id in (None, NULL_REVISION):
 
1277
                wt._set_root_id(generate_ids.gen_root_id())
 
1278
                wt.flush()
 
1279
                wt.current_dirstate()._validate()
 
1280
            wt.set_last_revision(revision_id)
 
1281
            wt.flush()
 
1282
            basis = wt.basis_tree()
 
1283
            basis.lock_read()
 
1284
            # if the basis has a root id we have to use that; otherwise we use
 
1285
            # a new random one
 
1286
            basis_root_id = basis.get_root_id()
 
1287
            if basis_root_id is not None:
 
1288
                wt._set_root_id(basis_root_id)
 
1289
                wt.flush()
 
1290
            transform.build_tree(basis, wt)
 
1291
            basis.unlock()
 
1292
        finally:
 
1293
            control_files.unlock()
 
1294
            wt.unlock()
 
1295
        return wt
 
1296
 
 
1297
    def _open(self, a_bzrdir, control_files):
 
1298
        """Open the tree itself.
 
1299
 
 
1300
        :param a_bzrdir: the dir for the tree.
 
1301
        :param control_files: the control files for the tree.
 
1302
        """
 
1303
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1304
                           branch=a_bzrdir.open_branch(),
 
1305
                           _format=self,
 
1306
                           _bzrdir=a_bzrdir,
 
1307
                           _control_files=control_files)
 
1308
 
 
1309
    def __get_matchingbzrdir(self):
 
1310
        # please test against something that will let us do tree references
 
1311
        return bzrdir.format_registry.make_bzrdir(
 
1312
            'dirstate-with-subtree')
 
1313
 
 
1314
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1315
 
 
1316
 
 
1317
class DirStateRevisionTree(Tree):
 
1318
    """A revision tree pulling the inventory from a dirstate."""
 
1319
 
 
1320
    def __init__(self, dirstate, revision_id, repository):
 
1321
        self._dirstate = dirstate
 
1322
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1323
        self._repository = repository
 
1324
        self._inventory = None
 
1325
        self._locked = 0
 
1326
        self._dirstate_locked = False
 
1327
 
 
1328
    def __repr__(self):
 
1329
        return "<%s of %s in %s>" % \
 
1330
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1331
 
 
1332
    def annotate_iter(self, file_id):
 
1333
        """See Tree.annotate_iter"""
 
1334
        w = self._repository.weave_store.get_weave(file_id,
 
1335
                           self._repository.get_transaction())
 
1336
        return w.annotate_iter(self.inventory[file_id].revision)
 
1337
 
 
1338
    def _comparison_data(self, entry, path):
 
1339
        """See Tree._comparison_data."""
 
1340
        if entry is None:
 
1341
            return None, False, None
 
1342
        # trust the entry as RevisionTree does, but this may not be
 
1343
        # sensible: the entry might not have come from us?
 
1344
        return entry.kind, entry.executable, None
 
1345
 
 
1346
    def _file_size(self, entry, stat_value):
 
1347
        return entry.text_size
 
1348
 
 
1349
    def filter_unversioned_files(self, paths):
 
1350
        """Filter out paths that are not versioned.
 
1351
 
 
1352
        :return: set of paths.
 
1353
        """
 
1354
        pred = self.has_filename
 
1355
        return set((p for p in paths if not pred(p)))
 
1356
 
 
1357
    def get_root_id(self):
 
1358
        return self.path2id('')
 
1359
 
 
1360
    def _get_parent_index(self):
 
1361
        """Return the index in the dirstate referenced by this tree."""
 
1362
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1363
 
 
1364
    def _get_entry(self, file_id=None, path=None):
 
1365
        """Get the dirstate row for file_id or path.
 
1366
 
 
1367
        If either file_id or path is supplied, it is used as the key to lookup.
 
1368
        If both are supplied, the fastest lookup is used, and an error is
 
1369
        raised if they do not both point at the same row.
 
1370
        
 
1371
        :param file_id: An optional unicode file_id to be looked up.
 
1372
        :param path: An optional unicode path to be looked up.
 
1373
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1374
        """
 
1375
        if file_id is None and path is None:
 
1376
            raise errors.BzrError('must supply file_id or path')
 
1377
        file_id = osutils.safe_file_id(file_id)
 
1378
        if path is not None:
 
1379
            path = path.encode('utf8')
 
1380
        parent_index = self._get_parent_index()
 
1381
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1382
 
 
1383
    def _generate_inventory(self):
 
1384
        """Create and set self.inventory from the dirstate object.
 
1385
 
 
1386
        (So this is only called the first time the inventory is requested for
 
1387
        this tree; it then remains in memory until it's out of date.)
 
1388
 
 
1389
        This is relatively expensive: we have to walk the entire dirstate.
 
1390
        """
 
1391
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1392
            'dirstate revision tree'
 
1393
        # separate call for profiling - makes it clear where the costs are.
 
1394
        self._dirstate._read_dirblocks_if_needed()
 
1395
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1396
            'parent %s has disappeared from %s' % (
 
1397
            self._revision_id, self._dirstate.get_parent_ids())
 
1398
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1399
        # This is identical now to the WorkingTree _generate_inventory except
 
1400
        # for the tree index use.
 
1401
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1402
        current_id = root_key[2]
 
1403
        assert current_entry[parent_index][0] == 'd'
 
1404
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1405
        inv.root.revision = current_entry[parent_index][4]
 
1406
        # Turn some things into local variables
 
1407
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1408
        factory = entry_factory
 
1409
        utf8_decode = cache_utf8._utf8_decode
 
1410
        inv_byid = inv._byid
 
1411
        # we could do this straight out of the dirstate; it might be fast
 
1412
        # and should be profiled - RBC 20070216
 
1413
        parent_ies = {'' : inv.root}
 
1414
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1415
            dirname = block[0]
 
1416
            try:
 
1417
                parent_ie = parent_ies[dirname]
 
1418
            except KeyError:
 
1419
                # all the paths in this block are not versioned in this tree
 
1420
                continue
 
1421
            for key, entry in block[1]:
 
1422
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1423
                if minikind in ('a', 'r'): # absent, relocated
 
1424
                    # not this tree
 
1425
                    continue
 
1426
                name = key[1]
 
