/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2007-03-19 07:26:06 UTC
  • mfrom: (1551.13.3 Aaron's mergeable stuff)
  • Revision ID: pqm@pqm.ubuntu.com-20070319072606-0b45228c7ad6e0ae
Remove the merge-revert alias

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
 
47
    dirstate,
 
48
    errors,
 
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revisiontree,
 
56
    textui,
 
57
    transform,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
 
61
    )
 
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
 
64
import bzrlib.ui
 
65
""")
 
66
 
 
67
from bzrlib import symbol_versioning
 
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
 
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
 
73
from bzrlib.mutabletree import needs_tree_write_lock
 
74
from bzrlib.osutils import (
 
75
    file_kind,
 
76
    isdir,
 
77
    normpath,
 
78
    pathjoin,
 
79
    rand_chars,
 
80
    realpath,
 
81
    safe_unicode,
 
82
    splitpath,
 
83
    )
 
84
from bzrlib.trace import mutter, note
 
85
from bzrlib.transport.local import LocalTransport
 
86
from bzrlib.tree import InterTree
 
87
from bzrlib.progress import DummyProgress, ProgressPhase
 
88
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
89
from bzrlib.rio import RioReader, rio_file, Stanza
 
90
from bzrlib.symbol_versioning import (deprecated_passed,
 
91
        deprecated_method,
 
92
        deprecated_function,
 
93
        DEPRECATED_PARAMETER,
 
94
        )
 
95
from bzrlib.tree import Tree
 
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
97
 
 
98
 
 
99
class WorkingTree4(WorkingTree3):
 
100
    """This is the Format 4 working tree.
 
101
 
 
102
    This differs from WorkingTree3 by:
 
103
     - Having a consolidated internal dirstate, stored in a
 
104
       randomly-accessible sorted file on disk.
 
105
     - Not having a regular inventory attribute.  One can be synthesized 
 
106
       on demand but this is expensive and should be avoided.
 
107
 
 
108
    This is new in bzr 0.15.
 
109
    """
 
110
 
 
111
    def __init__(self, basedir,
 
112
                 branch,
 
113
                 _control_files=None,
 
114
                 _format=None,
 
115
                 _bzrdir=None):
 
116
        """Construct a WorkingTree for basedir.
 
117
 
 
118
        If the branch is not supplied, it is opened automatically.
 
119
        If the branch is supplied, it must be the branch for this basedir.
 
120
        (branch.base is not cross checked, because for remote branches that
 
121
        would be meaningless).
 
122
        """
 
123
        self._format = _format
 
124
        self.bzrdir = _bzrdir
 
125
        from bzrlib.trace import note, mutter
 
126
        assert isinstance(basedir, basestring), \
 
127
            "base directory %r is not a string" % basedir
 
128
        basedir = safe_unicode(basedir)
 
129
        mutter("opening working tree %r", basedir)
 
130
        self._branch = branch
 
131
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
132
            "branch %r is not a Branch" % self.branch
 
133
        self.basedir = realpath(basedir)
 
134
        # if branch is at our basedir and is a format 6 or less
 
135
        # assume all other formats have their own control files.
 
136
        assert isinstance(_control_files, LockableFiles), \
 
137
            "_control_files must be a LockableFiles, not %r" % _control_files
 
138
        self._control_files = _control_files
 
139
        self._dirty = None
 
140
        #-------------
 
141
        # during a read or write lock these objects are set, and are
 
142
        # None the rest of the time.
 
143
        self._dirstate = None
 
144
        self._inventory = None
 
145
        #-------------
 
146
 
 
147
    @needs_tree_write_lock
 
148
    def _add(self, files, ids, kinds):
 
149
        """See MutableTree._add."""
 
150
        state = self.current_dirstate()
 
151
        for f, file_id, kind in zip(files, ids, kinds):
 
152
            f = f.strip('/')
 
153
            assert '//' not in f
 
154
            assert '..' not in f
 
155
            if self.path2id(f):
 
156
                # special case tree root handling.
 
157
                if f == '' and self.path2id(f) == ROOT_ID:
 
158
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
159
                continue
 
160
            if file_id is None:
 
161
                file_id = generate_ids.gen_file_id(f)
 
162
            # deliberately add the file with no cached stat or sha1
 
163
            # - on the first access it will be gathered, and we can
 
164
            # always change this once tests are all passing.
 
165
            state.add(f, file_id, kind, None, '')
 
166
        self._make_dirty(reset_inventory=True)
 
167
 
 
168
    def _make_dirty(self, reset_inventory):
 
169
        """Make the tree state dirty.
 
170
 
 
171
        :param reset_inventory: True if the cached inventory should be removed
 
172
            (presuming there is one).
 
173
        """
 
174
        self._dirty = True
 
175
        if reset_inventory and self._inventory is not None:
 
176
            self._inventory = None
 
177
 
 
178
    @needs_tree_write_lock
 
179
    def add_reference(self, sub_tree):
 
180
        # use standard implementation, which calls back to self._add
 
181
        # 
 
182
        # So we don't store the reference_revision in the working dirstate,
 
183
        # it's just recorded at the moment of commit. 
 
184
        self._add_reference(sub_tree)
 
185
 
 
186
    def break_lock(self):
 
187
        """Break a lock if one is present from another instance.
 
188
 
 
189
        Uses the ui factory to ask for confirmation if the lock may be from
 
190
        an active process.
 
191
 
 
192
        This will probe the repository for its lock as well.
 
193
        """
 
194
        # if the dirstate is locked by an active process, reject the break lock
 
195
        # call.
 
196
        try:
 
197
            if self._dirstate is None:
 
198
                clear = True
 
199
            else:
 
200
                clear = False
 
201
            state = self._current_dirstate()
 
202
            if state._lock_token is not None:
 
203
                # we already have it locked. sheese, cant break our own lock.
 
204
                raise errors.LockActive(self.basedir)
 
205
            else:
 
206
                try:
 
207
                    # try for a write lock - need permission to get one anyhow
 
208
                    # to break locks.
 
209
                    state.lock_write()
 
210
                except errors.LockContention:
 
211
                    # oslocks fail when a process is still live: fail.
 
212
                    # TODO: get the locked lockdir info and give to the user to
 
213
                    # assist in debugging.
 
214
                    raise errors.LockActive(self.basedir)
 
215
                else:
 
216
                    state.unlock()
 
217
        finally:
 
218
            if clear:
 
219
                self._dirstate = None
 
220
        self._control_files.break_lock()
 
221
        self.branch.break_lock()
 
222
 
 
223
    def _comparison_data(self, entry, path):
 
224
        kind, executable, stat_value = \
 
225
            WorkingTree3._comparison_data(self, entry, path)
 
226
        # it looks like a plain directory, but it's really a reference -- see
 
227
        # also kind()
 
228
        if (self._repo_supports_tree_reference and
 
229
            kind == 'directory' and
 
230
            self._directory_is_tree_reference(path)):
 
231
            kind = 'tree-reference'
 
232
        return kind, executable, stat_value
 
233
 
 
234
    @needs_write_lock
 
235
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
236
        # mark the tree as dirty post commit - commit
 
237
        # can change the current versioned list by doing deletes.
 
238
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
239
        self._make_dirty(reset_inventory=True)
 
240
        return result
 
241
 
 
242
    def current_dirstate(self):
 
243
        """Return the current dirstate object.
 
244
 
 
245
        This is not part of the tree interface and only exposed for ease of
 
246
        testing.
 
247
 
 
248
        :raises errors.NotWriteLocked: when not in a lock.
 
249
        """
 
250
        self._must_be_locked()
 
251
        return self._current_dirstate()
 
252
 
 
253
    def _current_dirstate(self):
 
254
        """Internal function that does not check lock status.
 
255
 
 
256
        This is needed for break_lock which also needs the dirstate.
 
257
        """
 
258
        if self._dirstate is not None:
 
259
            return self._dirstate
 
260
        local_path = self.bzrdir.get_workingtree_transport(None
 
261
            ).local_abspath('dirstate')
 
262
        self._dirstate = dirstate.DirState.on_file(local_path)
 
263
        return self._dirstate
 
264
 
 
265
    def _directory_is_tree_reference(self, relpath):
 
266
        # as a special case, if a directory contains control files then 
 
267
        # it's a tree reference, except that the root of the tree is not
 
268
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
269
        # TODO: We could ask all the control formats whether they
 
270
        # recognize this directory, but at the moment there's no cheap api
 
271
        # to do that.  Since we probably can only nest bzr checkouts and
 
272
        # they always use this name it's ok for now.  -- mbp 20060306
 
273
        #
 
274
        # FIXME: There is an unhandled case here of a subdirectory
 
275
        # containing .bzr but not a branch; that will probably blow up
 
276
        # when you try to commit it.  It might happen if there is a
 
277
        # checkout in a subdirectory.  This can be avoided by not adding
 
278
        # it.  mbp 20070306
 
279
 
 
280
    def filter_unversioned_files(self, paths):
 
281
        """Filter out paths that are versioned.
 
282
 
 
283
        :return: set of paths.
 
284
        """
 
285
        # TODO: make a generic multi-bisect routine roughly that should list
 
286
        # the paths, then process one half at a time recursively, and feed the
 
287
        # results of each bisect in further still
 
288
        paths = sorted(paths)
 
289
        result = set()
 
290
        state = self.current_dirstate()
 
291
        # TODO we want a paths_to_dirblocks helper I think
 
292
        for path in paths:
 
293
            dirname, basename = os.path.split(path.encode('utf8'))
 
294
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
295
                dirname, basename, 0)
 
296
            if not path_is_versioned:
 
297
                result.add(path)
 
298
        return result
 
299
 
 
300
    def flush(self):
 
301
        """Write all cached data to disk."""
 
302
        if self._control_files._lock_mode != 'w':
 
303
            raise errors.NotWriteLocked(self)
 
304
        self.current_dirstate().save()
 
305
        self._inventory = None
 
306
        self._dirty = False
 
307
 
 
308
    @needs_tree_write_lock
 
309
    def _gather_kinds(self, files, kinds):
 
310
        """See MutableTree._gather_kinds."""
 
311
        for pos, f in enumerate(files):
 
312
            if kinds[pos] is None:
 
313
                kinds[pos] = self._kind(f)
 
314
 
 
315
    def _generate_inventory(self):
 
316
        """Create and set self.inventory from the dirstate object.
 
317
        
 
318
        This is relatively expensive: we have to walk the entire dirstate.
 
319
        Ideally we would not, and can deprecate this function.
 
320
        """
 
321
        #: uncomment to trap on inventory requests.
 
322
        # import pdb;pdb.set_trace()
 
323
        state = self.current_dirstate()
 
324
        state._read_dirblocks_if_needed()
 
325
        root_key, current_entry = self._get_entry(path='')
 
326
        current_id = root_key[2]
 
327
        assert current_entry[0][0] == 'd' # directory
 
328
        inv = Inventory(root_id=current_id)
 
329
        # Turn some things into local variables
 
330
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
331
        factory = entry_factory
 
332
        utf8_decode = cache_utf8._utf8_decode
 
333
        inv_byid = inv._byid
 
334
        # we could do this straight out of the dirstate; it might be fast
 
335
        # and should be profiled - RBC 20070216
 
336
        parent_ies = {'' : inv.root}
 
337
        for block in state._dirblocks[1:]: # skip the root
 
338
            dirname = block[0]
 
339
            try:
 
340
                parent_ie = parent_ies[dirname]
 
341
            except KeyError:
 
342
                # all the paths in this block are not versioned in this tree
 
343
                continue
 
344
            for key, entry in block[1]:
 
345
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
346
                if minikind in ('a', 'r'): # absent, relocated
 
347
                    # a parent tree only entry
 
348
                    continue
 
349
                name = key[1]
 
350
                name_unicode = utf8_decode(name)[0]
 
351
                file_id = key[2]
 
352
                kind = minikind_to_kind[minikind]
 
353
                inv_entry = factory[kind](file_id, name_unicode,
 
354
                                          parent_ie.file_id)
 
355
                if kind == 'file':
 
356
                    # This is only needed on win32, where this is the only way
 
357
                    # we know the executable bit.
 
