~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Vincent Ladeuil
  • Date: 2007-05-25 17:12:52 UTC
  • mto: (2485.8.44 bzr.connection.sharing)
  • mto: This revision was merged to the branch mainline in revision 2646.
  • Revision ID: v.ladeuil+lp@free.fr-20070525171252-hsooebpkqtb3j6ef
Add test for cat and missing.

* bzrlib/tests/commands/__init__.py:
(test_suite): Add test_cat and test_missing.

* bzrlib/tests/commands/test_cat.py: 
New file.

* bzrlib/tests/commands/test_missing.py: 
New file.

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
 
47
    dirstate,
 
48
    errors,
 
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revisiontree,
 
56
    textui,
 
57
    transform,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
 
61
    )
 
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
 
64
import bzrlib.ui
 
65
""")
 
66
 
 
67
from bzrlib import symbol_versioning
 
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
 
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
 
73
from bzrlib.mutabletree import needs_tree_write_lock
 
74
from bzrlib.osutils import (
 
75
    file_kind,
 
76
    isdir,
 
77
    normpath,
 
78
    pathjoin,
 
79
    rand_chars,
 
80
    realpath,
 
81
    safe_unicode,
 
82
    splitpath,
 
83
    )
 
84
from bzrlib.trace import mutter, note
 
85
from bzrlib.transport.local import LocalTransport
 
86
from bzrlib.tree import InterTree
 
87
from bzrlib.progress import DummyProgress, ProgressPhase
 
88
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
89
from bzrlib.rio import RioReader, rio_file, Stanza
 
90
from bzrlib.symbol_versioning import (deprecated_passed,
 
91
        deprecated_method,
 
92
        deprecated_function,
 
93
        DEPRECATED_PARAMETER,
 
94
        )
 
95
from bzrlib.tree import Tree
 
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
97
 
 
98
 
 
99
# This is the Windows equivalent of ENOTDIR
 
100
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
101
# just an error code.
 
102
ERROR_PATH_NOT_FOUND = 3
 
103
ERROR_DIRECTORY = 267
 
104
 
 
105
 
 
106
class WorkingTree4(WorkingTree3):
 
107
    """This is the Format 4 working tree.
 
108
 
 
109
    This differs from WorkingTree3 by:
 
110
     - Having a consolidated internal dirstate, stored in a
 
111
       randomly-accessible sorted file on disk.
 
112
     - Not having a regular inventory attribute.  One can be synthesized 
 
113
       on demand but this is expensive and should be avoided.
 
114
 
 
115
    This is new in bzr 0.15.
 
116
    """
 
117
 
 
118
    def __init__(self, basedir,
 
119
                 branch,
 
120
                 _control_files=None,
 
121
                 _format=None,
 
122
                 _bzrdir=None):
 
123
        """Construct a WorkingTree for basedir.
 
124
 
 
125
        If the branch is not supplied, it is opened automatically.
 
126
        If the branch is supplied, it must be the branch for this basedir.
 
127
        (branch.base is not cross checked, because for remote branches that
 
128
        would be meaningless).
 
129
        """
 
130
        self._format = _format
 
131
        self.bzrdir = _bzrdir
 
132
        from bzrlib.trace import note, mutter
 
133
        assert isinstance(basedir, basestring), \
 
134
            "base directory %r is not a string" % basedir
 
135
        basedir = safe_unicode(basedir)
 
136
        mutter("opening working tree %r", basedir)
 
137
        self._branch = branch
 
138
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
139
            "branch %r is not a Branch" % self.branch
 
140
        self.basedir = realpath(basedir)
 
141
        # if branch is at our basedir and is a format 6 or less
 
142
        # assume all other formats have their own control files.
 
143
        assert isinstance(_control_files, LockableFiles), \
 
144
            "_control_files must be a LockableFiles, not %r" % _control_files
 
145
        self._control_files = _control_files
 
146
        self._dirty = None
 
147
        #-------------
 
148
        # during a read or write lock these objects are set, and are
 
149
        # None the rest of the time.
 
150
        self._dirstate = None
 
151
        self._inventory = None
 
152
        #-------------
 
153
 
 
154
    @needs_tree_write_lock
 
155
    def _add(self, files, ids, kinds):
 
156
        """See MutableTree._add."""
 
157
        state = self.current_dirstate()
 
158
        for f, file_id, kind in zip(files, ids, kinds):
 
159
            f = f.strip('/')
 
160
            assert '//' not in f
 
161
            assert '..' not in f
 
162
            if self.path2id(f):
 
163
                # special case tree root handling.
 
164
                if f == '' and self.path2id(f) == ROOT_ID:
 
165
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
166
                continue
 
167
            if file_id is None:
 
168
                file_id = generate_ids.gen_file_id(f)
 
169
            # deliberately add the file with no cached stat or sha1
 
170
            # - on the first access it will be gathered, and we can
 
171
            # always change this once tests are all passing.
 
172
            state.add(f, file_id, kind, None, '')
 
173
        self._make_dirty(reset_inventory=True)
 
174
 
 
175
    def _make_dirty(self, reset_inventory):
 
176
        """Make the tree state dirty.
 
177
 
 
178
        :param reset_inventory: True if the cached inventory should be removed
 
179
            (presuming there is one).
 
180
        """
 
181
        self._dirty = True
 
182
        if reset_inventory and self._inventory is not None:
 
183
            self._inventory = None
 
184
 
 
185
    @needs_tree_write_lock
 
186
    def add_reference(self, sub_tree):
 
187
        # use standard implementation, which calls back to self._add
 
188
        # 
 
189
        # So we don't store the reference_revision in the working dirstate,
 
190
        # it's just recorded at the moment of commit. 
 
191
        self._add_reference(sub_tree)
 
192
 
 
193
    def break_lock(self):
 
194
        """Break a lock if one is present from another instance.
 
195
 
 
196
        Uses the ui factory to ask for confirmation if the lock may be from
 
197
        an active process.
 