1427
                name_unicode = utf8_decode(name)[0]
 
1428
                file_id = key[2]
 
1429
                kind = minikind_to_kind[minikind]
 
1430
                inv_entry = factory[kind](file_id, name_unicode,
 
1431
                                          parent_ie.file_id)
 
1432
                inv_entry.revision = revid
 
1433
                if kind == 'file':
 
1434
                    inv_entry.executable = executable
 
1435
                    inv_entry.text_size = size
 
1436
                    inv_entry.text_sha1 = fingerprint
 
1437
                elif kind == 'directory':
 
1438
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1439
                elif kind == 'symlink':
 
1440
                    inv_entry.executable = False
 
1441
                    inv_entry.text_size = None
 
1442
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1443
                elif kind == 'tree-reference':
 
1444
                    inv_entry.reference_revision = fingerprint or None
 
1445
                else:
 
1446
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1447
                            % entry)
 
1448
                # These checks cost us around 40ms on a 55k entry tree
 
1449
                assert file_id not in inv_byid
 
1450
                assert name_unicode not in parent_ie.children
 
1451
                inv_byid[file_id] = inv_entry
 
1452
                parent_ie.children[name_unicode] = inv_entry
 
1453
        self._inventory = inv
 
1454
 
 
1455
    def get_file_mtime(self, file_id, path=None):
 
1456
        """Return the modification time for this record.
 
1457
 
 
1458
        We return the timestamp of the last-changed revision.
 
1459
        """
 
1460
        # Make sure the file exists
 
1461
        entry = self._get_entry(file_id, path=path)
 
1462
        if entry == (None, None): # do we raise?
 
1463
            return None
 
1464
        parent_index = self._get_parent_index()
 
1465
        last_changed_revision = entry[1][parent_index][4]
 
1466
        return self._repository.get_revision(last_changed_revision).timestamp
 
1467
 
 
1468
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1469
        entry = self._get_entry(file_id=file_id, path=path)
 
1470
        parent_index = self._get_parent_index()
 
1471
        parent_details = entry[1][parent_index]
 
1472
        if parent_details[0] == 'f':
 
1473
            return parent_details[1]
 
1474
        return None
 
1475
 
 
1476
    def get_file(self, file_id):
 
1477
        return StringIO(self.get_file_text(file_id))
 
1478
 
 
1479
    def get_file_lines(self, file_id):
 
1480
        ie = self.inventory[file_id]
 
1481
        return self._repository.weave_store.get_weave(file_id,
 
1482
                self._repository.get_transaction()).get_lines(ie.revision)
 
1483
 
 
1484
    def get_file_size(self, file_id):
 
1485
        return self.inventory[file_id].text_size
 
1486
 
 
1487
    def get_file_text(self, file_id):
 
1488
        return ''.join(self.get_file_lines(file_id))
 
1489
 
 
1490
    def get_reference_revision(self, file_id, path=None):
 
1491
        return self.inventory[file_id].reference_revision
 
1492
 
 
1493
    def get_symlink_target(self, file_id):
 
1494
        entry = self._get_entry(file_id=file_id)
 
1495
        parent_index = self._get_parent_index()
 
1496
        if entry[1][parent_index][0] != 'l':
 
1497
            return None
 
1498
        else:
 
1499
            # At present, none of the tree implementations supports non-ascii
 
1500
            # symlink targets. So we will just assume that the dirstate path is
 
1501
            # correct.
 
1502
            return entry[1][parent_index][1]
 
1503
 
 
1504
    def get_revision_id(self):
 
1505
        """Return the revision id for this tree."""
 
1506
        return self._revision_id
 
1507
 
 
1508
    def _get_inventory(self):
 
1509
        if self._inventory is not None:
 
1510
            return self._inventory
 
1511
        self._must_be_locked()
 
1512
        self._generate_inventory()
 
1513
        return self._inventory
 
1514
 
 
1515
    inventory = property(_get_inventory,
 
1516
                         doc="Inventory of this Tree")
 
1517
 
 
1518
    def get_parent_ids(self):
 
1519
        """The parents of a tree in the dirstate are not cached."""
 
1520
        return self._repository.get_revision(self._revision_id).parent_ids
 
1521
 
 
1522
    def has_filename(self, filename):
 
1523
        return bool(self.path2id(filename))
 
1524
 
 
1525
    def kind(self, file_id):
 
1526
        return self.inventory[file_id].kind
 
1527
 
 
1528
    def is_executable(self, file_id, path=None):
 
1529
        ie = self.inventory[file_id]
 
1530
        if ie.kind != "file":
 
1531
            return None
 
1532
        return ie.executable
 
1533
 
 
1534
    def list_files(self, include_root=False):
 
1535
        # We use a standard implementation, because DirStateRevisionTree is
 
1536
        # dealing with one of the parents of the current state
 
1537
        inv = self._get_inventory()
 
1538
        entries = inv.iter_entries()
 
1539
        if self.inventory.root is not None and not include_root:
 
1540
            entries.next()
 
1541
        for path, entry in entries:
 
1542
            yield path, 'V', entry.kind, entry.file_id, entry
 
1543
 
 
1544
    def lock_read(self):
 
1545
        """Lock the tree for a set of operations."""
 
1546
        if not self._locked:
 
1547
            self._repository.lock_read()
 
1548
            if self._dirstate._lock_token is None:
 
1549
                self._dirstate.lock_read()
 
1550
                self._dirstate_locked = True
 
1551
        self._locked += 1
 
1552
 
 
1553
    def _must_be_locked(self):
 
1554
        if not self._locked:
 
1555
            raise errors.ObjectNotLocked(self)
 
1556
 
 
1557
    @needs_read_lock
 
1558
    def path2id(self, path):
 
1559
        """Return the id for path in this tree."""
 
1560
        # lookup by path: faster than splitting and walking the ivnentory.
 
1561
        entry = self._get_entry(path=path)
 
1562
        if entry == (None, None):
 
1563
            return None
 
1564
        return entry[0][2]
 
1565
 
 
1566
    def unlock(self):
 
1567
        """Unlock, freeing any cache memory used during the lock."""
 
1568
        # outside of a lock, the inventory is suspect: release it.
 