358
                    inv_entry.executable = executable
 
359
                    # not strictly needed: working tree
 
360
                    #inv_entry.text_size = size
 
361
                    #inv_entry.text_sha1 = sha1
 
362
                elif kind == 'directory':
 
363
                    # add this entry to the parent map.
 
364
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
365
                elif kind == 'tree-reference':
 
366
                    assert self._repo_supports_tree_reference, \
 
367
                        "repository of %r " \
 
368
                        "doesn't support tree references " \
 
369
                        "required by entry %r" \
 
370
                        % (self, name)
 
371
                    inv_entry.reference_revision = link_or_sha1 or None
 
372
                elif kind != 'symlink':
 
373
                    raise AssertionError("unknown kind %r" % kind)
 
374
                # These checks cost us around 40ms on a 55k entry tree
 
375
                assert file_id not in inv_byid, ('file_id %s already in'
 
376
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
377
                assert name_unicode not in parent_ie.children
 
378
                inv_byid[file_id] = inv_entry
 
379
                parent_ie.children[name_unicode] = inv_entry
 
380
        self._inventory = inv
 
381
 
 
382
    def _get_entry(self, file_id=None, path=None):
 
383
        """Get the dirstate row for file_id or path.
 
384
 
 
385
        If either file_id or path is supplied, it is used as the key to lookup.
 
386
        If both are supplied, the fastest lookup is used, and an error is
 
387
        raised if they do not both point at the same row.
 
388
        
 
389
        :param file_id: An optional unicode file_id to be looked up.
 
390
        :param path: An optional unicode path to be looked up.
 
391
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
392
        """
 
393
        if file_id is None and path is None:
 
394
            raise errors.BzrError('must supply file_id or path')
 
395
        state = self.current_dirstate()
 
396
        if path is not None:
 
397
            path = path.encode('utf8')
 
398
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
399
 
 
400
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
401
        # check file id is valid unconditionally.
 
402
        entry = self._get_entry(file_id=file_id, path=path)
 
403
        assert entry[0] is not None, 'what error should this raise'
 
404
        if path is None:
 
405
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
406
 
 
407
        file_abspath = self.abspath(path)
 
408
        state = self.current_dirstate()
 
409
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
410
                                          stat_value=stat_value)
 
411
        if entry[1][0][0] == 'f':
 
412
            return link_or_sha1
 
413
        return None
 
414
 
 
415
    def _get_inventory(self):
 
416
        """Get the inventory for the tree. This is only valid within a lock."""
 
417
        if self._inventory is not None:
 
418
            return self._inventory
 
419
        self._must_be_locked()
 
420
        self._generate_inventory()
 
421
        return self._inventory
 
422
 
 
423
    inventory = property(_get_inventory,
 
424
                         doc="Inventory of this Tree")
 
425
 
 
426
    @needs_read_lock
 
427
    def get_parent_ids(self):
 
428
        """See Tree.get_parent_ids.
 
429
        
 
430
        This implementation requests the ids list from the dirstate file.
 
431
        """
 
432
        return self.current_dirstate().get_parent_ids()
 
433
 
 
434
    def get_reference_revision(self, file_id, path=None):
 
435
        # referenced tree's revision is whatever's currently there
 
436
        return self.get_nested_tree(file_id, path).last_revision()
 
437
 
 
438
    def get_nested_tree(self, file_id, path=None):
 
439
        if path is None:
 
440
            path = self.id2path(file_id)
 
441
        # else: check file_id is at path?
 
442
        return WorkingTree.open(self.abspath(path))
 
443
 
 
444
    @needs_read_lock
 
445
    def get_root_id(self):
 
446
        """Return the id of this trees root"""
 
447
        return self._get_entry(path='')[0][2]
 
448
 
 
449
    def has_id(self, file_id):
 
450
        state = self.current_dirstate()
 
451
        file_id = osutils.safe_file_id(file_id)
 
452
        row, parents = self._get_entry(file_id=file_id)
 
453
        if row is None:
 
454
            return False
 
455
        return osutils.lexists(pathjoin(
 
456
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
457
 
 
458
    @needs_read_lock
 
459
    def id2path(self, file_id):
 
460
        file_id = osutils.safe_file_id(file_id)
 
461
        state = self.current_dirstate()
 
462
        entry = self._get_entry(file_id=file_id)
 
463
        if entry == (None, None):
 
464
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
465
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
466
        return path_utf8.decode('utf8')
 
467
 
 
468
    if not osutils.supports_executable():
 
469
        @needs_read_lock
 
470
        def is_executable(self, file_id, path=None):
 
471
            file_id = osutils.safe_file_id(file_id)
 
472
            entry = self._get_entry(file_id=file_id, path=path)
 
473
            if entry == (None, None):
 
474
                return False
 
475
            return entry[1][0][3]
 
476
    else:
 
477
        @needs_read_lock
 
478
        def is_executable(self, file_id, path=None):
 
479
            if not path:
 
480
                file_id = osutils.safe_file_id(file_id)
 
481
                path = self.id2path(file_id)
 
482
            mode = os.lstat(self.abspath(path)).st_mode
 
483
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
484
 
 
485
    @needs_read_lock
 
486
    def __iter__(self):
 
487
        """Iterate through file_ids for this tree.
 
488
 
 
489
        file_ids are in a WorkingTree if they are in the working inventory
 
490
        and the working file exists.
 
491
        """
 
492
        result = []
 
493
        for key, tree_details in self.current_dirstate()._iter_entries():
 
494
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
495
                # not relevant to the working tree
 
496
                continue
 
497
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
498
            if osutils.lexists(path):
 
499
                result.append(key[2])
 
500
        return iter(result)
 
501
 
 
502
    def iter_references(self):
 
503
        for key, tree_details in self.current_dirstate()._iter_entries():
 
504
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
505
                # not relevant to the working tree
 
506
                continue
 
507
            if not key[1]:
 
508
                # the root is not a reference.
 
509
                continue
 
510
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
511
            try:
 
512
                if self._kind(path) == 'tree-reference':
 
513
                    yield path, key[2]
 
514
            except errors.NoSuchFile:
 
515
                # path is missing on disk.
 
516
                continue
 
517
 
 
518
    @needs_read_lock
 
519
    def kind(self, file_id):
 
520
        """Return the kind of a file.
 
521
 
 
522
        This is always the actual kind that's on disk, regardless of what it
 
523
        was added as.
 
524
        """
 
525
        relpath = self.id2path(file_id)
 
526
        assert relpath != None, \
 
527
            "path for id {%s} is None!" % file_id
 
528
        return self._kind(relpath)
 
529
 
 
530
    def _kind(self, relpath):
 
531
        abspath = self.abspath(relpath)
 
532
        kind = file_kind(abspath)
 
533
        if (self._repo_supports_tree_reference and
 
534
            kind == 'directory' and
 
535
            self._directory_is_tree_reference(relpath)):
 
536
            kind = 'tree-reference'
 
537
        return kind
 
538
 
 
539
    @needs_read_lock
 
540
    def _last_revision(self):
 
541
        """See Mutable.last_revision."""
 
542
        parent_ids = self.current_dirstate().get_parent_ids()
 
543
        if parent_ids:
 
544
            return parent_ids[0]
 
545
        else:
 
546
            return None
 
547
 
 
548
    def lock_read(self):
 
549
        """See Branch.lock_read, and WorkingTree.unlock."""
 
550
        self.branch.lock_read()
 
551
        try:
 
552
            self._control_files.lock_read()
 
553
            try:
 
554
                state = self.current_dirstate()
 
555
                if not state._lock_token:
 
556
                    state.lock_read()
 
557
                # set our support for tree references from the repository in
 
558
                # use.
 
559
                self._repo_supports_tree_reference = getattr(
 
560
                    self.branch.repository._format, "support_tree_reference",
 
561
                    False)
 
562
            except:
 
563
                self._control_files.unlock()
 
564
                raise
 
565
        except:
 
566
            self.branch.unlock()
 
567
            raise
 
568
 
 
569
    def _lock_self_write(self):
 
570
        """This should be called after the branch is locked."""
 
571
        try:
 
572
            self._control_files.lock_write()
 
573
            try:
 
574
                state = self.current_dirstate()
 
575
                if not state._lock_token:
 
576
                    state.lock_write()
 
577
                # set our support for tree references from the repository in
 
578
                # use.
 
579
                self._repo_supports_tree_reference = getattr(
 
580
                    self.branch.repository._format, "support_tree_reference",
 
581
                    False)
 
582
            except:
 
583
                self._control_files.unlock()
 
584
                raise
 
585
        except:
 
586
            self.branch.unlock()
 
587
            raise
 
588
 
 
589
    def lock_tree_write(self):
 
590
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
591
        self.branch.lock_read()
 
592
        self._lock_self_write()
 
593
 
 
594
    def lock_write(self):
 
595
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
596
        self.branch.lock_write()
 
597
        self._lock_self_write()
 
598
 
 
599
    @needs_tree_write_lock
 
600
    def move(self, from_paths, to_dir, after=False):
 
601
        """See WorkingTree.move()."""
 
602
        result = []
 
603
        if not from_paths:
 
604
            return result
 
605
 
 
606
        state = self.current_dirstate()
 
607
 
 
608
        assert not isinstance(from_paths, basestring)
 
609
        to_dir_utf8 = to_dir.encode('utf8')
 
610
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
611
        id_index = state._get_id_index()
 
612
        # check destination directory
 
613
        # get the details for it
 
614
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
615
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
616
        if not entry_present:
 
617
            raise errors.BzrMoveFailedError('', to_dir,
 
618
                errors.NotVersionedError(to_dir))
 
619
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
620
        # get a handle on the block itself.
 
621
        to_block_index = state._ensure_block(
 
622
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
623
        to_block = state._dirblocks[to_block_index]
 
624
        to_abs = self.abspath(to_dir)
 
625
        if not isdir(to_abs):
 
626
            raise errors.BzrMoveFailedError('',to_dir,
 
627
                errors.NotADirectory(to_abs))
 
628
 
 
629
        if to_entry[1][0][0] != 'd':
 
630
            raise errors.BzrMoveFailedError('',to_dir,
 
631
                errors.NotADirectory(to_abs))
 
632
 
 
633
        if self._inventory is not None:
 
634
            update_inventory = True
 
635
            inv = self.inventory
 
636
            to_dir_ie = inv[to_dir_id]
 
637
            to_dir_id = to_entry[0][2]
 
638
        else:
 
639
            update_inventory = False
 
640
 
 
641
        rollbacks = []
 
642
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
643
                     fingerprint, packed_stat, size,
 
644
                     to_block, to_key, to_path_utf8):
 
645
            state._make_absent(old_entry)
 
646
            from_key = old_entry[0]
 
647
            rollbacks.append(
 
648
                lambda:state.update_minimal(from_key,
 
649
                    minikind,
 
650
                    executable=executable,
 
651
                    fingerprint=fingerprint,
 
652
                    packed_stat=packed_stat,
 
653
                    size=size,
 
654
                    path_utf8=from_path_utf8))
 
655
            state.update_minimal(to_key,
 
656
                    minikind,
 
657
                    executable=executable,
 
658
                    fingerprint=fingerprint,
 
659
                    packed_stat=packed_stat,
 
660
                    size=size,
 
661
                    path_utf8=to_path_utf8)
 
662
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
663
            new_entry = to_block[1][added_entry_index]
 
664
            rollbacks.append(lambda:state._make_absent(new_entry))
 
665
 
 
666
        # create rename entries and tuples
 
667
        for from_rel in from_paths:
 
668
            # from_rel is 'pathinroot/foo/bar'
 
669
            from_rel_utf8 = from_rel.encode('utf8')
 
670
            from_dirname, from_tail = osutils.split(from_rel)
 
671
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
672
            from_entry = self._get_entry(path=from_rel)
 
673
            if from_entry == (None, None):
 
674
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
675
                    errors.NotVersionedError(path=str(from_rel)))
 
676
 
 
677
            from_id = from_entry[0][2]
 
678
            to_rel = pathjoin(to_dir, from_tail)
 
679
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
680
            item_to_entry = self._get_entry(path=to_rel)
 
681
            if item_to_entry != (None, None):
 
682
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
683
                    "Target is already versioned.")
 
