/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: Robert Collins
  • Date: 2007-03-06 21:45:51 UTC
  • mto: (2321.1.1 integration)
  • mto: This revision was merged to the branch mainline in revision 2322.
  • Revision ID: robertc@robertcollins.net-20070306214551-ayvx2ziz2hgbkqm0
Add test showing reverts UI working as Aaron intended it to.

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