1569
        self._locked -=1
 
1570
        if not self._locked:
 
1571
            self._inventory = None
 
1572
            self._locked = 0
 
1573
            if self._dirstate_locked:
 
1574
                self._dirstate.unlock()
 
1575
                self._dirstate_locked = False
 
1576
            self._repository.unlock()
 
1577
 
 
1578
    def walkdirs(self, prefix=""):
 
1579
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1580
        # implementation based on an inventory.  
 
1581
        # This should be cleaned up to use the much faster Dirstate code
 
1582
        # So for now, we just build up the parent inventory, and extract
 
1583
        # it the same way RevisionTree does.
 
1584
        _directory = 'directory'
 
1585
        inv = self._get_inventory()
 
1586
        top_id = inv.path2id(prefix)
 
1587
        if top_id is None:
 
1588
            pending = []
 
1589
        else:
 
1590
            pending = [(prefix, top_id)]
 
1591
        while pending:
 
1592
            dirblock = []
 
1593
            relpath, file_id = pending.pop()
 
1594
            # 0 - relpath, 1- file-id
 
1595
            if relpath:
 
1596
                relroot = relpath + '/'
 
1597
            else:
 
1598
                relroot = ""
 
1599
            # FIXME: stash the node in pending
 
1600
            entry = inv[file_id]
 
1601
            for name, child in entry.sorted_children():
 
1602
                toppath = relroot + name
 
1603
                dirblock.append((toppath, name, child.kind, None,
 
1604
                    child.file_id, child.kind
 
1605
                    ))
 
1606
            yield (relpath, entry.file_id), dirblock
 
1607
            # push the user specified dirs from dirblock
 
1608
            for dir in reversed(dirblock):
 
1609
                if dir[2] == _directory:
 
1610
                    pending.append((dir[0], dir[4]))
 
1611
 
 
1612
 
 
1613
class InterDirStateTree(InterTree):
 
1614
    """Fast path optimiser for changes_from with dirstate trees.
 
1615
    
 
1616
    This is used only when both trees are in the dirstate working file, and 
 
1617
    the source is any parent within the dirstate, and the destination is 
 
1618
    the current working tree of the same dirstate.
 
1619
    """
 
1620
    # this could be generalized to allow comparisons between any trees in the
 
1621
    # dirstate, and possibly between trees stored in different dirstates.
 
1622
 
 
1623
    def __init__(self, source, target):
 
1624
        super(InterDirStateTree, self).__init__(source, target)
 
1625
        if not InterDirStateTree.is_compatible(source, target):
 
1626
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1627
 
 
1628
    @staticmethod
 
1629
    def make_source_parent_tree(source, target):
 
1630
        """Change the source tree into a parent of the target."""
 
1631
        revid = source.commit('record tree')
 
1632
        target.branch.repository.fetch(source.branch.repository, revid)
 
1633
        target.set_parent_ids([revid])
 
1634
        return target.basis_tree(), target
 
1635
 
 
1636
    _matching_from_tree_format = WorkingTreeFormat4()
 
1637
    _matching_to_tree_format = WorkingTreeFormat4()
 
1638
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1639
 
 
1640
    def _iter_changes(self, include_unchanged=False,
 
1641
                      specific_files=None, pb=None, extra_trees=[],
 
1642
                      require_versioned=True, want_unversioned=False):
 
1643
        """Return the changes from source to target.
 
1644
 
 
1645
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1646
            for details.
 
1647
        :param specific_files: An optional list of file paths to restrict the
 
1648
            comparison to. When mapping filenames to ids, all matches in all
 
1649
            trees (including optional extra_trees) are used, and all children of
 
1650
            matched directories are included.
 
1651
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1652
            unchanged entries in the result.
 
1653
        :param extra_trees: An optional list of additional trees to use when
 
1654
            mapping the contents of specific_files (paths) to file_ids.
 
1655
        :param require_versioned: If True, all files in specific_files must be
 
1656
            versioned in one of source, target, extra_trees or
 
1657
            PathsNotVersionedError is raised.
 
1658
        :param want_unversioned: Should unversioned files be returned in the
 
1659
            output. An unversioned file is defined as one with (False, False)
 
1660
            for the versioned pair.
 
1661
        """
 
1662
        utf8_decode_or_none = cache_utf8._utf8_decode_with_None
 
1663
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1664
        # NB: show_status depends on being able to pass in non-versioned files
 
1665
        # and report them as unknown
 
1666
        # TODO: handle extra trees in the dirstate.
 
1667
        # TODO: handle comparisons as an empty tree as a different special
 
1668
        # case? mbp 20070226
 
1669
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1670
            # we can't fast-path these cases (yet)
 
1671
            for f in super(InterDirStateTree, self)._iter_changes(
 
1672
                include_unchanged, specific_files, pb, extra_trees,
 
1673
                require_versioned, want_unversioned=want_unversioned):
 
1674
                yield f
 
1675
            return
 
1676
        parent_ids = self.target.get_parent_ids()
 
1677
        assert (self.source._revision_id in parent_ids), \
 
1678
                "revision {%s} is not stored in {%s}, but %s " \
 
1679
                "can only be used for trees stored in the dirstate" \
 
1680
                % (self.source._revision_id, self.target, self._iter_changes)
 
1681
        target_index = 0
 
1682
        if self.source._revision_id == NULL_REVISION:
 
1683
            source_index = None
 
1684
            indices = (target_index,)
 
1685
        else:
 
1686
            assert (self.source._revision_id in parent_ids), \
 
1687
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1688
                self.source._revision_id, parent_ids)
 
1689
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1690
            indices = (source_index,target_index)
 
1691
        # -- make all specific_files utf8 --
 
1692
        if specific_files:
 
1693
            specific_files_utf8 = set()
 
1694
            for path in specific_files:
 
1695
                specific_files_utf8.add(path.encode('utf8'))
 
1696
            specific_files = specific_files_utf8
 
1697
        else:
 
1698
            specific_files = set([''])
 
1699
        # -- specific_files is now a utf8 path set --
 
1700
        # -- get the state object and prepare it.
 
1701
        state = self.target.current_dirstate()
 
1702
        state._read_dirblocks_if_needed()
 
1703
        def _entries_for_path(path):
 
1704
            """Return a list with all the entries that match path for all ids.
 