684
 
 
685
            if from_rel == to_rel:
 
686
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
687
                    "Source and target are identical.")
 
688
 
 
689
            from_missing = not self.has_filename(from_rel)
 
690
            to_missing = not self.has_filename(to_rel)
 
691
            if after:
 
692
                move_file = False
 
693
            else:
 
694
                move_file = True
 
695
            if to_missing:
 
696
                if not move_file:
 
697
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
698
                        errors.NoSuchFile(path=to_rel,
 
699
                        extra="New file has not been created yet"))
 
700
                elif from_missing:
 
701
                    # neither path exists
 
702
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
703
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
704
            else:
 
705
                if from_missing: # implicitly just update our path mapping
 
706
                    move_file = False
 
707
                elif not after:
 
708
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
709
                        extra="(Use --after to update the Bazaar id)")
 
710
 
 
711
            rollbacks = []
 
712
            def rollback_rename():
 
713
                """A single rename has failed, roll it back."""
 
714
                # roll back everything, even if we encounter trouble doing one
 
715
                # of them.
 
716
                #
 
717
                # TODO: at least log the other exceptions rather than just
 
718
                # losing them mbp 20070307
 
719
                exc_info = None
 
720
                for rollback in reversed(rollbacks):
 
721
                    try:
 
722
                        rollback()
 
723
                    except Exception, e:
 
724
                        exc_info = sys.exc_info()
 
725
                if exc_info:
 
726
                    raise exc_info[0], exc_info[1], exc_info[2]
 
727
 
 
728
            # perform the disk move first - its the most likely failure point.
 
729
            if move_file:
 
730
                from_rel_abs = self.abspath(from_rel)
 
731
                to_rel_abs = self.abspath(to_rel)
 
732
                try:
 
733
                    osutils.rename(from_rel_abs, to_rel_abs)
 
734
                except OSError, e:
 
735
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
736
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
737
            try:
 
738
                # perform the rename in the inventory next if needed: its easy
 
739
                # to rollback
 
740
                if update_inventory:
 
741
                    # rename the entry
 
742
                    from_entry = inv[from_id]
 
743
                    current_parent = from_entry.parent_id
 
744
                    inv.rename(from_id, to_dir_id, from_tail)
 
745
                    rollbacks.append(
 
746
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
747
                # finally do the rename in the dirstate, which is a little
 
748
                # tricky to rollback, but least likely to need it.
 
749
                old_block_index, old_entry_index, dir_present, file_present = \
 
750
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
751
                old_block = state._dirblocks[old_block_index][1]
 
752
                old_entry = old_block[old_entry_index]
 
753
                from_key, old_entry_details = old_entry
 
754
                cur_details = old_entry_details[0]
 
755
                # remove the old row
 
756
                to_key = ((to_block[0],) + from_key[1:3])
 
757
                minikind = cur_details[0]
 
758
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
759
                         minikind=minikind,
 
760
                         executable=cur_details[3],
 
761
                         fingerprint=cur_details[1],
 
762
                         packed_stat=cur_details[4],
 
763
                         size=cur_details[2],
 
764
                         to_block=to_block,
 
765
                         to_key=to_key,
 
766
                         to_path_utf8=to_rel_utf8)
 
767
 
 
768
                if minikind == 'd':
 
769
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
770
                        """all entries in this block need updating.
 
771
 
 
772
                        TODO: This is pretty ugly, and doesn't support
 
773
                        reverting, but it works.
 
774
                        """
 
775
                        assert from_dir != '', "renaming root not supported"
 
776
                        from_key = (from_dir, '')
 
777
                        from_block_idx, present = \
 
778
                            state._find_block_index_from_key(from_key)
 
779
                        if not present:
 
780
                            # This is the old record, if it isn't present, then
 
781
                            # there is theoretically nothing to update.
 
782
                            # (Unless it isn't present because of lazy loading,
 
783
                            # but we don't do that yet)
 
784
                            return
 
785
                        from_block = state._dirblocks[from_block_idx]
 
786
                        to_block_index, to_entry_index, _, _ = \
 
787
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
788
                        to_block_index = state._ensure_block(
 
789
                            to_block_index, to_entry_index, to_dir_utf8)
 
790
                        to_block = state._dirblocks[to_block_index]
 
791
                        for entry in from_block[1]:
 
792
                            assert entry[0][0] == from_dir
 
793
                            cur_details = entry[1][0]
 
794
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
795
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
796
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
797
                            minikind = cur_details[0]
 
798
                            move_one(entry, from_path_utf8=from_path_utf8,
 
799
                                     minikind=minikind,
 
800
                                     executable=cur_details[3],
 
801
                                     fingerprint=cur_details[1],
 
802
                                     packed_stat=cur_details[4],
 
803
                                     size=cur_details[2],
 
804
                                     to_block=to_block,
 
805
                                     to_key=to_key,
 
806
                                     to_path_utf8=to_rel_utf8)
 
807
                            if minikind == 'd':
 
808
                                # We need to move all the children of this
 
809
                                # entry
 
810
                                update_dirblock(from_path_utf8, to_key,
 
811
                                                to_path_utf8)
 
812
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
813
            except:
 
814
                rollback_rename()
 
815
                raise
 
816
            result.append((from_rel, to_rel))
 
817
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
818
            self._make_dirty(reset_inventory=False)
 
819
 
 
820
        return result
 
821
 
 
822
    def _must_be_locked(self):
 
823
        if not self._control_files._lock_count:
 
824
            raise errors.ObjectNotLocked(self)
 
825
 
 
826
    def _new_tree(self):
 
827
        """Initialize the state in this tree to be a new tree."""
 
828
        self._dirty = True
 
829
 
 
830
    @needs_read_lock
 
831
    def path2id(self, path):
 
832
        """Return the id for path in this tree."""
 
833
        path = path.strip('/')
 
834
        entry = self._get_entry(path=path)
 
835
        if entry == (None, None):
 
836
            return None
 
837
        return entry[0][2]
 
838
 
 
839
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
840
        """See Tree.paths2ids().
 
841
 
 
842
        This specialisation fast-paths the case where all the trees are in the
 
843
        dirstate.
 
844
        """
 
845
        if paths is None:
 
846
            return None
 
847
        parents = self.get_parent_ids()
 
848
        for tree in trees:
 
849
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
850
                parents):
 
851
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
852
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
853
        # -- make all paths utf8 --
 
854
        paths_utf8 = set()
 
855
        for path in paths:
 
856
            paths_utf8.add(path.encode('utf8'))
 
857
        paths = paths_utf8
 
858
        # -- paths is now a utf8 path set --
 
859
        # -- get the state object and prepare it.
 
860
        state = self.current_dirstate()
 
861
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
862
            and '' not in paths):
 
863
            paths2ids = self._paths2ids_using_bisect
 
864
        else:
 
865
            paths2ids = self._paths2ids_in_memory
 
866
        return paths2ids(paths, search_indexes,
 
867
                         require_versioned=require_versioned)
 
868
 
 
869
    def _paths2ids_in_memory(self, paths, search_indexes,
 
870
                             require_versioned=True):
 
871
        state = self.current_dirstate()
 
872
        state._read_dirblocks_if_needed()
 
873
        def _entries_for_path(path):
 
874
            """Return a list with all the entries that match path for all ids.
 
875
            """
 
876
            dirname, basename = os.path.split(path)
 
877
            key = (dirname, basename, '')
 
878
            block_index, present = state._find_block_index_from_key(key)
 
879
            if not present:
 
880
                # the block which should contain path is absent.
 
881
                return []
 
882
            result = []
 
883
            block = state._dirblocks[block_index][1]
 
884
            entry_index, _ = state._find_entry_index(key, block)
 
885
            # we may need to look at multiple entries at this path: walk while the paths match.
 
886
            while (entry_index < len(block) and
 
887
                block[entry_index][0][0:2] == key[0:2]):
 
888
                result.append(block[entry_index])
 
889
                entry_index += 1
 
890
            return result
 
891
        if require_versioned:
 
892
            # -- check all supplied paths are versioned in a search tree. --
 
893
            all_versioned = True
 
894
            for path in paths:
 
895
                path_entries = _entries_for_path(path)
 
896
                if not path_entries:
 
897
                    # this specified path is not present at all: error
 
898
                    all_versioned = False
 
899
                    break
 
900
                found_versioned = False
 
901
                # for each id at this path
 
902
                for entry in path_entries:
 
903
                    # for each tree.
 
904
                    for index in search_indexes:
 
905
                        if entry[1][index][0] != 'a': # absent
 
906
                            found_versioned = True
 
907
                            # all good: found a versioned cell
 
908
                            break
 
909
                if not found_versioned:
 
910
                    # none of the indexes was not 'absent' at all ids for this
 
911
                    # path.
 
912
                    all_versioned = False
 
913
                    break
 
914
            if not all_versioned:
 
915
                raise errors.PathsNotVersionedError(paths)
 
916
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
917
        search_paths = set()
 
918
        for path in paths:
 
919
            other_paths = paths.difference(set([path]))
 
920
            if not osutils.is_inside_any(other_paths, path):
 
921
                # this is a top level path, we must check it.
 
922
                search_paths.add(path)
 
923
        # sketch: 
 
924
        # for all search_indexs in each path at or under each element of
 
925
        # search_paths, if the detail is relocated: add the id, and add the
 
926
        # relocated path as one to search if its not searched already. If the
 
927
        # detail is not relocated, add the id.
 
928
        searched_paths = set()
 
929
        found_ids = set()
 
930
        def _process_entry(entry):
 
931
            """Look at search_indexes within entry.
 
932
 
 
933
            If a specific tree's details are relocated, add the relocation
 
934
            target to search_paths if not searched already. If it is absent, do
 
935
            nothing. Otherwise add the id to found_ids.
 
936
            """
 
937
            for index in search_indexes:
 
938
                if entry[1][index][0] == 'r': # relocated
 
939
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
940
                        search_paths.add(entry[1][index][1])
 
941
                elif entry[1][index][0] != 'a': # absent
 
942
                    found_ids.add(entry[0][2])
 
943
        while search_paths:
 
944
            current_root = search_paths.pop()
 
945
            searched_paths.add(current_root)
 
946
            # process the entries for this containing directory: the rest will be
 
947
            # found by their parents recursively.
 
948
            root_entries = _entries_for_path(current_root)
 
949
            if not root_entries:
 
950
                # this specified path is not present at all, skip it.
 
951
                continue
 
952
            for entry in root_entries:
 
953
                _process_entry(entry)
 
954
            initial_key = (current_root, '', '')
 
955
            block_index, _ = state._find_block_index_from_key(initial_key)
 
956
            while (block_index < len(state._dirblocks) and
 
957
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
958
                for entry in state._dirblocks[block_index][1]:
 
959
                    _process_entry(entry)
 
960
                block_index += 1
 
961
        return found_ids
 
962
 
 
963
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
964
                                require_versioned=True):
 
965
        state = self.current_dirstate()
 
966
        found_ids = set()
 
967
 
 
968
        split_paths = sorted(osutils.split(p) for p in paths)
 
969
        found = state._bisect_recursive(split_paths)
 
970
 
 
971
        if require_versioned:
 
972
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
973
            for dir_name in split_paths:
 
974
                if dir_name not in found_dir_names:
 
975
                    raise errors.PathsNotVersionedError(paths)
 
976
 
 
977
        for dir_name_id, trees_info in found.iteritems():
 
978
            for index in search_indexes:
 
979
                if trees_info[index][0] not in ('r', 'a'):
 
980
                    found_ids.add(dir_name_id[2])
 
981
        return found_ids
 
982
 
 
983
    def read_working_inventory(self):
 
984
        """Read the working inventory.
 