198
 
 
199
        This will probe the repository for its lock as well.
 
200
        """
 
201
        # if the dirstate is locked by an active process, reject the break lock
 
202
        # call.
 
203
        try:
 
204
            if self._dirstate is None:
 
205
                clear = True
 
206
            else:
 
207
                clear = False
 
208
            state = self._current_dirstate()
 
209
            if state._lock_token is not None:
 
210
                # we already have it locked. sheese, cant break our own lock.
 
211
                raise errors.LockActive(self.basedir)
 
212
            else:
 
213
                try:
 
214
                    # try for a write lock - need permission to get one anyhow
 
215
                    # to break locks.
 
216
                    state.lock_write()
 
217
                except errors.LockContention:
 
218
                    # oslocks fail when a process is still live: fail.
 
219
                    # TODO: get the locked lockdir info and give to the user to
 
220
                    # assist in debugging.
 
221
                    raise errors.LockActive(self.basedir)
 
222
                else:
 
223
                    state.unlock()
 
224
        finally:
 
225
            if clear:
 
226
                self._dirstate = None
 
227
        self._control_files.break_lock()
 
228
        self.branch.break_lock()
 
229
 
 
230
    def _comparison_data(self, entry, path):
 
231
        kind, executable, stat_value = \
 
232
            WorkingTree3._comparison_data(self, entry, path)
 
233
        # it looks like a plain directory, but it's really a reference -- see
 
234
        # also kind()
 
235
        if (self._repo_supports_tree_reference and
 
236
            kind == 'directory' and
 
237
            self._directory_is_tree_reference(path)):
 
238
            kind = 'tree-reference'
 
239
        return kind, executable, stat_value
 
240
 
 
241
    @needs_write_lock
 
242
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
243
        # mark the tree as dirty post commit - commit
 
244
        # can change the current versioned list by doing deletes.
 
245
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
246
        self._make_dirty(reset_inventory=True)
 
247
        return result
 
248
 
 
249
    def current_dirstate(self):
 
250
        """Return the current dirstate object.
 
251
 
 
252
        This is not part of the tree interface and only exposed for ease of
 
253
        testing.
 
254
 
 
255
        :raises errors.NotWriteLocked: when not in a lock.
 
256
        """
 
257
        self._must_be_locked()
 
258
        return self._current_dirstate()
 
259
 
 
260
    def _current_dirstate(self):
 
261
        """Internal function that does not check lock status.
 
262
 
 
263
        This is needed for break_lock which also needs the dirstate.
 
264
        """
 
265
        if self._dirstate is not None:
 
266
            return self._dirstate
 
267
        local_path = self.bzrdir.get_workingtree_transport(None
 
268
            ).local_abspath('dirstate')
 
269
        self._dirstate = dirstate.DirState.on_file(local_path)
 
270
        return self._dirstate
 
271
 
 
272
    def _directory_is_tree_reference(self, relpath):
 
273
        # as a special case, if a directory contains control files then 
 
274
        # it's a tree reference, except that the root of the tree is not
 
275
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
276
        # TODO: We could ask all the control formats whether they
 
277
        # recognize this directory, but at the moment there's no cheap api
 
278
        # to do that.  Since we probably can only nest bzr checkouts and
 
279
        # they always use this name it's ok for now.  -- mbp 20060306
 
280
        #
 
281
        # FIXME: There is an unhandled case here of a subdirectory
 
282
        # containing .bzr but not a branch; that will probably blow up
 
283
        # when you try to commit it.  It might happen if there is a
 
284
        # checkout in a subdirectory.  This can be avoided by not adding
 
285
        # it.  mbp 20070306
 
286
 
 
287
    def filter_unversioned_files(self, paths):
 
288
        """Filter out paths that are versioned.
 
289
 
 
290
        :return: set of paths.
 
291
        """
 
292
        # TODO: make a generic multi-bisect routine roughly that should list
 
293
        # the paths, then process one half at a time recursively, and feed the
 
294
        # results of each bisect in further still
 
295
        paths = sorted(paths)
 
296
        result = set()
 
297
        state = self.current_dirstate()
 
298
        # TODO we want a paths_to_dirblocks helper I think
 
299
        for path in paths:
 
300
            dirname, basename = os.path.split(path.encode('utf8'))
 
301
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
302
                dirname, basename, 0)
 
303
            if not path_is_versioned:
 
304
                result.add(path)
 
305
        return result
 
306
 
 
307
    def flush(self):
 
308
        """Write all cached data to disk."""
 
309
        if self._control_files._lock_mode != 'w':
 
310
            raise errors.NotWriteLocked(self)
 
311
        self.current_dirstate().save()
 
312
        self._inventory = None
 
313
        self._dirty = False
 
314
 
 
315
    @needs_tree_write_lock
 
316
    def _gather_kinds(self, files, kinds):
 
317
        """See MutableTree._gather_kinds."""
 
318
        for pos, f in enumerate(files):
 
319
            if kinds[pos] is None:
 
320
                kinds[pos] = self._kind(f)
 
321
 
 
322
    def _generate_inventory(self):
 
323
        """Create and set self.inventory from the dirstate object.
 
324
        
 
325
        This is relatively expensive: we have to walk the entire dirstate.
 
326
        Ideally we would not, and can deprecate this function.
 
327
        """
 
328
        #: uncomment to trap on inventory requests.
 