1705
            """
 
1706
            dirname, basename = os.path.split(path)
 
1707
            key = (dirname, basename, '')
 
1708
            block_index, present = state._find_block_index_from_key(key)
 
1709
            if not present:
 
1710
                # the block which should contain path is absent.
 
1711
                return []
 
1712
            result = []
 
1713
            block = state._dirblocks[block_index][1]
 
1714
            entry_index, _ = state._find_entry_index(key, block)
 
1715
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1716
            while (entry_index < len(block) and
 
1717
                block[entry_index][0][0:2] == key[0:2]):
 
1718
                result.append(block[entry_index])
 
1719
                entry_index += 1
 
1720
            return result
 
1721
        if require_versioned:
 
1722
            # -- check all supplied paths are versioned in a search tree. --
 
1723
            all_versioned = True
 
1724
            for path in specific_files:
 
1725
                path_entries = _entries_for_path(path)
 
1726
                if not path_entries:
 
1727
                    # this specified path is not present at all: error
 
1728
                    all_versioned = False
 
1729
                    break
 
1730
                found_versioned = False
 
1731
                # for each id at this path
 
1732
                for entry in path_entries:
 
1733
                    # for each tree.
 
1734
                    for index in indices:
 
1735
                        if entry[1][index][0] != 'a': # absent
 
1736
                            found_versioned = True
 
1737
                            # all good: found a versioned cell
 
1738
                            break
 
1739
                if not found_versioned:
 
1740
                    # none of the indexes was not 'absent' at all ids for this
 
1741
                    # path.
 
1742
                    all_versioned = False
 
1743
                    break
 
1744
            if not all_versioned:
 
1745
                raise errors.PathsNotVersionedError(specific_files)
 
1746
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1747
        search_specific_files = set()
 
1748
        for path in specific_files:
 
1749
            other_specific_files = specific_files.difference(set([path]))
 
1750
            if not osutils.is_inside_any(other_specific_files, path):
 
1751
                # this is a top level path, we must check it.
 
1752
                search_specific_files.add(path)
 
1753
        # sketch: 
 
1754
        # compare source_index and target_index at or under each element of search_specific_files.
 
1755
        # follow the following comparison table. Note that we only want to do diff operations when
 
1756
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1757
        # for the target.
 
1758
        # cases:
 
1759
        # 
 
1760
        # Source | Target | disk | action
 
1761
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1762
        #        |        |      | diff check on source-target
 
1763
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1764
        #        |        |      | ???
 
1765
        #   r    |  a     |      | add source to search
 
1766
        #   r    |  a     |  a   | 
 
1767
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1768
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1769
        #   a    | fdlt   |      | add new id
 
1770
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1771
        #   a    |  a     |      | not present in either tree, skip
 
1772
        #   a    |  a     |  a   | not present in any tree, skip
 
1773
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1774
        #        |        |      | may not be selected by the users list of paths.
 
1775
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1776
        #        |        |      | may not be selected by the users list of paths.
 
1777
        #  fdlt  | fdlt   |      | content in both: diff them
 
1778
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1779
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1780
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1781
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1782
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1783
        #        |        |      | this id at the other path.
 
1784
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1785
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1786
        #        |        |      | this id at the other path.
 
1787
 
 
1788
        # for all search_indexs in each path at or under each element of
 
1789
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1790
        # relocated path as one to search if its not searched already. If the
 
1791
        # detail is not relocated, add the id.
 
1792
        searched_specific_files = set()
 
1793
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1794
        # Using a list so that we can access the values and change them in
 
1795
        # nested scope. Each one is [path, file_id, entry]
 
1796
        last_source_parent = [None, None, None]
 
1797
        last_target_parent = [None, None, None]
 
1798
 
 
1799
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1800
 
 
1801
        def _process_entry(entry, path_info):
 
1802
            """Compare an entry and real disk to generate delta information.
 
1803
 
 
1804
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1805
                the path of entry. If None, then the path is considered absent.
 
1806
                (Perhaps we should pass in a concrete entry for this ?)
 
1807
                Basename is returned as a utf8 string because we expect this
 
1808
                tuple will be ignored, and don't want to take the time to
 
1809
                decode.
 
1810
            """
 
1811
            if source_index is None:
 
1812
                source_details = NULL_PARENT_DETAILS
 
1813
            else:
 
1814
                source_details = entry[1][source_index]
 
1815
            target_details = entry[1][target_index]
 
1816
            target_minikind = target_details[0]
 
1817
            if path_info is not None and target_minikind in 'fdlt':
 
1818
                assert target_index == 0
 
1819
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1820
                                                  stat_value=path_info[3])
 
1821
                # The entry may have been modified by update_entry
 
1822
                target_details = entry[1][target_index]
 
1823
                target_minikind = target_details[0]
 
1824
            else:
 
1825
                link_or_sha1 = None
 
1826
            source_minikind = source_details[0]
 
1827
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1828
                # claimed content in both: diff
 
1829
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1830
                #        |        |      | diff check on source-target
 
1831
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1832
                #        |        |      | ???
 
1833
                if source_minikind in 'r':
 
1834
                    # add the source to the search path to find any children it
 
1835
                    # has.  TODO ? : only add if it is a container ?
 
1836
                    if not osutils.is_inside_any(searched_specific_files,
 
1837
                                                 source_details[1]):
 
1838
                        search_specific_files.add(source_details[1])
 
1839
                    # generate the old path; this is needed for stating later
 
1840
                    # as well.
 
1841
                    old_path = source_details[1]
 
1842
                    old_dirname, old_basename = os.path.split(old_path)
 
1843
                    path = pathjoin(entry[0][0], entry[0][1])
 
1844
                    old_entry = state._get_entry(source_index,
 
1845
                                                 path_utf8=old_path)
 
1846
                    # update the source details variable to be the real
 
1847
                    # location.
 
1848
                    source_details = old_entry[1][source_index]
 
1849
                    source_minikind = source_details[0]
 
1850
                else:
 
1851
                    old_dirname = entry[0][0]
 
1852
                    old_basename = entry[0][1]
 
1853
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1854
                if path_info is None:
 
1855
                    # the file is missing on disk, show as removed.
 