985
        
 
986
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
987
        """
 
988
        return self.inventory
 
989
 
 
990
    @needs_read_lock
 
991
    def revision_tree(self, revision_id):
 
992
        """See Tree.revision_tree.
 
993
 
 
994
        WorkingTree4 supplies revision_trees for any basis tree.
 
995
        """
 
996
        revision_id = osutils.safe_revision_id(revision_id)
 
997
        dirstate = self.current_dirstate()
 
998
        parent_ids = dirstate.get_parent_ids()
 
999
        if revision_id not in parent_ids:
 
1000
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1001
        if revision_id in dirstate.get_ghosts():
 
1002
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1003
        return DirStateRevisionTree(dirstate, revision_id,
 
1004
            self.branch.repository)
 
1005
 
 
1006
    @needs_tree_write_lock
 
1007
    def set_last_revision(self, new_revision):
 
1008
        """Change the last revision in the working tree."""
 
1009
        new_revision = osutils.safe_revision_id(new_revision)
 
1010
        parents = self.get_parent_ids()
 
1011
        if new_revision in (NULL_REVISION, None):
 
1012
            assert len(parents) < 2, (
 
1013
                "setting the last parent to none with a pending merge is "
 
1014
                "unsupported.")
 
1015
            self.set_parent_ids([])
 
1016
        else:
 
1017
            self.set_parent_ids([new_revision] + parents[1:],
 
1018
                allow_leftmost_as_ghost=True)
 
1019
 
 
1020
    @needs_tree_write_lock
 
1021
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1022
        """Set the parent ids to revision_ids.
 
1023
        
 
1024
        See also set_parent_trees. This api will try to retrieve the tree data
 
1025
        for each element of revision_ids from the trees repository. If you have
 
1026
        tree data already available, it is more efficient to use
 
1027
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1028
        an easier API to use.
 
1029
 
 
1030
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1031
            working tree. Any of these may be ghosts.
 
1032
        """
 
1033
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1034
        trees = []
 
1035
        for revision_id in revision_ids:
 
1036
            try:
 
1037
                revtree = self.branch.repository.revision_tree(revision_id)
 
1038
                # TODO: jam 20070213 KnitVersionedFile raises
 
1039
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1040
                #       given revision_id is not present. Should Repository be
 
1041
                #       catching it and re-raising NoSuchRevision?
 
1042
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1043
                revtree = None
 
1044
            trees.append((revision_id, revtree))
 
1045
        self.current_dirstate()._validate()
 
1046
        self.set_parent_trees(trees,
 
1047
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1048
        self.current_dirstate()._validate()
 
1049
 
 
1050
    @needs_tree_write_lock
 
1051
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1052
        """Set the parents of the working tree.
 
1053
 
 
1054
        :param parents_list: A list of (revision_id, tree) tuples.
 
1055
            If tree is None, then that element is treated as an unreachable
 
1056
            parent tree - i.e. a ghost.
 
1057
        """
 
1058
        dirstate = self.current_dirstate()
 
1059
        dirstate._validate()
 
1060
        if len(parents_list) > 0:
 
1061
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1062
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1063
        real_trees = []
 
1064
        ghosts = []
 
1065
        # convert absent trees to the null tree, which we convert back to
 
1066
        # missing on access.
 
1067
        for rev_id, tree in parents_list:
 
1068
            rev_id = osutils.safe_revision_id(rev_id)
 
1069
            if tree is not None:
 
1070
                real_trees.append((rev_id, tree))
 
1071
            else:
 
1072
                real_trees.append((rev_id,
 
1073
                    self.branch.repository.revision_tree(None)))
 
1074
                ghosts.append(rev_id)
 
1075
        dirstate._validate()
 
1076
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1077
        dirstate._validate()
 
1078
        self._make_dirty(reset_inventory=False)
 
1079
        dirstate._validate()
 
1080
 
 
1081
    def _set_root_id(self, file_id):
 
1082
        """See WorkingTree.set_root_id."""
 
1083
        state = self.current_dirstate()
 
1084
        state.set_path_id('', file_id)
 
1085
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1086
            self._make_dirty(reset_inventory=True)
 
1087
 
 
1088
    @needs_read_lock
 
1089
    def supports_tree_reference(self):
 
1090
        return self._repo_supports_tree_reference
 
1091
 
 
1092
    def unlock(self):
 
1093
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1094
        if self._control_files._lock_count == 1:
 
1095
            # eventually we should do signature checking during read locks for
 
1096
            # dirstate updates.
 
1097
            if self._control_files._lock_mode == 'w':
 
1098
                if self._dirty:
 
1099
                    self.flush()
 
1100
            if self._dirstate is not None:
 
1101
                # This is a no-op if there are no modifications.
 
1102
                self._dirstate.save()
 
1103
                self._dirstate.unlock()
 
1104
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1105
            #       point. Instead, it could check if the header has been
 
1106
            #       modified when it is locked, and if not, it can hang on to
 
1107
            #       the data it has in memory.
 
1108
            self._dirstate = None
 
1109
            self._inventory = None
 
1110
        # reverse order of locking.
 
1111
        try:
 
1112
            return self._control_files.unlock()
 
1113
        finally:
 
1114
            self.branch.unlock()
 
1115
 
 
1116
    @needs_tree_write_lock
 
1117
    def unversion(self, file_ids):
 
1118
        """Remove the file ids in file_ids from the current versioned set.
 
1119
 
 
1120
        When a file_id is unversioned, all of its children are automatically
 
1121
        unversioned.
 
1122
 
 
1123
        :param file_ids: The file ids to stop versioning.
 
1124
        :raises: NoSuchId if any fileid is not currently versioned.
 
1125
        """
 
1126
        if not file_ids:
 
1127
            return
 
1128
        state = self.current_dirstate()
 
1129
        state._read_dirblocks_if_needed()
 
1130
        ids_to_unversion = set()
 
1131
        for file_id in file_ids:
 
1132
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1133
        paths_to_unversion = set()
 
1134
        # sketch:
 
1135
        # check if the root is to be unversioned, if so, assert for now.
 
1136
        # walk the state marking unversioned things as absent.
 
1137
        # if there are any un-unversioned ids at the end, raise
 
1138
        for key, details in state._dirblocks[0][1]:
 
1139
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1140
                key[2] in ids_to_unversion):
 
1141
                # I haven't written the code to unversion / yet - it should be
 
1142
                # supported.
 
1143
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1144
        block_index = 0
 
1145
        while block_index < len(state._dirblocks):
 
1146
            # process one directory at a time.
 
1147
            block = state._dirblocks[block_index]
 
1148
            # first check: is the path one to remove - it or its children
 
1149
            delete_block = False
 
1150
            for path in paths_to_unversion:
 
1151
                if (block[0].startswith(path) and
 
1152
                    (len(block[0]) == len(path) or
 
1153
                     block[0][len(path)] == '/')):
 
1154
                    # this entire block should be deleted - its the block for a
 
1155
                    # path to unversion; or the child of one
 
1156
                    delete_block = True
 
1157
                    break
 
1158
            # TODO: trim paths_to_unversion as we pass by paths
 
1159
            if delete_block:
 
1160
                # this block is to be deleted: process it.
 
1161
                # TODO: we can special case the no-parents case and
 
1162
                # just forget the whole block.
 
1163
                entry_index = 0
 
1164
                while entry_index < len(block[1]):
 
1165
                    # Mark this file id as having been removed
 
1166
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1167
                    if not state._make_absent(block[1][entry_index]):
 
1168
                        entry_index += 1
 
1169
                # go to the next block. (At the moment we dont delete empty
 
1170
                # dirblocks)
 
1171
                block_index += 1
 
1172
                continue
 
1173
            entry_index = 0
 
1174
            while entry_index < len(block[1]):
 
1175
                entry = block[1][entry_index]
 
1176
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1177
                    # ^ some parent row.
 
1178
                    entry[0][2] not in ids_to_unversion):
 
1179
                    # ^ not an id to unversion
 
1180
                    entry_index += 1
 
1181
                    continue
 
1182
                if entry[1][0][0] == 'd':
 
1183
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1184
                if not state._make_absent(entry):
 
1185
                    entry_index += 1
 
1186
                # we have unversioned this id
 
1187
                ids_to_unversion.remove(entry[0][2])
 
1188
            block_index += 1
 
1189
        if ids_to_unversion:
 
1190
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1191
        self._make_dirty(reset_inventory=False)
 
1192
        # have to change the legacy inventory too.
 
1193
        if self._inventory is not None:
 
1194
            for file_id in file_ids:
 
1195
                self._inventory.remove_recursive_id(file_id)
 
1196
 
 
1197
    @needs_tree_write_lock
 
1198
    def _write_inventory(self, inv):
 
1199
        """Write inventory as the current inventory."""
 
1200
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1201
        self.current_dirstate().set_state_from_inventory(inv)
 
1202
        self._make_dirty(reset_inventory=False)
 
1203
        if self._inventory is not None:
 
1204
            self._inventory = inv
 
1205
        self.flush()
 
1206
 
 
1207
 
 
1208
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1209
    """The first consolidated dirstate working tree format.
 
1210
 
 
1211
    This format:
 
1212
        - exists within a metadir controlling .bzr
 
1213
        - includes an explicit version marker for the workingtree control
 
1214
          files, separate from the BzrDir format
 
1215
        - modifies the hash cache format
 
1216
        - is new in bzr 0.15
 
1217
        - uses a LockDir to guard access to it.
 
1218
    """
 
1219
 
 
1220
    def get_format_string(self):
 
1221
        """See WorkingTreeFormat.get_format_string()."""
 
1222
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1223
 
 
1224
    def get_format_description(self):
 
1225
        """See WorkingTreeFormat.get_format_description()."""
 
1226
        return "Working tree format 4"
 
1227
 
 
1228
    def initialize(self, a_bzrdir, revision_id=None):
 
1229
        """See WorkingTreeFormat.initialize().
 
1230
 
 
1231
        :param revision_id: allows creating a working tree at a different
 
1232
        revision than the branch is at.
 
1233
 
 
1234
        These trees get an initial random root id.
 
1235
        """
 
1236
        revision_id = osutils.safe_revision_id(revision_id)
 
1237
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1238
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1239
        transport = a_bzrdir.get_workingtree_transport(self)
 
1240
        control_files = self._open_control_files(a_bzrdir)
 
1241
        control_files.create_lock()
 
1242
        control_files.lock_write()
 
1243
        control_files.put_utf8('format', self.get_format_string())
 
1244
        branch = a_bzrdir.open_branch()
 
1245
        if revision_id is None:
 
1246
            revision_id = branch.last_revision()
 
1247
        local_path = transport.local_abspath('dirstate')
 
1248
        # write out new dirstate (must exist when we create the tree)
 
1249
        state = dirstate.DirState.initialize(local_path)
 
1250
        state.unlock()
 
1251
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1252
                         branch,
 
1253
                         _format=self,
 
1254
                         _bzrdir=a_bzrdir,
 
1255
                         _control_files=control_files)
 
1256
        wt._new_tree()
 
1257
        wt.lock_tree_write()
 
1258
        state._validate()
 
1259
        try:
 
1260
            if revision_id in (None, NULL_REVISION):
 
1261
                wt._set_root_id(generate_ids.gen_root_id())
 
1262
                wt.flush()
 
1263
                wt.current_dirstate()._validate()
 
1264
            wt.set_last_revision(revision_id)
 
1265
            wt.flush()
 
1266
            basis = wt.basis_tree()
 
1267
            basis.lock_read()
 
1268
            # if the basis has a root id we have to use that; otherwise we use
 
1269
            # a new random one
 
1270
            basis_root_id = basis.get_root_id()
 
1271
            if basis_root_id is not None:
 
1272
                wt._set_root_id(basis_root_id)
 
1273
                wt.flush()
 
1274
            transform.build_tree(basis, wt)
 
1275
            basis.unlock()
 
1276
        finally:
 
1277
            control_files.unlock()
 
1278
            wt.unlock()
 
1279
        return wt
 
1280
 
 
1281
    def _open(self, a_bzrdir, control_files):
 
1282
        """Open the tree itself.
 