329
        # import pdb;pdb.set_trace()
 
330
        state = self.current_dirstate()
 
331
        state._read_dirblocks_if_needed()
 
332
        root_key, current_entry = self._get_entry(path='')
 
333
        current_id = root_key[2]
 
334
        assert current_entry[0][0] == 'd' # directory
 
335
        inv = Inventory(root_id=current_id)
 
336
        # Turn some things into local variables
 
337
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
338
        factory = entry_factory
 
339
        utf8_decode = cache_utf8._utf8_decode
 
340
        inv_byid = inv._byid
 
341
        # we could do this straight out of the dirstate; it might be fast
 
342
        # and should be profiled - RBC 20070216
 
343
        parent_ies = {'' : inv.root}
 
344
        for block in state._dirblocks[1:]: # skip the root
 
345
            dirname = block[0]
 
346
            try:
 
347
                parent_ie = parent_ies[dirname]
 
348
            except KeyError:
 
349
                # all the paths in this block are not versioned in this tree
 
350
                continue
 
351
            for key, entry in block[1]:
 
352
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
353
                if minikind in ('a', 'r'): # absent, relocated
 
354
                    # a parent tree only entry
 
355
                    continue
 
356
                name = key[1]
 
357
                name_unicode = utf8_decode(name)[0]
 
358
                file_id = key[2]
 
359
                kind = minikind_to_kind[minikind]
 
360
                inv_entry = factory[kind](file_id, name_unicode,
 
361
                                          parent_ie.file_id)
 
362
                if kind == 'file':
 
363
                    # This is only needed on win32, where this is the only way
 
364
                    # we know the executable bit.
 
365
                    inv_entry.executable = executable
 
366
                    # not strictly needed: working tree
 
367
                    #inv_entry.text_size = size
 
368
                    #inv_entry.text_sha1 = sha1
 
369
                elif kind == 'directory':
 
370
                    # add this entry to the parent map.
 
371
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
372
                elif kind == 'tree-reference':
 
373
                    assert self._repo_supports_tree_reference, \
 
374
                        "repository of %r " \
 
375
                        "doesn't support tree references " \
 
376
                        "required by entry %r" \
 
377
                        % (self, name)
 
378
                    inv_entry.reference_revision = link_or_sha1 or None
 
379
                elif kind != 'symlink':
 
380
                    raise AssertionError("unknown kind %r" % kind)
 
381
                # These checks cost us around 40ms on a 55k entry tree
 
382
                assert file_id not in inv_byid, ('file_id %s already in'
 
383
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
384
                assert name_unicode not in parent_ie.children
 
385
                inv_byid[file_id] = inv_entry
 
386
                parent_ie.children[name_unicode] = inv_entry
 
387
        self._inventory = inv
 
388
 
 
389
    def _get_entry(self, file_id=None, path=None):
 
390
        """Get the dirstate row for file_id or path.
 
391
 
 
392
        If either file_id or path is supplied, it is used as the key to lookup.
 
393
        If both are supplied, the fastest lookup is used, and an error is
 
394
        raised if they do not both point at the same row.
 
395
        
 
396
        :param file_id: An optional unicode file_id to be looked up.
 
397
        :param path: An optional unicode path to be looked up.
 
398
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
399
        """
 
400
        if file_id is None and path is None:
 
401
            raise errors.BzrError('must supply file_id or path')
 
402
        state = self.current_dirstate()
 
403
        if path is not None:
 
404
            path = path.encode('utf8')
 
405
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
406
 
 
407
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
408
        # check file id is valid unconditionally.
 
409
        entry = self._get_entry(file_id=file_id, path=path)
 
410
        assert entry[0] is not None, 'what error should this raise'
 
411
        if path is None:
 
412
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
413
 
 
414
        file_abspath = self.abspath(path)
 
415
        state = self.current_dirstate()
 
416
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
417
                                          stat_value=stat_value)
 
418
        if entry[1][0][0] == 'f':
 
419
            return link_or_sha1
 
420
        return None
 
421
 
 
422
    def _get_inventory(self):
 
423
        """Get the inventory for the tree. This is only valid within a lock."""
 
424
        if self._inventory is not None:
 
425
            return self._inventory
 
426
        self._must_be_locked()
 
427
        self._generate_inventory()
 
428
        return self._inventory
 
429
 
 
430
    inventory = property(_get_inventory,
 
431
                         doc="Inventory of this Tree")
 
432
 
 
433
    @needs_read_lock
 
434
    def get_parent_ids(self):
 
435
        """See Tree.get_parent_ids.
 
436
        
 
437
        This implementation requests the ids list from the dirstate file.
 
438
        """
 
439
        return self.current_dirstate().get_parent_ids()
 
440
 
 
441
    def get_reference_revision(self, file_id, path=None):
 
442
        # referenced tree's revision is whatever's currently there
 
443
        return self.get_nested_tree(file_id, path).last_revision()
 
444
 
 
445
    def get_nested_tree(self, file_id, path=None):
 
446
        if path is None:
 
447
            path = self.id2path(file_id)
 
448
        # else: check file_id is at path?
 
449
        return WorkingTree.open(self.abspath(path))
 
450
 
 
451
    @needs_read_lock
 
452
    def get_root_id(self):
 
453
        """Return the id of this trees root"""
 
454
        return self._get_entry(path='')[0][2]
 
455
 
 
456
    def has_id(self, file_id):
 
457
        state = self.current_dirstate()
 
458
        file_id = osutils.safe_file_id(file_id)
 
459
        row, parents = self._get_entry(file_id=file_id)
 
460
        if row is None:
 
461
            return False
 
462
        return osutils.lexists(pathjoin(
 
463
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
464
 
 
465
    @needs_read_lock
 
466
    def id2path(self, file_id):
 
467
        file_id = osutils.safe_file_id(file_id)
 
468
        state = self.current_dirstate()
 
469
        entry = self._get_entry(file_id=file_id)
 
470
        if entry == (None, None):
 
471
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
472
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
473
        return path_utf8.decode('utf8')
 
474
 
 
475
    if not osutils.supports_executable():
 
476
        @needs_read_lock
 
477
        def is_executable(self, file_id, path=None):
 
478
            file_id = osutils.safe_file_id(file_id)
 
479
            entry = self._get_entry(file_id=file_id, path=path)
 
480
            if entry == (None, None):
 
481
                return False
 
482
            return entry[1][0][3]
 
483
    else:
 
484
        @needs_read_lock
 
485
        def is_executable(self, file_id, path=None):
 
486
            if not path:
 
487
                file_id = osutils.safe_file_id(file_id)
 
488
                path = self.id2path(file_id)
 
489
            mode = os.lstat(self.abspath(path)).st_mode
 
490
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
491
 
 
492
    @needs_read_lock
 
493
    def __iter__(self):
 
494
        """Iterate through file_ids for this tree.
 