1856
                    content_change = True
 
1857
                    target_kind = None
 
1858
                    target_exec = False
 
1859
                else:
 
1860
                    # source and target are both versioned and disk file is present.
 
1861
                    target_kind = path_info[2]
 
1862
                    if target_kind == 'directory':
 
1863
                        if source_minikind != 'd':
 
1864
                            content_change = True
 
1865
                        else:
 
1866
                            # directories have no fingerprint
 
1867
                            content_change = False
 
1868
                        target_exec = False
 
1869
                    elif target_kind == 'file':
 
1870
                        if source_minikind != 'f':
 
1871
                            content_change = True
 
1872
                        else:
 
1873
                            # We could check the size, but we already have the
 
1874
                            # sha1 hash.
 
1875
                            content_change = (link_or_sha1 != source_details[1])
 
1876
                        # Target details is updated at update_entry time
 
1877
                        if use_filesystem_for_exec:
 
1878
                            # We don't need S_ISREG here, because we are sure
 
1879
                            # we are dealing with a file.
 
1880
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1881
                        else:
 
1882
                            target_exec = target_details[3]
 
1883
                    elif target_kind == 'symlink':
 
1884
                        if source_minikind != 'l':
 
1885
                            content_change = True
 
1886
                        else:
 
1887
                            content_change = (link_or_sha1 != source_details[1])
 
1888
                        target_exec = False
 
1889
                    elif target_kind == 'tree-reference':
 
1890
                        if source_minikind != 't':
 
1891
                            content_change = True
 
1892
                        else:
 
1893
                            content_change = False
 
1894
                        target_exec = False
 
1895
                    else:
 
1896
                        raise Exception, "unknown kind %s" % path_info[2]
 
1897
                # parent id is the entry for the path in the target tree
 
1898
                if old_dirname == last_source_parent[0]:
 
1899
                    source_parent_id = last_source_parent[1]
 
1900
                else:
 
1901
                    source_parent_entry = state._get_entry(source_index,
 
1902
                                                           path_utf8=old_dirname)
 
1903
                    source_parent_id = source_parent_entry[0][2]
 
1904
                    if source_parent_id == entry[0][2]:
 
1905
                        # This is the root, so the parent is None
 
1906
                        source_parent_id = None
 
1907
                    else:
 
1908
                        last_source_parent[0] = old_dirname
 
1909
                        last_source_parent[1] = source_parent_id
 
1910
                        last_source_parent[2] = source_parent_entry
 
1911
                new_dirname = entry[0][0]
 
1912
                if new_dirname == last_target_parent[0]:
 
1913
                    target_parent_id = last_target_parent[1]
 
1914
                else:
 
1915
                    # TODO: We don't always need to do the lookup, because the
 
1916
                    #       parent entry will be the same as the source entry.
 
1917
                    target_parent_entry = state._get_entry(target_index,
 
1918
                                                           path_utf8=new_dirname)
 
1919
                    target_parent_id = target_parent_entry[0][2]
 
1920
                    if target_parent_id == entry[0][2]:
 
1921
                        # This is the root, so the parent is None
 
1922
                        target_parent_id = None
 
1923
                    else:
 
1924
                        last_target_parent[0] = new_dirname
 
1925
                        last_target_parent[1] = target_parent_id
 
1926
                        last_target_parent[2] = target_parent_entry
 
1927
 
 
1928
                source_exec = source_details[3]
 
1929
                return ((entry[0][2], (old_path, path), content_change,
 
1930
                        (True, True),
 
1931
                        (source_parent_id, target_parent_id),
 
1932
                        (old_basename, entry[0][1]),
 
1933
                        (_minikind_to_kind[source_minikind], target_kind),
 
1934
                        (source_exec, target_exec)),)
 
1935
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1936
                # looks like a new file
 
1937
                if path_info is not None:
 
1938
                    path = pathjoin(entry[0][0], entry[0][1])
 
1939
                    # parent id is the entry for the path in the target tree
 
1940
                    # TODO: these are the same for an entire directory: cache em.
 
1941
                    parent_id = state._get_entry(target_index,
 
1942
                                                 path_utf8=entry[0][0])[0][2]
 
1943
                    if parent_id == entry[0][2]:
 
1944
                        parent_id = None
 
1945
                    if use_filesystem_for_exec:
 
1946
                        # We need S_ISREG here, because we aren't sure if this
 
1947
                        # is a file or not.
 
1948
                        target_exec = bool(
 
1949
                            stat.S_ISREG(path_info[3].st_mode)
 
1950
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1951
                    else:
 
1952
                        target_exec = target_details[3]
 
1953
                    return ((entry[0][2], (None, path), True,
 
1954
                            (False, True),
 
1955
                            (None, parent_id),
 
1956
                            (None, entry[0][1]),
 
1957
                            (None, path_info[2]),
 
1958
                            (None, target_exec)),)
 
1959
                else:
 
1960
                    # but its not on disk: we deliberately treat this as just
 
1961
                    # never-present. (Why ?! - RBC 20070224)
 
1962
                    pass
 
1963
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1964
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1965
                # if its still on disk, *and* theres no other entry at this
 
1966
                # path [we dont know this in this routine at the moment -
 
1967
                # perhaps we should change this - then it would be an unknown.
 
1968
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1969
                # parent id is the entry for the path in the target tree
 
1970
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1971
                if parent_id == entry[0][2]:
 
1972
                    parent_id = None
 
1973
                return ((entry[0][2], (old_path, None), True,
 
1974
                        (True, False),
 
1975
                        (parent_id, None),
 
1976
                        (entry[0][1], None),
 
1977
                        (_minikind_to_kind[source_minikind], None),
 
1978
                        (source_details[3], None)),)
 
1979
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1980
                # a rename; could be a true rename, or a rename inherited from
 
1981
                # a renamed parent. TODO: handle this efficiently. Its not
 
1982
                # common case to rename dirs though, so a correct but slow
 
1983
                # implementation will do.
 