1283
 
 
1284
        :param a_bzrdir: the dir for the tree.
 
1285
        :param control_files: the control files for the tree.
 
1286
        """
 
1287
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1288
                           branch=a_bzrdir.open_branch(),
 
1289
                           _format=self,
 
1290
                           _bzrdir=a_bzrdir,
 
1291
                           _control_files=control_files)
 
1292
 
 
1293
    def __get_matchingbzrdir(self):
 
1294
        # please test against something that will let us do tree references
 
1295
        return bzrdir.format_registry.make_bzrdir(
 
1296
            'dirstate-with-subtree')
 
1297
 
 
1298
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1299
 
 
1300
 
 
1301
class DirStateRevisionTree(Tree):
 
1302
    """A revision tree pulling the inventory from a dirstate."""
 
1303
 
 
1304
    def __init__(self, dirstate, revision_id, repository):
 
1305
        self._dirstate = dirstate
 
1306
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1307
        self._repository = repository
 
1308
        self._inventory = None
 
1309
        self._locked = 0
 
1310
        self._dirstate_locked = False
 
1311
 
 
1312
    def __repr__(self):
 
1313
        return "<%s of %s in %s>" % \
 
1314
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1315
 
 
1316
    def annotate_iter(self, file_id):
 
1317
        """See Tree.annotate_iter"""
 
1318
        w = self._repository.weave_store.get_weave(file_id,
 
1319
                           self._repository.get_transaction())
 
1320
        return w.annotate_iter(self.inventory[file_id].revision)
 
1321
 
 
1322
    def _comparison_data(self, entry, path):
 
1323
        """See Tree._comparison_data."""
 
1324
        if entry is None:
 
1325
            return None, False, None
 
1326
        # trust the entry as RevisionTree does, but this may not be
 
1327
        # sensible: the entry might not have come from us?
 
1328
        return entry.kind, entry.executable, None
 
1329
 
 
1330
    def _file_size(self, entry, stat_value):
 
1331
        return entry.text_size
 
1332
 
 
1333
    def filter_unversioned_files(self, paths):
 
1334
        """Filter out paths that are not versioned.
 
1335
 
 
1336
        :return: set of paths.
 
1337
        """
 
1338
        pred = self.has_filename
 
1339
        return set((p for p in paths if not pred(p)))
 
1340
 
 
1341
    def get_root_id(self):
 
1342
        return self.path2id('')
 
1343
 
 
1344
    def _get_parent_index(self):
 
1345
        """Return the index in the dirstate referenced by this tree."""
 
1346
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1347
 
 
1348
    def _get_entry(self, file_id=None, path=None):
 
1349
        """Get the dirstate row for file_id or path.
 
1350
 
 
1351
        If either file_id or path is supplied, it is used as the key to lookup.
 
1352
        If both are supplied, the fastest lookup is used, and an error is
 
1353
        raised if they do not both point at the same row.
 
1354
        
 
1355
        :param file_id: An optional unicode file_id to be looked up.
 
1356
        :param path: An optional unicode path to be looked up.
 
1357
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1358
        """
 
1359
        if file_id is None and path is None:
 
1360
            raise errors.BzrError('must supply file_id or path')
 
1361
        file_id = osutils.safe_file_id(file_id)
 
1362
        if path is not None:
 
1363
            path = path.encode('utf8')
 
1364
        parent_index = self._get_parent_index()
 
1365
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1366
 
 
1367
    def _generate_inventory(self):
 
1368
        """Create and set self.inventory from the dirstate object.
 
1369
 
 
1370
        (So this is only called the first time the inventory is requested for
 
1371
        this tree; it then remains in memory until it's out of date.)
 
1372
 
 
1373
        This is relatively expensive: we have to walk the entire dirstate.
 
1374
        """
 
1375
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1376
            'dirstate revision tree'
 
1377
        # separate call for profiling - makes it clear where the costs are.
 
1378
        self._dirstate._read_dirblocks_if_needed()
 
1379
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1380
            'parent %s has disappeared from %s' % (
 
1381
            self._revision_id, self._dirstate.get_parent_ids())
 
1382
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1383
        # This is identical now to the WorkingTree _generate_inventory except
 
1384
        # for the tree index use.
 
1385
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1386
        current_id = root_key[2]
 
1387
        assert current_entry[parent_index][0] == 'd'
 
1388
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1389
        inv.root.revision = current_entry[parent_index][4]
 
1390
        # Turn some things into local variables
 
1391
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1392
        factory = entry_factory
 
1393
        utf8_decode = cache_utf8._utf8_decode
 
1394
        inv_byid = inv._byid
 
1395
        # we could do this straight out of the dirstate; it might be fast
 
1396
        # and should be profiled - RBC 20070216
 
1397
        parent_ies = {'' : inv.root}
 
1398
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1399
            dirname = block[0]
 
1400
            try:
 
1401
                parent_ie = parent_ies[dirname]
 
1402
            except KeyError:
 
1403
                # all the paths in this block are not versioned in this tree
 
1404
                continue
 
1405
            for key, entry in block[1]:
 
1406
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1407
                if minikind in ('a', 'r'): # absent, relocated
 
1408
                    # not this tree
 
1409
                    continue
 
1410
                name = key[1]
 
1411
                name_unicode = utf8_decode(name)[0]
 
1412
                file_id = key[2]
 
1413
                kind = minikind_to_kind[minikind]
 
1414
                inv_entry = factory[kind](file_id, name_unicode,
 
1415
                                          parent_ie.file_id)
 
1416
                inv_entry.revision = revid
 
1417
                if kind == 'file':
 
1418
                    inv_entry.executable = executable
 
1419
                    inv_entry.text_size = size
 
1420
                    inv_entry.text_sha1 = fingerprint
 
1421
                elif kind == 'directory':
 
1422
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1423
                elif kind == 'symlink':
 
1424
                    inv_entry.executable = False
 
1425
                    inv_entry.text_size = None
 
1426
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1427
                elif kind == 'tree-reference':
 
1428
                    inv_entry.reference_revision = fingerprint or None
 
1429
                else:
 
1430
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1431
                            % entry)
 
1432
                # These checks cost us around 40ms on a 55k entry tree
 
1433
                assert file_id not in inv_byid
 
1434
                assert name_unicode not in parent_ie.children
 
1435
                inv_byid[file_id] = inv_entry
 
1436
                parent_ie.children[name_unicode] = inv_entry
 
1437
        self._inventory = inv
 
1438
 
 
1439
    def get_file_mtime(self, file_id, path=None):
 
1440
        """Return the modification time for this record.
 
1441
 
 
1442
        We return the timestamp of the last-changed revision.
 
1443
        """
 
1444
        # Make sure the file exists
 
1445
        entry = self._get_entry(file_id, path=path)
 
1446
        if entry == (None, None): # do we raise?
 
1447
            return None
 
1448
        parent_index = self._get_parent_index()
 
1449
        last_changed_revision = entry[1][parent_index][4]
 
1450
        return self._repository.get_revision(last_changed_revision).timestamp
 
1451
 
 
1452
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1453
        entry = self._get_entry(file_id=file_id, path=path)
 
1454
        parent_index = self._get_parent_index()
 
1455
        parent_details = entry[1][parent_index]
 
1456
        if parent_details[0] == 'f':
 
1457
            return parent_details[1]
 
1458
        return None
 
1459
 
 
1460
    def get_file(self, file_id):
 
1461
        return StringIO(self.get_file_text(file_id))
 
1462
 
 
1463
    def get_file_lines(self, file_id):
 
1464
        ie = self.inventory[file_id]
 
1465
        return self._repository.weave_store.get_weave(file_id,
 
1466
                self._repository.get_transaction()).get_lines(ie.revision)
 
1467
 
 
1468
    def get_file_size(self, file_id):
 
1469
        return self.inventory[file_id].text_size
 
1470
 
 
1471
    def get_file_text(self, file_id):
 
1472
        return ''.join(self.get_file_lines(file_id))
 
1473
 
 
1474
    def get_reference_revision(self, file_id, path=None):
 
1475
        return self.inventory[file_id].reference_revision
 
1476
 
 
1477
    def get_symlink_target(self, file_id):
 
1478
        entry = self._get_entry(file_id=file_id)
 
1479
        parent_index = self._get_parent_index()
 
1480
        if entry[1][parent_index][0] != 'l':
 
1481
            return None
 
1482
        else:
 
1483
            # At present, none of the tree implementations supports non-ascii
 
1484
            # symlink targets. So we will just assume that the dirstate path is
 
1485
            # correct.
 
1486
            return entry[1][parent_index][1]
 
1487
 
 
1488
    def get_revision_id(self):
 
1489
        """Return the revision id for this tree."""
 
1490
        return self._revision_id
 
1491
 
 
1492
    def _get_inventory(self):
 
1493
        if self._inventory is not None:
 
1494
            return self._inventory
 
1495
        self._must_be_locked()
 
1496
        self._generate_inventory()
 
1497
        return self._inventory
 
1498
 
 
1499
    inventory = property(_get_inventory,
 
1500
                         doc="Inventory of this Tree")
 
1501
 
 
1502
    def get_parent_ids(self):
 
1503
        """The parents of a tree in the dirstate are not cached."""
 
1504
        return self._repository.get_revision(self._revision_id).parent_ids
 
1505
 
 
1506
    def has_filename(self, filename):
 
1507
        return bool(self.path2id(filename))
 
1508
 
 
1509
    def kind(self, file_id):
 
1510
        return self.inventory[file_id].kind
 
1511
 
 
1512
    def is_executable(self, file_id, path=None):
 
1513
        ie = self.inventory[file_id]
 
1514
        if ie.kind != "file":
 
1515
            return None
 
1516
        return ie.executable
 
1517
 
 
1518
    def list_files(self, include_root=False):
 
1519
        # We use a standard implementation, because DirStateRevisionTree is
 
1520
        # dealing with one of the parents of the current state
 
1521
        inv = self._get_inventory()
 
1522
        entries = inv.iter_entries()
 
1523
        if self.inventory.root is not None and not include_root:
 
1524
            entries.next()
 
1525
        for path, entry in entries:
 
1526
            yield path, 'V', entry.kind, entry.file_id, entry
 
1527
 
 
1528
    def lock_read(self):
 
1529
        """Lock the tree for a set of operations."""
 
1530
        if not self._locked:
 
1531
            self._repository.lock_read()
 
1532
            if self._dirstate._lock_token is None:
 
1533
                self._dirstate.lock_read()
 
1534
                self._dirstate_locked = True
 
1535
        self._locked += 1
 
1536
 
 
1537
    def _must_be_locked(self):
 
1538
        if not self._locked:
 
1539
            raise errors.ObjectNotLocked(self)
 
1540
 
 
1541
    @needs_read_lock
 
1542
    def path2id(self, path):
 
1543
        """Return the id for path in this tree."""
 
1544
        # lookup by path: faster than splitting and walking the ivnentory.
 
1545
        entry = self._get_entry(path=path)
 
1546
        if entry == (None, None):
 
1547
            return None
 
1548
        return entry[0][2]
 
1549
 
 
1550
    def unlock(self):
 
1551
        """Unlock, freeing any cache memory used during the lock."""
 
1552
        # outside of a lock, the inventory is suspect: release it.
 
1553
        self._locked -=1
 
1554
        if not self._locked:
 
1555
            self._inventory = None
 
1556
            self._locked = 0
 
1557
            if self._dirstate_locked:
 
1558
                self._dirstate.unlock()
 
1559
                self._dirstate_locked = False
 
1560
            self._repository.unlock()
 
1561
 
 
1562
    def walkdirs(self, prefix=""):
 
1563
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1564
        # implementation based on an inventory.  
 