495
 
 
496
        file_ids are in a WorkingTree if they are in the working inventory
 
497
        and the working file exists.
 
498
        """
 
499
        result = []
 
500
        for key, tree_details in self.current_dirstate()._iter_entries():
 
501
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
502
                # not relevant to the working tree
 
503
                continue
 
504
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
505
            if osutils.lexists(path):
 
506
                result.append(key[2])
 
507
        return iter(result)
 
508
 
 
509
    def iter_references(self):
 
510
        for key, tree_details in self.current_dirstate()._iter_entries():
 
511
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
512
                # not relevant to the working tree
 
513
                continue
 
514
            if not key[1]:
 
515
                # the root is not a reference.
 
516
                continue
 
517
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
518
            try:
 
519
                if self._kind(path) == 'tree-reference':
 
520
                    yield path, key[2]
 
521
            except errors.NoSuchFile:
 
522
                # path is missing on disk.
 
523
                continue
 
524
 
 
525
    @needs_read_lock
 
526
    def kind(self, file_id):
 
527
        """Return the kind of a file.
 
528
 
 
529
        This is always the actual kind that's on disk, regardless of what it
 
530
        was added as.
 
531
        """
 
532
        relpath = self.id2path(file_id)
 
533
        assert relpath != None, \
 
534
            "path for id {%s} is None!" % file_id
 
535
        return self._kind(relpath)
 
536
 
 
537
    def _kind(self, relpath):
 
538
        abspath = self.abspath(relpath)
 
539
        kind = file_kind(abspath)
 
540
        if (self._repo_supports_tree_reference and
 
541
            kind == 'directory' and
 
542
            self._directory_is_tree_reference(relpath)):
 
543
            kind = 'tree-reference'
 
544
        return kind
 
545
 
 
546
    @needs_read_lock
 
547
    def _last_revision(self):
 
548
        """See Mutable.last_revision."""
 
549
        parent_ids = self.current_dirstate().get_parent_ids()
 
550
        if parent_ids:
 
551
            return parent_ids[0]
 
552
        else:
 
553
            return None
 
554
 
 
555
    def lock_read(self):
 
556
        """See Branch.lock_read, and WorkingTree.unlock."""
 
557
        self.branch.lock_read()
 
558
        try:
 
559
            self._control_files.lock_read()
 
560
            try:
 
561
                state = self.current_dirstate()
 
562
                if not state._lock_token:
 
563
                    state.lock_read()
 
564
                # set our support for tree references from the repository in
 
565
                # use.
 
566
                self._repo_supports_tree_reference = getattr(
 
567
                    self.branch.repository._format, "supports_tree_reference",
 
568
                    False)
 
569
            except:
 
570
                self._control_files.unlock()
 
571
                raise
 
572
        except:
 
573
            self.branch.unlock()
 
574
            raise
 
575
 
 
576
    def _lock_self_write(self):
 
577
        """This should be called after the branch is locked."""
 
578
        try:
 
579
            self._control_files.lock_write()
 
580
            try:
 
581
                state = self.current_dirstate()
 
582
                if not state._lock_token:
 
583
                    state.lock_write()
 
584
                # set our support for tree references from the repository in
 
585
                # use.
 
586
                self._repo_supports_tree_reference = getattr(
 
587
                    self.branch.repository._format, "supports_tree_reference",
 
588
                    False)
 
589
            except:
 
590
                self._control_files.unlock()
 
591
                raise
 
592
        except:
 
593
            self.branch.unlock()
 
594
            raise
 
595
 
 
596
    def lock_tree_write(self):
 
597
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
598
        self.branch.lock_read()
 
599
        self._lock_self_write()
 
600
 
 
601
    def lock_write(self):
 
602
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
603
        self.branch.lock_write()
 
604
        self._lock_self_write()
 
605
 
 
606
    @needs_tree_write_lock
 
607
    def move(self, from_paths, to_dir, after=False):
 
608
        """See WorkingTree.move()."""
 
609
        result = []
 
610
        if not from_paths:
 
611
            return result
 
612
 
 
613
        state = self.current_dirstate()
 
614
 
 
615
        assert not isinstance(from_paths, basestring)
 
616
        to_dir_utf8 = to_dir.encode('utf8')
 
617
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
618
        id_index = state._get_id_index()
 
619
        # check destination directory
 
620
        # get the details for it
 
621
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
622
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
623
        if not entry_present:
 
624
            raise errors.BzrMoveFailedError('', to_dir,
 
625
                errors.NotVersionedError(to_dir))
 
626
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
627
        # get a handle on the block itself.
 
628
        to_block_index = state._ensure_block(
 
629
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
630
        to_block = state._dirblocks[to_block_index]
 
631
        to_abs = self.abspath(to_dir)
 
632
        if not isdir(to_abs):
 
633
            raise errors.BzrMoveFailedError('',to_dir,
 
634
                errors.NotADirectory(to_abs))
 
635
 
 
636
        if to_entry[1][0][0] != 'd':
 
637
            raise errors.BzrMoveFailedError('',to_dir,
 
638
                errors.NotADirectory(to_abs))
 
639
 
 
640
        if self._inventory is not None:
 
641
            update_inventory = True
 
642
            inv = self.inventory
 
643
            to_dir_ie = inv[to_dir_id]
 
644
            to_dir_id = to_entry[0][2]
 
645
        else:
 
646
            update_inventory = False
 
647
 
 
648
        rollbacks = []
 
649
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
650
                     fingerprint, packed_stat, size,
 
651
                     to_block, to_key, to_path_utf8):
 
652
            state._make_absent(old_entry)
 