1984
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1985
                    search_specific_files.add(target_details[1])
 
1986
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
1987
                # neither of the selected trees contain this file,
 
1988
                # so skip over it. This is not currently directly tested, but
 
1989
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1990
                pass
 
1991
            else:
 
1992
                raise AssertionError("don't know how to compare "
 
1993
                    "source_minikind=%r, target_minikind=%r"
 
1994
                    % (source_minikind, target_minikind))
 
1995
                ## import pdb;pdb.set_trace()
 
1996
            return ()
 
1997
 
 
1998
        while search_specific_files:
 
1999
            # TODO: the pending list should be lexically sorted?  the
 
2000
            # interface doesn't require it.
 
2001
            current_root = search_specific_files.pop()
 
2002
            current_root_unicode = current_root.decode('utf8')
 
2003
            searched_specific_files.add(current_root)
 
2004
            # process the entries for this containing directory: the rest will be
 
2005
            # found by their parents recursively.
 
2006
            root_entries = _entries_for_path(current_root)
 
2007
            root_abspath = self.target.abspath(current_root_unicode)
 
2008
            try:
 
2009
                root_stat = os.lstat(root_abspath)
 
2010
            except OSError, e:
 
2011
                if e.errno == errno.ENOENT:
 
2012
                    # the path does not exist: let _process_entry know that.
 
2013
                    root_dir_info = None
 
2014
                else:
 
2015
                    # some other random error: hand it up.
 
2016
                    raise
 
2017
            else:
 
2018
                root_dir_info = ('', current_root,
 
2019
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2020
                    root_abspath)
 
2021
                if root_dir_info[2] == 'directory':
 
2022
                    if self.target._directory_is_tree_reference(
 
2023
                        current_root.decode('utf8')):
 
2024
                        root_dir_info = root_dir_info[:2] + \
 
2025
                            ('tree-reference',) + root_dir_info[3:]
 
2026
 
 
2027
            if not root_entries and not root_dir_info:
 
2028
                # this specified path is not present at all, skip it.
 
2029
                continue
 
2030
            path_handled = False
 
2031
            for entry in root_entries:
 
2032
                for result in _process_entry(entry, root_dir_info):
 
2033
                    # this check should probably be outside the loop: one
 
2034
                    # 'iterate two trees' api, and then _iter_changes filters
 
2035
                    # unchanged pairs. - RBC 20070226
 
2036
                    path_handled = True
 
2037
                    if (include_unchanged
 
2038
                        or result[2]                    # content change
 
2039
                        or result[3][0] != result[3][1] # versioned status
 
2040
                        or result[4][0] != result[4][1] # parent id
 
2041
                        or result[5][0] != result[5][1] # name
 
2042
                        or result[6][0] != result[6][1] # kind
 
2043
                        or result[7][0] != result[7][1] # executable
 
2044
                        ):
 
2045
                        yield (result[0],
 
2046
                               (utf8_decode_or_none(result[1][0]),
 
2047
                                utf8_decode_or_none(result[1][1])),
 
2048
                               result[2],
 
2049
                               result[3],
 
2050
                               result[4],
 
2051
                               (utf8_decode_or_none(result[5][0]),
 
2052
                                utf8_decode_or_none(result[5][1])),
 
2053
                               result[6],
 
2054
                               result[7],
 
2055
                              )
 
2056
            if want_unversioned and not path_handled and root_dir_info:
 
2057
                new_executable = bool(
 
2058
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2059
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2060
                yield (None,
 
2061
                       (None, current_root_unicode),
 
2062
                       True,
 
2063
                       (False, False),
 
2064
                       (None, None),
 
2065
                       (None, splitpath(current_root_unicode)[-1]),
 
2066
                       (None, root_dir_info[2]),
 
2067
                       (None, new_executable)
 
2068
                      )
 
2069
            initial_key = (current_root, '', '')
 
2070
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2071
            if block_index == 0:
 
2072
                # we have processed the total root already, but because the
 
2073
                # initial key matched it we should skip it here.
 
2074
                block_index +=1
 
2075
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2076
                current_dir_info = None
 
2077
            else:
 
2078
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2079
                try:
 
2080
                    current_dir_info = dir_iterator.next()
 
2081
                except OSError, e:
 
2082
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2083
                    # python 2.5 has e.errno == EINVAL,
 
2084
                    #            and e.winerror == ERROR_DIRECTORY
 
2085
                    e_winerror = getattr(e, 'winerror', None)
 
2086
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2087
                    # there may be directories in the inventory even though
 
2088
                    # this path is not a file on disk: so mark it as end of
 
2089
                    # iterator
 
2090
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2091
                        current_dir_info = None
 
2092
                    elif (sys.platform == 'win32'
 
2093
                          and (e.errno in win_errors
 
2094
                               or e_winerror in win_errors)):
 
2095
                        current_dir_info = None
 
2096
                    else:
 
2097
                        raise
 
2098
                else:
 
2099
                    if current_dir_info[0][0] == '':
 
2100
                        # remove .bzr from iteration
 
2101
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2102
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2103
                        del current_dir_info[1][bzr_index]
 
2104
            # walk until both the directory listing and the versioned metadata
 
2105
            # are exhausted. 
 
2106
            if (block_index < len(state._dirblocks) and
 
2107
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2108
                current_block = state._dirblocks[block_index]
 
2109
            else:
 
2110
                current_block = None
 
2111
            while (current_dir_info is not None or
 
2112
                   current_block is not None):
 
2113
                if (current_dir_info and current_block
 
2114
                    and current_dir_info[0][0] != current_block[0]):
 
2115
                    if current_dir_info[0][0] < current_block[0] :
 
2116
                        # filesystem data refers to paths not covered by the dirblock.
 
2117
                        # this has two possibilities:
 
2118
                        # A) it is versioned but empty, so there is no block for it
 
2119
                        # B) it is not versioned.
 
2120
                        # in either case it was processed by the containing directories walk:
 
2121
                        # if it is root/foo, when we walked root we emitted it,
 
2122
                        # or if we ere given root/foo to walk specifically, we
 
2123
                        # emitted it when checking the walk-root entries
 
2124
                        # advance the iterator and loop - we dont need to emit it.
 
2125
                        try:
 
2126
                            current_dir_info = dir_iterator.next()
 
2127
                        except StopIteration:
 
2128
                            current_dir_info = None
 
2129
                    else:
 
2130
                        # We have a dirblock entry for this location, but there
 
2131
                        # is no filesystem path for this. This is most likely
 
2132
                        # because a directory was removed from the disk.
 