1565
        # This should be cleaned up to use the much faster Dirstate code
 
1566
        # So for now, we just build up the parent inventory, and extract
 
1567
        # it the same way RevisionTree does.
 
1568
        _directory = 'directory'
 
1569
        inv = self._get_inventory()
 
1570
        top_id = inv.path2id(prefix)
 
1571
        if top_id is None:
 
1572
            pending = []
 
1573
        else:
 
1574
            pending = [(prefix, top_id)]
 
1575
        while pending:
 
1576
            dirblock = []
 
1577
            relpath, file_id = pending.pop()
 
1578
            # 0 - relpath, 1- file-id
 
1579
            if relpath:
 
1580
                relroot = relpath + '/'
 
1581
            else:
 
1582
                relroot = ""
 
1583
            # FIXME: stash the node in pending
 
1584
            entry = inv[file_id]
 
1585
            for name, child in entry.sorted_children():
 
1586
                toppath = relroot + name
 
1587
                dirblock.append((toppath, name, child.kind, None,
 
1588
                    child.file_id, child.kind
 
1589
                    ))
 
1590
            yield (relpath, entry.file_id), dirblock
 
1591
            # push the user specified dirs from dirblock
 
1592
            for dir in reversed(dirblock):
 
1593
                if dir[2] == _directory:
 
1594
                    pending.append((dir[0], dir[4]))
 
1595
 
 
1596
 
 
1597
class InterDirStateTree(InterTree):
 
1598
    """Fast path optimiser for changes_from with dirstate trees.
 
1599
    
 
1600
    This is used only when both trees are in the dirstate working file, and 
 
1601
    the source is any parent within the dirstate, and the destination is 
 
1602
    the current working tree of the same dirstate.
 
1603
    """
 
1604
    # this could be generalized to allow comparisons between any trees in the
 
1605
    # dirstate, and possibly between trees stored in different dirstates.
 
1606
 
 
1607
    def __init__(self, source, target):
 
1608
        super(InterDirStateTree, self).__init__(source, target)
 
1609
        if not InterDirStateTree.is_compatible(source, target):
 
1610
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1611
 
 
1612
    @staticmethod
 
1613
    def make_source_parent_tree(source, target):
 
1614
        """Change the source tree into a parent of the target."""
 
1615
        revid = source.commit('record tree')
 
1616
        target.branch.repository.fetch(source.branch.repository, revid)
 
1617
        target.set_parent_ids([revid])
 
1618
        return target.basis_tree(), target
 
1619
 
 
1620
    _matching_from_tree_format = WorkingTreeFormat4()
 
1621
    _matching_to_tree_format = WorkingTreeFormat4()
 
1622
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1623
 
 
1624
    def _iter_changes(self, include_unchanged=False,
 
1625
                      specific_files=None, pb=None, extra_trees=[],
 
1626
                      require_versioned=True, want_unversioned=False):
 
1627
        """Return the changes from source to target.
 
1628
 
 
1629
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1630
            for details.
 
1631
        :param specific_files: An optional list of file paths to restrict the
 
1632
            comparison to. When mapping filenames to ids, all matches in all
 
1633
            trees (including optional extra_trees) are used, and all children of
 
1634
            matched directories are included.
 
1635
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1636
            unchanged entries in the result.
 
1637
        :param extra_trees: An optional list of additional trees to use when
 
1638
            mapping the contents of specific_files (paths) to file_ids.
 
1639
        :param require_versioned: If True, all files in specific_files must be
 
1640
            versioned in one of source, target, extra_trees or
 
1641
            PathsNotVersionedError is raised.
 
1642
        :param want_unversioned: Should unversioned files be returned in the
 
1643
            output. An unversioned file is defined as one with (False, False)
 
1644
            for the versioned pair.
 
1645
        """
 
1646
        utf8_decode_or_none = cache_utf8._utf8_decode_with_None
 
1647
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1648
        # NB: show_status depends on being able to pass in non-versioned files
 
1649
        # and report them as unknown
 
1650
        # TODO: handle extra trees in the dirstate.
 
1651
        # TODO: handle comparisons as an empty tree as a different special
 
1652
        # case? mbp 20070226
 
1653
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1654
            # we can't fast-path these cases (yet)
 
1655
            for f in super(InterDirStateTree, self)._iter_changes(
 
1656
                include_unchanged, specific_files, pb, extra_trees,
 
1657
                require_versioned, want_unversioned=want_unversioned):
 
1658
                yield f
 
1659
            return
 
1660
        parent_ids = self.target.get_parent_ids()
 
1661
        assert (self.source._revision_id in parent_ids), \
 
1662
                "revision {%s} is not stored in {%s}, but %s " \
 
1663
                "can only be used for trees stored in the dirstate" \
 
1664
                % (self.source._revision_id, self.target, self._iter_changes)
 
1665
        target_index = 0
 
1666
        if self.source._revision_id == NULL_REVISION:
 
1667
            source_index = None
 
1668
            indices = (target_index,)
 
1669
        else:
 
1670
            assert (self.source._revision_id in parent_ids), \
 
1671
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1672
                self.source._revision_id, parent_ids)
 
1673
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1674
            indices = (source_index,target_index)
 
1675
        # -- make all specific_files utf8 --
 
1676
        if specific_files:
 
1677
            specific_files_utf8 = set()
 
1678
            for path in specific_files:
 
1679
                specific_files_utf8.add(path.encode('utf8'))
 
1680
            specific_files = specific_files_utf8
 
1681
        else:
 
1682
            specific_files = set([''])
 
1683
        # -- specific_files is now a utf8 path set --
 
1684
        # -- get the state object and prepare it.
 
1685
        state = self.target.current_dirstate()
 
1686
        state._read_dirblocks_if_needed()
 
1687
        def _entries_for_path(path):
 
1688
            """Return a list with all the entries that match path for all ids.
 
1689
            """
 
1690
            dirname, basename = os.path.split(path)
 
1691
            key = (dirname, basename, '')
 
1692
            block_index, present = state._find_block_index_from_key(key)
 
1693
            if not present:
 
1694
                # the block which should contain path is absent.
 
1695
                return []
 
1696
            result = []
 
1697
            block = state._dirblocks[block_index][1]
 
1698
            entry_index, _ = state._find_entry_index(key, block)
 
1699
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1700
            while (entry_index < len(block) and
 
1701
                block[entry_index][0][0:2] == key[0:2]):
 
1702
                result.append(block[entry_index])
 
1703
                entry_index += 1
 
1704
            return result
 
1705
        if require_versioned:
 
1706
            # -- check all supplied paths are versioned in a search tree. --
 
1707
            all_versioned = True
 
1708
            for path in specific_files:
 
1709
                path_entries = _entries_for_path(path)
 
1710
                if not path_entries:
 
1711
                    # this specified path is not present at all: error
 
1712
                    all_versioned = False
 
1713
                    break
 
1714
                found_versioned = False
 
1715
                # for each id at this path
 
1716
                for entry in path_entries:
 
1717
                    # for each tree.
 
1718
                    for index in indices:
 
1719
                        if entry[1][index][0] != 'a': # absent
 
1720
                            found_versioned = True
 
1721
                            # all good: found a versioned cell
 
1722
                            break
 
1723
                if not found_versioned:
 
1724
                    # none of the indexes was not 'absent' at all ids for this
 
1725
                    # path.
 
1726
                    all_versioned = False
 
1727
                    break
 
1728
            if not all_versioned:
 
1729
                raise errors.PathsNotVersionedError(specific_files)
 
1730
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1731
        search_specific_files = set()
 
1732
        for path in specific_files:
 
1733
            other_specific_files = specific_files.difference(set([path]))
 
1734
            if not osutils.is_inside_any(other_specific_files, path):
 
1735
                # this is a top level path, we must check it.
 
1736
                search_specific_files.add(path)
 
1737
        # sketch: 
 
1738
        # compare source_index and target_index at or under each element of search_specific_files.
 
1739
        # follow the following comparison table. Note that we only want to do diff operations when
 
1740
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1741
        # for the target.
 
1742
        # cases:
 
1743
        # 
 
1744
        # Source | Target | disk | action
 
1745
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1746
        #        |        |      | diff check on source-target
 
1747
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1748
        #        |        |      | ???
 
1749
        #   r    |  a     |      | add source to search
 
1750
        #   r    |  a     |  a   | 
 
1751
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1752
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1753
        #   a    | fdlt   |      | add new id
 
1754
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1755
        #   a    |  a     |      | not present in either tree, skip
 
1756
        #   a    |  a     |  a   | not present in any tree, skip
 
1757
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1758
        #        |        |      | may not be selected by the users list of paths.
 
1759
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1760
        #        |        |      | may not be selected by the users list of paths.
 
1761
        #  fdlt  | fdlt   |      | content in both: diff them
 
1762
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1763
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1764
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1765
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1766
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1767
        #        |        |      | this id at the other path.
 
1768
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1769
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1770
        #        |        |      | this id at the other path.
 
1771
 
 
1772
        # for all search_indexs in each path at or under each element of
 
1773
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1774
        # relocated path as one to search if its not searched already. If the
 
1775
        # detail is not relocated, add the id.
 
1776
        searched_specific_files = set()
 
1777
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1778
        # Using a list so that we can access the values and change them in
 
1779
        # nested scope. Each one is [path, file_id, entry]
 
1780
        last_source_parent = [None, None, None]
 
1781
        last_target_parent = [None, None, None]
 
1782
 
 
1783
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1784
 
 
1785
        def _process_entry(entry, path_info):
 
1786
            """Compare an entry and real disk to generate delta information.
 
1787
 
 
1788
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1789
                the path of entry. If None, then the path is considered absent.
 
1790
                (Perhaps we should pass in a concrete entry for this ?)
 
1791
                Basename is returned as a utf8 string because we expect this
 
1792
                tuple will be ignored, and don't want to take the time to
 
1793
                decode.
 
1794
            """
 
1795
            if source_index is None:
 
1796
                source_details = NULL_PARENT_DETAILS
 
1797
            else:
 
1798
                source_details = entry[1][source_index]
 
1799
            target_details = entry[1][target_index]
 
1800
            target_minikind = target_details[0]
 
1801
            if path_info is not None and target_minikind in 'fdlt':
 
1802
                assert target_index == 0
 
1803
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1804
                                                  stat_value=path_info[3])
 
1805
                # The entry may have been modified by update_entry
 
1806
                target_details = entry[1][target_index]
 
1807
                target_minikind = target_details[0]
 
1808
            else:
 
1809
                link_or_sha1 = None
 
1810
            source_minikind = source_details[0]
 
1811
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1812
                # claimed content in both: diff
 
1813
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1814
                #        |        |      | diff check on source-target
 
1815
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1816
                #        |        |      | ???
 
1817
                if source_minikind in 'r':
 
1818
                    # add the source to the search path to find any children it
 
1819
                    # has.  TODO ? : only add if it is a container ?
 
1820
                    if not osutils.is_inside_any(searched_specific_files,
 
1821
                                                 source_details[1]):
 
1822
                        search_specific_files.add(source_details[1])
 
1823
                    # generate the old path; this is needed for stating later
 
1824
                    # as well.
 
1825
                    old_path = source_details[1]
 
1826
                    old_dirname, old_basename = os.path.split(old_path)
 
1827
                    path = pathjoin(entry[0][0], entry[0][1])
 
1828
                    old_entry = state._get_entry(source_index,
 
1829
                                                 path_utf8=old_path)
 
1830
                    # update the source details variable to be the real
 
1831
                    # location.
 
1832
                    source_details = old_entry[1][source_index]
 
1833
                    source_minikind = source_details[0]
 
1834
                else:
 
1835
                    old_dirname = entry[0][0]
 
1836
                    old_basename = entry[0][1]
 
1837
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1838
                if path_info is None:
 
1839
                    # the file is missing on disk, show as removed.
 