653
            from_key = old_entry[0]
 
654
            rollbacks.append(
 
655
                lambda:state.update_minimal(from_key,
 
656
                    minikind,
 
657
                    executable=executable,
 
658
                    fingerprint=fingerprint,
 
659
                    packed_stat=packed_stat,
 
660
                    size=size,
 
661
                    path_utf8=from_path_utf8))
 
662
            state.update_minimal(to_key,
 
663
                    minikind,
 
664
                    executable=executable,
 
665
                    fingerprint=fingerprint,
 
666
                    packed_stat=packed_stat,
 
667
                    size=size,
 
668
                    path_utf8=to_path_utf8)
 
669
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
670
            new_entry = to_block[1][added_entry_index]
 
671
            rollbacks.append(lambda:state._make_absent(new_entry))
 
672
 
 
673
        for from_rel in from_paths:
 
674
            # from_rel is 'pathinroot/foo/bar'
 
675
            from_rel_utf8 = from_rel.encode('utf8')
 
676
            from_dirname, from_tail = osutils.split(from_rel)
 
677
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
678
            from_entry = self._get_entry(path=from_rel)
 
679
            if from_entry == (None, None):
 
680
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
681
                    errors.NotVersionedError(path=str(from_rel)))
 
682
 
 
683
            from_id = from_entry[0][2]
 
684
            to_rel = pathjoin(to_dir, from_tail)
 
685
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
686
            item_to_entry = self._get_entry(path=to_rel)
 
687
            if item_to_entry != (None, None):
 
688
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
689
                    "Target is already versioned.")
 
690
 
 
691
            if from_rel == to_rel:
 
692
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
693
                    "Source and target are identical.")
 
694
 
 
695
            from_missing = not self.has_filename(from_rel)
 
696
            to_missing = not self.has_filename(to_rel)
 
697
            if after:
 
698
                move_file = False
 
699
            else:
 
700
                move_file = True
 
701
            if to_missing:
 
702
                if not move_file:
 
703
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
704
                        errors.NoSuchFile(path=to_rel,
 
705
                        extra="New file has not been created yet"))
 
706
                elif from_missing:
 
707
                    # neither path exists
 
708
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
709
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
710
            else:
 
711
                if from_missing: # implicitly just update our path mapping
 
712
                    move_file = False
 
713
                elif not after:
 
714
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
715
                        extra="(Use --after to update the Bazaar id)")
 
716
 
 
717
            rollbacks = []
 
718
            def rollback_rename():
 
719
                """A single rename has failed, roll it back."""
 
720
                # roll back everything, even if we encounter trouble doing one
 
721
                # of them.
 
722
                #
 
723
                # TODO: at least log the other exceptions rather than just
 
724
                # losing them mbp 20070307
 
725
                exc_info = None
 
726
                for rollback in reversed(rollbacks):
 
727
                    try:
 
728
                        rollback()
 
729
                    except Exception, e:
 
730
                        exc_info = sys.exc_info()
 
731
                if exc_info:
 
732
                    raise exc_info[0], exc_info[1], exc_info[2]
 
733
 
 
734
            # perform the disk move first - its the most likely failure point.
 
735
            if move_file:
 
736
                from_rel_abs = self.abspath(from_rel)
 
737
                to_rel_abs = self.abspath(to_rel)
 
738
                try:
 
739
                    osutils.rename(from_rel_abs, to_rel_abs)
 
740
                except OSError, e:
 
741
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
742
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
743
            try:
 
744
                # perform the rename in the inventory next if needed: its easy
 
745
                # to rollback
 
746
                if update_inventory:
 
747
                    # rename the entry
 
748
                    from_entry = inv[from_id]
 
749
                    current_parent = from_entry.parent_id
 
750
                    inv.rename(from_id, to_dir_id, from_tail)
 
751
                    rollbacks.append(
 
752
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
753
                # finally do the rename in the dirstate, which is a little
 
754
                # tricky to rollback, but least likely to need it.
 
755
                old_block_index, old_entry_index, dir_present, file_present = \
 
756
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
757
                old_block = state._dirblocks[old_block_index][1]
 
758
                old_entry = old_block[old_entry_index]
 
759
                from_key, old_entry_details = old_entry
 
760
                cur_details = old_entry_details[0]
 
761
                # remove the old row
 
762
                to_key = ((to_block[0],) + from_key[1:3])
 
763
                minikind = cur_details[0]
 
764
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
765
                         minikind=minikind,
 
766
                         executable=cur_details[3],
 
767
                         fingerprint=cur_details[1],
 
768
                         packed_stat=cur_details[4],
 
769
                         size=cur_details[2],
 
770
                         to_block=to_block,
 
771
                         to_key=to_key,
 
772
                         to_path_utf8=to_rel_utf8)
 
773
 
 
774
                if minikind == 'd':
 
775
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
776
                        """Recursively update all entries in this dirblock."""
 
777
                        assert from_dir != '', "renaming root not supported"
 
778
                        from_key = (from_dir, '')
 
779
                        from_block_idx, present = \
 
780
                            state._find_block_index_from_key(from_key)
 
781
                        if not present:
 
782
                            # This is the old record, if it isn't present, then
 
783
                            # there is theoretically nothing to update.
 
784
                            # (Unless it isn't present because of lazy loading,
 
785
                            # but we don't do that yet)
 
786
                            return
 
787
                        from_block = state._dirblocks[from_block_idx]
 
788
                        to_block_index, to_entry_index, _, _ = \
 
789
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
790
                        to_block_index = state._ensure_block(
 
791
                            to_block_index, to_entry_index, to_dir_utf8)
 
792
                        to_block = state._dirblocks[to_block_index]
 
793
 
 
794
                        # Grab a copy since move_one may update the list.
 
795
                        for entry in from_block[1][:]:
 
796
                            assert entry[0][0] == from_dir
 
797
                            cur_details = entry[1][0]
 
798
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
799
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
800
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
801
                            minikind = cur_details[0]
 
802
                            if minikind in 'ar':
 
803
                                # Deleted children of a renamed directory
 
804
                                # Do not need to be updated.
 