2133
                        # We don't have to report the missing directory,
 
2134
                        # because that should have already been handled, but we
 
2135
                        # need to handle all of the files that are contained
 
2136
                        # within.
 
2137
                        for current_entry in current_block[1]:
 
2138
                            # entry referring to file not present on disk.
 
2139
                            # advance the entry only, after processing.
 
2140
                            for result in _process_entry(current_entry, None):
 
2141
                                # this check should probably be outside the loop: one
 
2142
                                # 'iterate two trees' api, and then _iter_changes filters
 
2143
                                # unchanged pairs. - RBC 20070226
 
2144
                                if (include_unchanged
 
2145
                                    or result[2]                    # content change
 
2146
                                    or result[3][0] != result[3][1] # versioned status
 
2147
                                    or result[4][0] != result[4][1] # parent id
 
2148
                                    or result[5][0] != result[5][1] # name
 
2149
                                    or result[6][0] != result[6][1] # kind
 
2150
                                    or result[7][0] != result[7][1] # executable
 
2151
                                    ):
 
2152
                                    yield (result[0],
 
2153
                                           (utf8_decode_or_none(result[1][0]),
 
2154
                                            utf8_decode_or_none(result[1][1])),
 
2155
                                           result[2],
 
2156
                                           result[3],
 
2157
                                           result[4],
 
2158
                                           (utf8_decode_or_none(result[5][0]),
 
2159
                                            utf8_decode_or_none(result[5][1])),
 
2160
                                           result[6],
 
2161
                                           result[7],
 
2162
                                          )
 
2163
                        block_index +=1
 
2164
                        if (block_index < len(state._dirblocks) and
 
2165
                            osutils.is_inside(current_root,
 
2166
                                              state._dirblocks[block_index][0])):
 
2167
                            current_block = state._dirblocks[block_index]
 
2168
                        else:
 
2169
                            current_block = None
 
2170
                    continue
 
2171
                entry_index = 0
 
2172
                if current_block and entry_index < len(current_block[1]):
 
2173
                    current_entry = current_block[1][entry_index]
 
2174
                else:
 
2175
                    current_entry = None
 
2176
                advance_entry = True
 
2177
                path_index = 0
 
2178
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2179
                    current_path_info = current_dir_info[1][path_index]
 
2180
                    if current_path_info[2] == 'directory':
 
2181
                        if self.target._directory_is_tree_reference(
 
2182
                            current_path_info[0].decode('utf8')):
 
2183
                            current_path_info = current_path_info[:2] + \
 
2184
                                ('tree-reference',) + current_path_info[3:]
 
2185
                else:
 
2186
                    current_path_info = None
 
2187
                advance_path = True
 
2188
                path_handled = False
 
2189
                while (current_entry is not None or
 
2190
                    current_path_info is not None):
 
2191
                    if current_entry is None:
 
2192
                        # the check for path_handled when the path is adnvaced
 
2193
                        # will yield this path if needed.
 
2194
                        pass
 
2195
                    elif current_path_info is None:
 
2196
                        # no path is fine: the per entry code will handle it.
 
2197
                        for result in _process_entry(current_entry, current_path_info):
 
2198
                            # this check should probably be outside the loop: one
 
2199
                            # 'iterate two trees' api, and then _iter_changes filters
 
2200
                            # unchanged pairs. - RBC 20070226
 
2201
                            if (include_unchanged
 
2202
                                or result[2]                    # content change
 
2203
                                or result[3][0] != result[3][1] # versioned status
 
2204
                                or result[4][0] != result[4][1] # parent id
 
2205
                                or result[5][0] != result[5][1] # name
 
2206
                                or result[6][0] != result[6][1] # kind
 
2207
                                or result[7][0] != result[7][1] # executable
 
2208
                                ):
 
2209
                                yield (result[0],
 
2210
                                       (utf8_decode_or_none(result[1][0]),
 
2211
                                        utf8_decode_or_none(result[1][1])),
 
2212
                                       result[2],
 
2213
                                       result[3],
 
2214
                                       result[4],
 
2215
                                       (utf8_decode_or_none(result[5][0]),
 
2216
                                        utf8_decode_or_none(result[5][1])),
 
2217
                                       result[6],
 
2218
                                       result[7],
 
2219
                                      )
 
2220
                    elif current_entry[0][1] != current_path_info[1]:
 
2221
                        if current_path_info[1] < current_entry[0][1]:
 
2222
                            # extra file on disk: pass for now, but only
 
2223
                            # increment the path, not the entry
 
2224
                            advance_entry = False
 
2225
                        else:
 
2226
                            # entry referring to file not present on disk.
 
2227
                            # advance the entry only, after processing.
 
2228
                            for result in _process_entry(current_entry, None):
 
2229
                                # this check should probably be outside the loop: one
 
2230
                                # 'iterate two trees' api, and then _iter_changes filters
 
2231
                                # unchanged pairs. - RBC 20070226
 
2232
                                path_handled = True
 
2233
                                if (include_unchanged
 
2234
                                    or result[2]                    # content change
 
2235
                                    or result[3][0] != result[3][1] # versioned status
 
2236
                                    or result[4][0] != result[4][1] # parent id
 
2237
                                    or result[5][0] != result[5][1] # name
 
2238
                                    or result[6][0] != result[6][1] # kind
 
2239
                                    or result[7][0] != result[7][1] # executable
 
2240
                                    ):
 
2241
                                    yield (result[0],
 
2242
                                           (utf8_decode_or_none(result[1][0]),
 
2243
                                            utf8_decode_or_none(result[1][1])),
 
2244
                                           result[2],
 
2245
                                           result[3],
 
2246
                                           result[4],
 
2247
                                           (utf8_decode_or_none(result[5][0]),
 
2248
                                            utf8_decode_or_none(result[5][1])),
 
2249
                                           result[6],
 
2250
                                           result[7],
 
2251
                                          )
 
2252
                            advance_path = False
 
2253
                    else:
 
2254
                        for result in _process_entry(current_entry, current_path_info):
 
2255
                            # this check should probably be outside the loop: one
 