1840
                    content_change = True
 
1841
                    target_kind = None
 
1842
                    target_exec = False
 
1843
                else:
 
1844
                    # source and target are both versioned and disk file is present.
 
1845
                    target_kind = path_info[2]
 
1846
                    if target_kind == 'directory':
 
1847
                        if source_minikind != 'd':
 
1848
                            content_change = True
 
1849
                        else:
 
1850
                            # directories have no fingerprint
 
1851
                            content_change = False
 
1852
                        target_exec = False
 
1853
                    elif target_kind == 'file':
 
1854
                        if source_minikind != 'f':
 
1855
                            content_change = True
 
1856
                        else:
 
1857
                            # We could check the size, but we already have the
 
1858
                            # sha1 hash.
 
1859
                            content_change = (link_or_sha1 != source_details[1])
 
1860
                        # Target details is updated at update_entry time
 
1861
                        if use_filesystem_for_exec:
 
1862
                            # We don't need S_ISREG here, because we are sure
 
1863
                            # we are dealing with a file.
 
1864
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1865
                        else:
 
1866
                            target_exec = target_details[3]
 
1867
                    elif target_kind == 'symlink':
 
1868
                        if source_minikind != 'l':
 
1869
                            content_change = True
 
1870
                        else:
 
1871
                            content_change = (link_or_sha1 != source_details[1])
 
1872
                        target_exec = False
 
1873
                    elif target_kind == 'tree-reference':
 
1874
                        if source_minikind != 't':
 
1875
                            content_change = True
 
1876
                        else:
 
1877
                            content_change = False
 
1878
                        target_exec = False
 
1879
                    else:
 
1880
                        raise Exception, "unknown kind %s" % path_info[2]
 
1881
                # parent id is the entry for the path in the target tree
 
1882
                if old_dirname == last_source_parent[0]:
 
1883
                    source_parent_id = last_source_parent[1]
 
1884
                else:
 
1885
                    source_parent_entry = state._get_entry(source_index,
 
1886
                                                           path_utf8=old_dirname)
 
1887
                    source_parent_id = source_parent_entry[0][2]
 
1888
                    if source_parent_id == entry[0][2]:
 
1889
                        # This is the root, so the parent is None
 
1890
                        source_parent_id = None
 
1891
                    else:
 
1892
                        last_source_parent[0] = old_dirname
 
1893
                        last_source_parent[1] = source_parent_id
 
1894
                        last_source_parent[2] = source_parent_entry
 
1895
                new_dirname = entry[0][0]
 
1896
                if new_dirname == last_target_parent[0]:
 
1897
                    target_parent_id = last_target_parent[1]
 
1898
                else:
 
1899
                    # TODO: We don't always need to do the lookup, because the
 
1900
                    #       parent entry will be the same as the source entry.
 
1901
                    target_parent_entry = state._get_entry(target_index,
 
1902
                                                           path_utf8=new_dirname)
 
1903
                    target_parent_id = target_parent_entry[0][2]
 
1904
                    if target_parent_id == entry[0][2]:
 
1905
                        # This is the root, so the parent is None
 
1906
                        target_parent_id = None
 
1907
                    else:
 
1908
                        last_target_parent[0] = new_dirname
 
1909
                        last_target_parent[1] = target_parent_id
 
1910
                        last_target_parent[2] = target_parent_entry
 
1911
 
 
1912
                source_exec = source_details[3]
 
1913
                return ((entry[0][2], (old_path, path), content_change,
 
1914
                        (True, True),
 
1915
                        (source_parent_id, target_parent_id),
 
1916
                        (old_basename, entry[0][1]),
 
1917
                        (_minikind_to_kind[source_minikind], target_kind),
 
1918
                        (source_exec, target_exec)),)
 
1919
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1920
                # looks like a new file
 
1921
                if path_info is not None:
 
1922
                    path = pathjoin(entry[0][0], entry[0][1])
 
1923
                    # parent id is the entry for the path in the target tree
 
1924
                    # TODO: these are the same for an entire directory: cache em.
 
1925
                    parent_id = state._get_entry(target_index,
 
1926
                                                 path_utf8=entry[0][0])[0][2]
 
1927
                    if parent_id == entry[0][2]:
 
1928
                        parent_id = None
 
1929
                    if use_filesystem_for_exec:
 
1930
                        # We need S_ISREG here, because we aren't sure if this
 
1931
                        # is a file or not.
 
1932
                        target_exec = bool(
 
1933
                            stat.S_ISREG(path_info[3].st_mode)
 
1934
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1935
                    else:
 
1936
                        target_exec = target_details[3]
 
1937
                    return ((entry[0][2], (None, path), True,
 
1938
                            (False, True),
 
1939
                            (None, parent_id),
 
1940
                            (None, entry[0][1]),
 
1941
                            (None, path_info[2]),
 
1942
                            (None, target_exec)),)
 
1943
                else:
 
1944
                    # but its not on disk: we deliberately treat this as just
 
1945
                    # never-present. (Why ?! - RBC 20070224)
 
1946
                    pass
 
1947
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1948
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1949
                # if its still on disk, *and* theres no other entry at this
 
1950
                # path [we dont know this in this routine at the moment -
 
1951
                # perhaps we should change this - then it would be an unknown.
 
1952
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1953
                # parent id is the entry for the path in the target tree
 
1954
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1955
                if parent_id == entry[0][2]:
 
1956
                    parent_id = None
 
1957
                return ((entry[0][2], (old_path, None), True,
 
1958
                        (True, False),
 
1959
                        (parent_id, None),
 
1960
                        (entry[0][1], None),
 
1961
                        (_minikind_to_kind[source_minikind], None),
 
1962
                        (source_details[3], None)),)
 
1963
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1964
                # a rename; could be a true rename, or a rename inherited from
 
1965
                # a renamed parent. TODO: handle this efficiently. Its not
 
1966
                # common case to rename dirs though, so a correct but slow
 
1967
                # implementation will do.
 
1968
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1969
                    search_specific_files.add(target_details[1])
 
1970
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
1971
                # neither of the selected trees contain this file,
 
1972
                # so skip over it. This is not currently directly tested, but
 
1973
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1974
                pass
 
1975
            else:
 
1976
                raise AssertionError("don't know how to compare "
 
1977
                    "source_minikind=%r, target_minikind=%r"
 
1978
                    % (source_minikind, target_minikind))
 
1979
                ## import pdb;pdb.set_trace()
 
1980
            return ()
 
1981
 
 
1982
        while search_specific_files:
 
1983
            # TODO: the pending list should be lexically sorted?  the
 
1984
            # interface doesn't require it.
 
1985
            current_root = search_specific_files.pop()
 
1986
            current_root_unicode = current_root.decode('utf8')
 
1987
            searched_specific_files.add(current_root)
 
1988
            # process the entries for this containing directory: the rest will be
 
1989
            # found by their parents recursively.
 
1990
            root_entries = _entries_for_path(current_root)
 
1991
            root_abspath = self.target.abspath(current_root_unicode)
 
1992
            try:
 
1993
                root_stat = os.lstat(root_abspath)
 
1994
            except OSError, e:
 
1995
                if e.errno == errno.ENOENT:
 
1996
                    # the path does not exist: let _process_entry know that.
 
1997
                    root_dir_info = None
 
1998
                else:
 
1999
                    # some other random error: hand it up.
 
2000
                    raise
 
2001
            else:
 
2002
                root_dir_info = ('', current_root,
 
2003
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2004
                    root_abspath)
 
2005
                if root_dir_info[2] == 'directory':
 
2006
                    if self.target._directory_is_tree_reference(
 
2007
                        current_root.decode('utf8')):
 
2008
                        root_dir_info = root_dir_info[:2] + \
 
2009
                            ('tree-reference',) + root_dir_info[3:]
 
2010
 
 
2011
            if not root_entries and not root_dir_info:
 
2012
                # this specified path is not present at all, skip it.
 
2013
                continue
 
2014
            path_handled = False
 
2015
            for entry in root_entries:
 
2016
                for result in _process_entry(entry, root_dir_info):
 
2017
                    # this check should probably be outside the loop: one
 
2018
                    # 'iterate two trees' api, and then _iter_changes filters
 
2019
                    # unchanged pairs. - RBC 20070226
 
2020
                    path_handled = True
 
2021
                    if (include_unchanged
 
2022
                        or result[2]                    # content change
 
2023
                        or result[3][0] != result[3][1] # versioned status
 
2024
                        or result[4][0] != result[4][1] # parent id
 
2025
                        or result[5][0] != result[5][1] # name
 
2026
                        or result[6][0] != result[6][1] # kind
 
2027
                        or result[7][0] != result[7][1] # executable
 
2028
                        ):
 
2029
                        yield (result[0],
 
2030
                               (utf8_decode_or_none(result[1][0]),
 
2031
                                utf8_decode_or_none(result[1][1])),
 
2032
                               result[2],
 
2033
                               result[3],
 
2034
                               result[4],
 
2035
                               (utf8_decode_or_none(result[5][0]),
 
2036
                                utf8_decode_or_none(result[5][1])),
 
2037
                               result[6],
 
2038
                               result[7],
 
2039
                              )
 
2040
            if want_unversioned and not path_handled and root_dir_info:
 
2041
                new_executable = bool(
 
2042
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2043
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2044
                yield (None,
 
2045
                       (None, current_root_unicode),
 
2046
                       True,
 
2047
                       (False, False),
 
2048
                       (None, None),
 
2049
                       (None, splitpath(current_root_unicode)[-1]),
 
2050
                       (None, root_dir_info[2]),
 
2051
                       (None, new_executable)
 
2052
                      )
 
2053
            initial_key = (current_root, '', '')
 
2054
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2055
            if block_index == 0:
 
2056
                # we have processed the total root already, but because the
 
2057
                # initial key matched it we should skip it here.
 
2058
                block_index +=1
 
2059
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2060
                current_dir_info = None
 
2061
            else:
 
2062
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2063
                try:
 
2064
                    current_dir_info = dir_iterator.next()
 
2065
                except OSError, e:
 
2066
                    if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
2067
                        # there may be directories in the inventory even though
 
2068
                        # this path is not a file on disk: so mark it as end of
 
2069
                        # iterator
 
2070
                        current_dir_info = None
 
2071
                    else:
 
2072
                        raise
 
2073
                else:
 
2074
                    if current_dir_info[0][0] == '':
 
2075
                        # remove .bzr from iteration
 
2076
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2077
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2078
                        del current_dir_info[1][bzr_index]
 
2079
            # walk until both the directory listing and the versioned metadata
 
2080
            # are exhausted. 
 
2081
            if (block_index < len(state._dirblocks) and
 
2082
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2083
                current_block = state._dirblocks[block_index]
 
2084
            else:
 
2085
                current_block = None
 
2086
            while (current_dir_info is not None or
 
2087
                   current_block is not None):
 
2088
                if (current_dir_info and current_block
 
2089
                    and current_dir_info[0][0] != current_block[0]):
 
2090
                    if current_dir_info[0][0] < current_block[0] :
 
2091
                        # filesystem data refers to paths not covered by the dirblock.
 
2092
                        # this has two possibilities:
 
2093
                        # A) it is versioned but empty, so there is no block for it
 
2094
                        # B) it is not versioned.
 
2095
                        # in either case it was processed by the containing directories walk:
 
2096
                        # if it is root/foo, when we walked root we emitted it,
 
2097
                        # or if we ere given root/foo to walk specifically, we
 
2098
                        # emitted it when checking the walk-root entries
 
2099
                        # advance the iterator and loop - we dont need to emit it.
 
2100
                        try:
 
2101
                            current_dir_info = dir_iterator.next()
 
2102
                        except StopIteration:
 
2103
                            current_dir_info = None
 
2104
                    else:
 
2105
                        # We have a dirblock entry for this location, but there
 
2106
                        # is no filesystem path for this. This is most likely
 
2107
                        # because a directory was removed from the disk.
 