805
                                # Children that have been renamed out of this
 
806
                                # directory should also not be updated
 
807
                                continue
 
808
                            move_one(entry, from_path_utf8=from_path_utf8,
 
809
                                     minikind=minikind,
 
810
                                     executable=cur_details[3],
 
811
                                     fingerprint=cur_details[1],
 
812
                                     packed_stat=cur_details[4],
 
813
                                     size=cur_details[2],
 
814
                                     to_block=to_block,
 
815
                                     to_key=to_key,
 
816
                                     to_path_utf8=to_path_utf8)
 
817
                            if minikind == 'd':
 
818
                                # We need to move all the children of this
 
819
                                # entry
 
820
                                update_dirblock(from_path_utf8, to_key,
 
821
                                                to_path_utf8)
 
822
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
823
            except:
 
824
                rollback_rename()
 
825
                raise
 
826
            result.append((from_rel, to_rel))
 
827
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
828
            self._make_dirty(reset_inventory=False)
 
829
 
 
830
        return result
 
831
 
 
832
    def _must_be_locked(self):
 
833
        if not self._control_files._lock_count:
 
834
            raise errors.ObjectNotLocked(self)
 
835
 
 
836
    def _new_tree(self):
 
837
        """Initialize the state in this tree to be a new tree."""
 
838
        self._dirty = True
 
839
 
 
840
    @needs_read_lock
 
841
    def path2id(self, path):
 
842
        """Return the id for path in this tree."""
 
843
        path = path.strip('/')
 
844
        entry = self._get_entry(path=path)
 
845
        if entry == (None, None):
 
846
            return None
 
847
        return entry[0][2]
 
848
 
 
849
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
850
        """See Tree.paths2ids().
 
851
 
 
852
        This specialisation fast-paths the case where all the trees are in the
 
853
        dirstate.
 
854
        """
 
855
        if paths is None:
 
856
            return None
 
857
        parents = self.get_parent_ids()
 
858
        for tree in trees:
 
859
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
860
                parents):
 
861
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
862
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
863
        # -- make all paths utf8 --
 
864
        paths_utf8 = set()
 
865
        for path in paths:
 
866
            paths_utf8.add(path.encode('utf8'))
 
867
        paths = paths_utf8
 
868
        # -- paths is now a utf8 path set --
 
869
        # -- get the state object and prepare it.
 
870
        state = self.current_dirstate()
 
871
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
872
            and '' not in paths):
 
873
            paths2ids = self._paths2ids_using_bisect
 
874
        else:
 
875
            paths2ids = self._paths2ids_in_memory
 
876
        return paths2ids(paths, search_indexes,
 
877
                         require_versioned=require_versioned)
 
878
 
 
879
    def _paths2ids_in_memory(self, paths, search_indexes,
 
880
                             require_versioned=True):
 
881
        state = self.current_dirstate()
 
882
        state._read_dirblocks_if_needed()
 
883
        def _entries_for_path(path):
 
884
            """Return a list with all the entries that match path for all ids.
 
885
            """
 
886
            dirname, basename = os.path.split(path)
 
887
            key = (dirname, basename, '')
 
888
            block_index, present = state._find_block_index_from_key(key)
 
889
            if not present:
 
890
                # the block which should contain path is absent.
 
891
                return []
 
892
            result = []
 
893
            block = state._dirblocks[block_index][1]
 
894
            entry_index, _ = state._find_entry_index(key, block)
 
895
            # we may need to look at multiple entries at this path: walk while the paths match.
 
896
            while (entry_index < len(block) and
 
897
                block[entry_index][0][0:2] == key[0:2]):
 
898
                result.append(block[entry_index])
 
899
                entry_index += 1
 
900
            return result
 
901
        if require_versioned:
 
902
            # -- check all supplied paths are versioned in a search tree. --
 
903
            all_versioned = True
 
904
            for path in paths:
 
905
                path_entries = _entries_for_path(path)
 
906
                if not path_entries:
 
907
                    # this specified path is not present at all: error
 
908
                    all_versioned = False
 
909
                    break
 
910
                found_versioned = False
 
911
                # for each id at this path
 
912
                for entry in path_entries:
 
913
                    # for each tree.
 
914
                    for index in search_indexes:
 
915
                        if entry[1][index][0] != 'a': # absent
 
916
                            found_versioned = True
 
917
                            # all good: found a versioned cell
 
918
                            break
 
919
                if not found_versioned:
 
920
                    # none of the indexes was not 'absent' at all ids for this
 
921
                    # path.
 
922
                    all_versioned = False
 
923
                    break
 
924
            if not all_versioned:
 
925
                raise errors.PathsNotVersionedError(paths)
 
926
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
927
        search_paths = set()
 
928
        for path in paths:
 
929
            other_paths = paths.difference(set([path]))
 
930
            if not osutils.is_inside_any(other_paths, path):
 
931
                # this is a top level path, we must check it.
 
932
                search_paths.add(path)
 
933
        # sketch: 
 
934
        # for all search_indexs in each path at or under each element of
 
935
        # search_paths, if the detail is relocated: add the id, and add the
 
936
        # relocated path as one to search if its not searched already. If the
 
937
        # detail is not relocated, add the id.
 
938
        searched_paths = set()
 
939
        found_ids = set()
 
940
        def _process_entry(entry):
 
941
            """Look at search_indexes within entry.
 
942
 
 
943
            If a specific tree's details are relocated, add the relocation
 
944
            target to search_paths if not searched already. If it is absent, do
 
945
            nothing. Otherwise add the id to found_ids.
 
