~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: John Arbash Meinel
  • Date: 2007-04-12 21:18:29 UTC
  • mto: This revision was merged to the branch mainline in revision 2415.
  • Revision ID: john@arbash-meinel.com-20070412211829-dduwktk8in0e6zpl
Clean up test ordering

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