2108
                        # We don't have to report the missing directory,
 
2109
                        # because that should have already been handled, but we
 
2110
                        # need to handle all of the files that are contained
 
2111
                        # within.
 
2112
                        for current_entry in current_block[1]:
 
2113
                            # entry referring to file not present on disk.
 
2114
                            # advance the entry only, after processing.
 
2115
                            for result in _process_entry(current_entry, None):
 
2116
                                # this check should probably be outside the loop: one
 
2117
                                # 'iterate two trees' api, and then _iter_changes filters
 
2118
                                # unchanged pairs. - RBC 20070226
 
2119
                                if (include_unchanged
 
2120
                                    or result[2]                    # content change
 
2121
                                    or result[3][0] != result[3][1] # versioned status
 
2122
                                    or result[4][0] != result[4][1] # parent id
 
2123
                                    or result[5][0] != result[5][1] # name
 
2124
                                    or result[6][0] != result[6][1] # kind
 
2125
                                    or result[7][0] != result[7][1] # executable
 
2126
                                    ):
 
2127
                                    yield (result[0],
 
2128
                                           (utf8_decode_or_none(result[1][0]),
 
2129
                                            utf8_decode_or_none(result[1][1])),
 
2130
                                           result[2],
 
2131
                                           result[3],
 
2132
                                           result[4],
 
2133
                                           (utf8_decode_or_none(result[5][0]),
 
2134
                                            utf8_decode_or_none(result[5][1])),
 
2135
                                           result[6],
 
2136
                                           result[7],
 
2137
                                          )
 
2138
                        block_index +=1
 
2139
                        if (block_index < len(state._dirblocks) and
 
2140
                            osutils.is_inside(current_root,
 
2141
                                              state._dirblocks[block_index][0])):
 
2142
                            current_block = state._dirblocks[block_index]
 
2143
                        else:
 
2144
                            current_block = None
 
2145
                    continue
 
2146
                entry_index = 0
 
2147
                if current_block and entry_index < len(current_block[1]):
 
2148
                    current_entry = current_block[1][entry_index]
 
2149
                else:
 
2150
                    current_entry = None
 
2151
                advance_entry = True
 
2152
                path_index = 0
 
2153
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2154
                    current_path_info = current_dir_info[1][path_index]
 
2155
                    if current_path_info[2] == 'directory':
 
2156
                        if self.target._directory_is_tree_reference(
 
2157
                            current_path_info[0].decode('utf8')):
 
2158
                            current_path_info = current_path_info[:2] + \
 
2159
                                ('tree-reference',) + current_path_info[3:]
 
2160
                else:
 
2161
                    current_path_info = None
 
2162
                advance_path = True
 
2163
                path_handled = False
 
2164
                while (current_entry is not None or
 
2165
                    current_path_info is not None):
 
2166
                    if current_entry is None:
 
2167
                        # the check for path_handled when the path is adnvaced
 
2168
                        # will yield this path if needed.
 
2169
                        pass
 
2170
                    elif current_path_info is None:
 
2171
                        # no path is fine: the per entry code will handle it.
 
2172
                        for result in _process_entry(current_entry, current_path_info):
 
2173
                            # this check should probably be outside the loop: one
 
2174
                            # 'iterate two trees' api, and then _iter_changes filters
 
2175
                            # unchanged pairs. - RBC 20070226
 
2176
                            if (include_unchanged
 
2177
                                or result[2]                    # content change
 
2178
                                or result[3][0] != result[3][1] # versioned status
 
2179
                                or result[4][0] != result[4][1] # parent id
 
2180
                                or result[5][0] != result[5][1] # name
 
2181
                                or result[6][0] != result[6][1] # kind
 
2182
                                or result[7][0] != result[7][1] # executable
 
2183
                                ):
 
2184
                                yield (result[0],
 
2185
                                       (utf8_decode_or_none(result[1][0]),
 
2186
                                        utf8_decode_or_none(result[1][1])),
 
2187
                                       result[2],
 
2188
                                       result[3],
 
2189
                                       result[4],
 
2190
                                       (utf8_decode_or_none(result[5][0]),
 
2191
                                        utf8_decode_or_none(result[5][1])),
 
2192
                                       result[6],
 
2193
                                       result[7],
 
2194
                                      )
 
2195
                    elif current_entry[0][1] != current_path_info[1]:
 
2196
                        if current_path_info[1] < current_entry[0][1]:
 
2197
                            # extra file on disk: pass for now, but only
 
2198
                            # increment the path, not the entry
 
2199
                            advance_entry = False
 
2200
                        else:
 
2201
                            # entry referring to file not present on disk.
 
2202
                            # advance the entry only, after processing.
 
2203
                            for result in _process_entry(current_entry, None):
 
2204
                                # this check should probably be outside the loop: one
 
2205
                                # 'iterate two trees' api, and then _iter_changes filters
 
2206
                                # unchanged pairs. - RBC 20070226
 
2207
                                path_handled = True
 
2208
                                if (include_unchanged
 
2209
                                    or result[2]                    # content change
 
2210
                                    or result[3][0] != result[3][1] # versioned status
 
2211
                                    or result[4][0] != result[4][1] # parent id
 
2212
                                    or result[5][0] != result[5][1] # name
 
2213
                                    or result[6][0] != result[6][1] # kind
 
2214
                                    or result[7][0] != result[7][1] # executable
 
2215
                                    ):
 
2216
                                    yield (result[0],
 
2217
                                           (utf8_decode_or_none(result[1][0]),
 
2218
                                            utf8_decode_or_none(result[1][1])),
 
2219
                                           result[2],
 
2220
                                           result[3],
 
2221
                                           result[4],
 
2222
                                           (utf8_decode_or_none(result[5][0]),
 
2223
                                            utf8_decode_or_none(result[5][1])),
 
2224
                                           result[6],
 
2225
                                           result[7],
 
2226
                                          )
 
2227
                            advance_path = False
 
2228
                    else:
 
2229
                        for result in _process_entry(current_entry, current_path_info):
 
2230
                            # this check should probably be outside the loop: one
 
2231
                            # 'iterate two trees' api, and then _iter_changes filters
 
2232
                            # unchanged pairs. - RBC 20070226
 
2233
                            path_handled = True
 
2234
                            if (include_unchanged
 
2235
                                or result[2]                    # content change
 
2236
                                or result[3][0] != result[3][1] # versioned status
 
2237
                                or result[4][0] != result[4][1] # parent id
 
2238
                                or result[5][0] != result[5][1] # name
 
2239
                                or result[6][0] != result[6][1] # kind
 
2240
                                or result[7][0] != result[7][1] # executable
 
2241
                                ):
 
2242
                                yield (result[0],
 
2243
                                       (utf8_decode_or_none(result[1][0]),
 
2244
                                        utf8_decode_or_none(result[1][1])),
 
2245
                                       result[2],
 
2246
                                       result[3],
 
2247
                                       result[4],
 
2248
                                       (utf8_decode_or_none(result[5][0]),
 
2249
                                        utf8_decode_or_none(result[5][1])),
 
2250
                                       result[6],
 
2251
                                       result[7],
 
2252
                                      )
 
2253
                    if advance_entry and current_entry is not None:
 
2254
                        entry_index += 1
 
2255
                        if entry_index < len(current_block[1]):
 
2256
                            current_entry = current_block[1][entry_index]
 
2257
                        else:
 
2258
                            current_entry = None
 
2259
                    else:
 
2260
                        advance_entry = True # reset the advance flaga
 
2261
                    if advance_path and current_path_info is not None:
 
2262
                        if not path_handled:
 
2263
                            # unversioned in all regards
 
2264
                            if want_unversioned:
 
2265
                                new_executable = bool(
 
2266
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2267
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2268
                                if want_unversioned:
 
2269
                                    yield (None,
 
2270
                                        (None, utf8_decode_or_none(current_path_info[0])),
 
2271
                                        True,
 
2272
                                        (False, False),
 
2273
                                        (None, None),
 
2274
                                        (None, utf8_decode_or_none(current_path_info[1])),
 
2275
                                        (None, current_path_info[2]),
 
2276
                                        (None, new_executable))
 
2277
                            # dont descend into this unversioned path if it is
 
2278
                            # a dir
 
2279
                            if current_path_info[2] in ('directory'):
 
2280
                                del current_dir_info[1][path_index]
 
2281
                                path_index -= 1
 
2282
                        # dont descend the disk iterator into any tree 
 
2283
                        # paths.
 
2284
                        if current_path_info[2] == 'tree-reference':
 
2285
                            del current_dir_info[1][path_index]
 
2286
                            path_index -= 1
 
2287
                        path_index += 1
 
2288
                        if path_index < len(current_dir_info[1]):
 
2289
                            current_path_info = current_dir_info[1][path_index]
 
2290
                            if current_path_info[2] == 'directory':
 
2291
                                if self.target._directory_is_tree_reference(
 
2292
                                    current_path_info[0].decode('utf8')):
 
2293
                                    current_path_info = current_path_info[:2] + \
 
2294
                                        ('tree-reference',) + current_path_info[3:]
 
2295
                        else:
 
2296
                            current_path_info = None
 
2297
                        path_handled = False
 
2298
                    else:
 
2299
                        advance_path = True # reset the advance flagg.
 
2300
                if current_block is not None:
 
2301
                    block_index += 1
 
2302
                    if (block_index < len(state._dirblocks) and
 
2303
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2304
                        current_block = state._dirblocks[block_index]
 
2305
                    else:
 
2306
                        current_block = None
 
2307
                if current_dir_info is not None:
 
2308
                    try:
 
2309
                        current_dir_info = dir_iterator.next()
 
2310
                    except StopIteration:
 
2311
                        current_dir_info = None
 
2312
 
 
2313
 
 
2314
    @staticmethod
 
2315
    def is_compatible(source, target):
 
2316
        # the target must be a dirstate working tree
 
2317
        if not isinstance(target, WorkingTree4):
 
2318
            return False
 
2319
        # the source must be a revtreee or dirstate rev tree.
 
2320
        if not isinstance(source,
 
2321
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2322
            return False
 
2323
        # the source revid must be in the target dirstate
 
2324
        if not (source._revision_id == NULL_REVISION or
 
2325
            source._revision_id in target.get_parent_ids()):
 
2326
            # TODO: what about ghosts? it may well need to 
 
2327
            # check for them explicitly.
 
2328
            return False
 
2329
        return True
 
2330
 
 
2331
InterTree.register_optimiser(InterDirStateTree)
 
2332
 
 
2333
 
 
2334
class Converter3to4(object):
 
2335
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2336
 
 
2337
    def __init__(self):
 
2338
        self.target_format = WorkingTreeFormat4()
 
2339
 
 
2340
    def convert(self, tree):
 
2341
        # lock the control files not the tree, so that we dont get tree
 
2342
        # on-unlock behaviours, and so that noone else diddles with the 
 
2343
        # tree during upgrade.
 
2344
        tree._control_files.lock_write()
 
2345
        try:
 
2346
            self.create_dirstate_data(tree)
 
2347
            self.update_format(tree)
 
2348
            self.remove_xml_files(tree)
 
2349
        finally:
 
2350
            tree._control_files.unlock()
 
2351
 
 
2352
    def create_dirstate_data(self, tree):
 
2353
        """Create the dirstate based data for tree."""
 
2354
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2355
            ).local_abspath('dirstate')
 
2356
        state = dirstate.DirState.from_tree(tree, local_path)
 
2357
        state.save()
 
2358
        state.unlock()
 
2359
 
 
2360
    def remove_xml_files(self, tree):
 
2361
        """Remove the oldformat 3 data."""
 
2362
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2363
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2364
            'pending-merges', 'stat-cache']:
 
2365
            try:
 
2366
                transport.delete(path)
 
2367
            except errors.NoSuchFile:
 
2368
                # some files are optional - just deal.
 
2369
                pass
 
2370
 
 
2371
    def update_format(self, tree):
 
2372
        """Change the format marker."""
 
2373
        tree._control_files.put_utf8('format',
 
2374
            self.target_format.get_format_string())