946
            """
 
947
            for index in search_indexes:
 
948
                if entry[1][index][0] == 'r': # relocated
 
949
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
950
                        search_paths.add(entry[1][index][1])
 
951
                elif entry[1][index][0] != 'a': # absent
 
952
                    found_ids.add(entry[0][2])
 
953
        while search_paths:
 
954
            current_root = search_paths.pop()
 
955
            searched_paths.add(current_root)
 
956
            # process the entries for this containing directory: the rest will be
 
957
            # found by their parents recursively.
 
958
            root_entries = _entries_for_path(current_root)
 
959
            if not root_entries:
 
960
                # this specified path is not present at all, skip it.
 
961
                continue
 
962
            for entry in root_entries:
 
963
                _process_entry(entry)
 
964
            initial_key = (current_root, '', '')
 
965
            block_index, _ = state._find_block_index_from_key(initial_key)
 
966
            while (block_index < len(state._dirblocks) and
 
967
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
968
                for entry in state._dirblocks[block_index][1]:
 
969
                    _process_entry(entry)
 
970
                block_index += 1
 
971
        return found_ids
 
972
 
 
973
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
974
                                require_versioned=True):
 
975
        state = self.current_dirstate()
 
976
        found_ids = set()
 
977
 
 
978
        split_paths = sorted(osutils.split(p) for p in paths)
 
979
        found = state._bisect_recursive(split_paths)
 
980
 
 
981
        if require_versioned:
 
982
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
983
            for dir_name in split_paths:
 
984
                if dir_name not in found_dir_names:
 
985
                    raise errors.PathsNotVersionedError(paths)
 
986
 
 
987
        for dir_name_id, trees_info in found.iteritems():
 
988
            for index in search_indexes:
 
989
                if trees_info[index][0] not in ('r', 'a'):
 
990
                    found_ids.add(dir_name_id[2])
 
991
        return found_ids
 
992
 
 
993
    def read_working_inventory(self):
 
994
        """Read the working inventory.
 
995
        
 
996
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
997
        """
 
998
        return self.inventory
 
999
 
 
1000
    @needs_read_lock
 
1001
    def revision_tree(self, revision_id):
 
1002
        """See Tree.revision_tree.
 
1003
 
 
1004
        WorkingTree4 supplies revision_trees for any basis tree.
 
1005
        """
 
1006
        revision_id = osutils.safe_revision_id(revision_id)
 
1007
        dirstate = self.current_dirstate()
 
1008
        parent_ids = dirstate.get_parent_ids()
 
1009
        if revision_id not in parent_ids:
 
1010
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1011
        if revision_id in dirstate.get_ghosts():
 
1012
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1013
        return DirStateRevisionTree(dirstate, revision_id,
 
1014
            self.branch.repository)
 
1015
 
 
1016
    @needs_tree_write_lock
 
1017
    def set_last_revision(self, new_revision):
 
1018
        """Change the last revision in the working tree."""
 
1019
        new_revision = osutils.safe_revision_id(new_revision)
 
1020
        parents = self.get_parent_ids()
 
1021
        if new_revision in (NULL_REVISION, None):
 
1022
            assert len(parents) < 2, (
 
1023
                "setting the last parent to none with a pending merge is "
 
1024
                "unsupported.")
 
1025
            self.set_parent_ids([])
 
1026
        else:
 
1027
            self.set_parent_ids([new_revision] + parents[1:],
 
1028
                allow_leftmost_as_ghost=True)
 
1029
 
 
1030
    @needs_tree_write_lock
 
1031
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1032
        """Set the parent ids to revision_ids.
 
1033
        
 
1034
        See also set_parent_trees. This api will try to retrieve the tree data
 
1035
        for each element of revision_ids from the trees repository. If you have
 
1036
        tree data already available, it is more efficient to use
 
1037
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1038
        an easier API to use.
 
1039
 
 
1040
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1041
            working tree. Any of these may be ghosts.
 
1042
        """
 
1043
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1044
        trees = []
 
1045
        for revision_id in revision_ids:
 
1046
            try:
 
1047
                revtree = self.branch.repository.revision_tree(revision_id)
 
1048
                # TODO: jam 20070213 KnitVersionedFile raises
 
1049
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1050
                #       given revision_id is not present. Should Repository be
 
1051
                #       catching it and re-raising NoSuchRevision?
 
1052
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1053
                revtree = None
 
1054
            trees.append((revision_id, revtree))
 
1055
        self.current_dirstate()._validate()
 
1056
        self.set_parent_trees(trees,
 
1057
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1058
        self.current_dirstate()._validate()
 
1059
 
 
1060
    @needs_tree_write_lock
 
1061
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1062
        """Set the parents of the working tree.
 
1063
 
 
1064
        :param parents_list: A list of (revision_id, tree) tuples.
 
1065
            If tree is None, then that element is treated as an unreachable
 
1066
            parent tree - i.e. a ghost.
 
1067
        """
 
1068
        dirstate = self.current_dirstate()
 
1069
        dirstate._validate()
 
1070
        if len(parents_list) > 0:
 
1071
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1072
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1073
        real_trees = []
 
1074
        ghosts = []
 
1075
        # convert absent trees to the null tree, which we convert back to
 
1076
        # missing on access.
 
1077
        for rev_id, tree in parents_list:
 
1078
            rev_id = osutils.safe_revision_id(rev_id)
 
1079
            if tree is not None:
 
1080
                real_trees.append((rev_id, tree))
 
1081
            else:
 
1082
                real_trees.append((rev_id,
 
1083
                    self.branch.repository.revision_tree(None)))
 
1084
                ghosts.append(rev_id)
 
1085
        dirstate._validate()
 
1086
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1087
        dirstate._validate()
 
1088
        self._make_dirty(reset_inventory=False)
 
1089
        dirstate._validate()
 
1090
 
 
1091
    def _set_root_id(self, file_id):
 
1092
        """See WorkingTree.set_root_id."""
 
