/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2007-03-13 12:01:34 UTC
  • mfrom: (2318.2.4 bzr.kg)
  • Revision ID: pqm@pqm.ubuntu.com-20070313120134-ddd44b75e5a2639c
Add support for status --versioned

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