2256
                            # 'iterate two trees' api, and then _iter_changes filters
 
2257
                            # unchanged pairs. - RBC 20070226
 
2258
                            path_handled = True
 
2259
                            if (include_unchanged
 
2260
                                or result[2]                    # content change
 
2261
                                or result[3][0] != result[3][1] # versioned status
 
2262
                                or result[4][0] != result[4][1] # parent id
 
2263
                                or result[5][0] != result[5][1] # name
 
2264
                                or result[6][0] != result[6][1] # kind
 
2265
                                or result[7][0] != result[7][1] # executable
 
2266
                                ):
 
2267
                                yield (result[0],
 
2268
                                       (utf8_decode_or_none(result[1][0]),
 
2269
                                        utf8_decode_or_none(result[1][1])),
 
2270
                                       result[2],
 
2271
                                       result[3],
 
2272
                                       result[4],
 
2273
                                       (utf8_decode_or_none(result[5][0]),
 
2274
                                        utf8_decode_or_none(result[5][1])),
 
2275
                                       result[6],
 
2276
                                       result[7],
 
2277
                                      )
 
2278
                    if advance_entry and current_entry is not None:
 
2279
                        entry_index += 1
 
2280
                        if entry_index < len(current_block[1]):
 
2281
                            current_entry = current_block[1][entry_index]
 
2282
                        else:
 
2283
                            current_entry = None
 
2284
                    else:
 
2285
                        advance_entry = True # reset the advance flaga
 
2286
                    if advance_path and current_path_info is not None:
 
2287
                        if not path_handled:
 
2288
                            # unversioned in all regards
 
2289
                            if want_unversioned:
 
2290
                                new_executable = bool(
 
2291
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2292
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2293
                                if want_unversioned:
 
2294
                                    yield (None,
 
2295
                                        (None, utf8_decode_or_none(current_path_info[0])),
 
2296
                                        True,
 
2297
                                        (False, False),
 
2298
                                        (None, None),
 
2299
                                        (None, utf8_decode_or_none(current_path_info[1])),
 
2300
                                        (None, current_path_info[2]),
 
2301
                                        (None, new_executable))
 
2302
                            # dont descend into this unversioned path if it is
 
2303
                            # a dir
 
2304
                            if current_path_info[2] in ('directory'):
 
2305
                                del current_dir_info[1][path_index]
 
2306
                                path_index -= 1
 
2307
                        # dont descend the disk iterator into any tree 
 
2308
                        # paths.
 
2309
                        if current_path_info[2] == 'tree-reference':
 
2310
                            del current_dir_info[1][path_index]
 
2311
                            path_index -= 1
 
2312
                        path_index += 1
 
2313
                        if path_index < len(current_dir_info[1]):
 
2314
                            current_path_info = current_dir_info[1][path_index]
 
2315
                            if current_path_info[2] == 'directory':
 
2316
                                if self.target._directory_is_tree_reference(
 
2317
                                    current_path_info[0].decode('utf8')):
 
2318
                                    current_path_info = current_path_info[:2] + \
 
2319
                                        ('tree-reference',) + current_path_info[3:]
 
2320
                        else:
 
2321
                            current_path_info = None
 
2322
                        path_handled = False
 
2323
                    else:
 
2324
                        advance_path = True # reset the advance flagg.
 
2325
                if current_block is not None:
 
2326
                    block_index += 1
 
2327
                    if (block_index < len(state._dirblocks) and
 
2328
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2329
                        current_block = state._dirblocks[block_index]
 
2330
                    else:
 
2331
                        current_block = None
 
2332
                if current_dir_info is not None:
 
2333
                    try:
 
2334
                        current_dir_info = dir_iterator.next()
 
2335
                    except StopIteration:
 
2336
                        current_dir_info = None
 
2337
 
 
2338
 
 
2339
    @staticmethod
 
2340
    def is_compatible(source, target):
 
2341
        # the target must be a dirstate working tree
 
2342
        if not isinstance(target, WorkingTree4):
 
2343
            return False
 
2344
        # the source must be a revtreee or dirstate rev tree.
 
2345
        if not isinstance(source,
 
2346
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2347
            return False
 
2348
        # the source revid must be in the target dirstate
 
2349
        if not (source._revision_id == NULL_REVISION or
 
2350
            source._revision_id in target.get_parent_ids()):
 
2351
            # TODO: what about ghosts? it may well need to 
 
2352
            # check for them explicitly.
 
2353
            return False
 
2354
        return True
 
2355
 
 
2356
InterTree.register_optimiser(InterDirStateTree)
 
2357
 
 
2358
 
 
2359
class Converter3to4(object):
 
2360
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2361
 
 
2362
    def __init__(self):
 
2363
        self.target_format = WorkingTreeFormat4()
 
2364
 
 
2365
    def convert(self, tree):
 
2366
        # lock the control files not the tree, so that we dont get tree
 
2367
        # on-unlock behaviours, and so that noone else diddles with the 
 
2368
        # tree during upgrade.
 
2369
        tree._control_files.lock_write()
 
2370
        try:
 
2371
            tree.read_working_inventory()
 
2372
            self.create_dirstate_data(tree)
 
2373
            self.update_format(tree)
 
2374
            self.remove_xml_files(tree)
 
2375
        finally:
 
2376
            tree._control_files.unlock()
 
2377
 
 
2378
    def create_dirstate_data(self, tree):
 
2379
        """Create the dirstate based data for tree."""
 
2380
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2381
            ).local_abspath('dirstate')
 
2382
        state = dirstate.DirState.from_tree(tree, local_path)
 
2383
        state.save()
 
2384
        state.unlock()
 
2385
 
 
2386
    def remove_xml_files(self, tree):
 
2387
        """Remove the oldformat 3 data."""
 
2388
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2389
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2390
            'pending-merges', 'stat-cache']:
 
2391
            try:
 
2392
                transport.delete(path)
 
2393
            except errors.NoSuchFile:
 
2394
                # some files are optional - just deal.
 
2395
                pass
 
2396
 
 
2397
    def update_format(self, tree):
 
2398
        """Change the format marker."""
 
2399
        tree._control_files.put_utf8('format',
 
2400
            self.target_format.get_format_string())