1093
        state = self.current_dirstate()
 
1094
        state.set_path_id('', file_id)
 
1095
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1096
            self._make_dirty(reset_inventory=True)
 
1097
 
 
1098
    @needs_read_lock
 
1099
    def supports_tree_reference(self):
 
1100
        return self._repo_supports_tree_reference
 
1101
 
 
1102
    def unlock(self):
 
1103
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1104
        if self._control_files._lock_count == 1:
 
1105
            # eventually we should do signature checking during read locks for
 
1106
            # dirstate updates.
 
1107
            if self._control_files._lock_mode == 'w':
 
1108
                if self._dirty:
 
1109
                    self.flush()
 
1110
            if self._dirstate is not None:
 
1111
                # This is a no-op if there are no modifications.
 
1112
                self._dirstate.save()
 
1113
                self._dirstate.unlock()
 
1114
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1115
            #       point. Instead, it could check if the header has been
 
1116
            #       modified when it is locked, and if not, it can hang on to
 
1117
            #       the data it has in memory.
 
1118
            self._dirstate = None
 
1119
            self._inventory = None
 
1120
        # reverse order of locking.
 
1121
        try:
 
1122
            return self._control_files.unlock()
 
1123
        finally:
 
1124
            self.branch.unlock()
 
1125
 
 
1126
    @needs_tree_write_lock
 
1127
    def unversion(self, file_ids):
 
1128
        """Remove the file ids in file_ids from the current versioned set.
 
1129
 
 
1130
        When a file_id is unversioned, all of its children are automatically
 
1131
        unversioned.
 
1132
 
 
1133
        :param file_ids: The file ids to stop versioning.
 
1134
        :raises: NoSuchId if any fileid is not currently versioned.
 
1135
        """
 
1136
        if not file_ids:
 
1137
            return
 
1138
        state = self.current_dirstate()
 
1139
        state._read_dirblocks_if_needed()
 
1140
        ids_to_unversion = set()
 
1141
        for file_id in file_ids:
 
1142
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1143
        paths_to_unversion = set()
 
1144
        # sketch:
 
1145
        # check if the root is to be unversioned, if so, assert for now.
 
1146
        # walk the state marking unversioned things as absent.
 
1147
        # if there are any un-unversioned ids at the end, raise
 
1148
        for key, details in state._dirblocks[0][1]:
 
1149
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1150
                key[2] in ids_to_unversion):
 
1151
                # I haven't written the code to unversion / yet - it should be
 
1152
                # supported.
 
1153
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1154
        block_index = 0
 
1155
        while block_index < len(state._dirblocks):
 
1156
            # process one directory at a time.
 
1157
            block = state._dirblocks[block_index]
 
1158
            # first check: is the path one to remove - it or its children
 
1159
            delete_block = False
 
1160
            for path in paths_to_unversion:
 
1161
                if (block[0].startswith(path) and
 
1162
                    (len(block[0]) == len(path) or
 
1163
                     block[0][len(path)] == '/')):
 
1164
                    # this entire block should be deleted - its the block for a
 
1165
                    # path to unversion; or the child of one
 
1166
                    delete_block = True
 
1167
                    break
 
1168
            # TODO: trim paths_to_unversion as we pass by paths
 
1169
            if delete_block:
 
1170
                # this block is to be deleted: process it.
 
1171
                # TODO: we can special case the no-parents case and
 
1172
                # just forget the whole block.
 
1173
                entry_index = 0
 
1174
                while entry_index < len(block[1]):
 
1175
                    # Mark this file id as having been removed
 
1176
                    entry = block[1][entry_index]
 
1177
                    ids_to_unversion.discard(entry[0][2])
 
1178
                    if (entry[1][0][0] == 'a'
 
1179
                        or not state._make_absent(entry)):
 
1180
                        entry_index += 1
 
1181
                # go to the next block. (At the moment we dont delete empty
 
1182
                # dirblocks)
 
1183
                block_index += 1
 
1184
                continue
 
1185
            entry_index = 0
 
1186
            while entry_index < len(block[1]):
 
1187
                entry = block[1][entry_index]
 
1188
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1189
                    # ^ some parent row.
 
1190
                    entry[0][2] not in ids_to_unversion):
 
1191
                    # ^ not an id to unversion
 
1192
                    entry_index += 1
 
1193
                    continue
 
1194
                if entry[1][0][0] == 'd':
 
1195
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1196
                if not state._make_absent(entry):
 
1197
                    entry_index += 1
 
1198
                # we have unversioned this id
 
1199
                ids_to_unversion.remove(entry[0][2])
 
1200
            block_index += 1
 
1201
        if ids_to_unversion:
 
1202
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1203
        self._make_dirty(reset_inventory=False)
 
1204
        # have to change the legacy inventory too.
 
1205
        if self._inventory is not None:
 
1206
            for file_id in file_ids:
 
1207
                self._inventory.remove_recursive_id(file_id)
 
1208
 
 
1209
    @needs_read_lock
 
1210
    def _validate(self):
 
1211
        self._dirstate._validate()
 
1212
 
 
1213
    @needs_tree_write_lock
 
1214
    def _write_inventory(self, inv):
 
1215
        """Write inventory as the current inventory."""
 
1216
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1217
        self.current_dirstate().set_state_from_inventory(inv)
 
1218
        self._make_dirty(reset_inventory=False)
 
1219
        if self._inventory is not None:
 
1220
            self._inventory = inv
 
1221
        self.flush()
 
1222
 
 
1223
 
 
1224
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1225
    """The first consolidated dirstate working tree format.
 
1226
 
 
1227
    This format:
 
1228
        - exists within a metadir controlling .bzr
 
1229
        - includes an explicit version marker for the workingtree control
 
1230
          files, separate from the BzrDir format
 
1231
        - modifies the hash cache format
 
1232
        - is new in bzr 0.15
 
1233
        - uses a LockDir to guard access to it.
 
1234
    """
 
1235
 
 
1236
    upgrade_recommended = False
 
1237
 
 
1238
    def get_format_string(self):
 
1239
        """See WorkingTreeFormat.get_format_string()."""
 
1240
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1241
 
 
1242
    def get_format_description(self):
 
1243
        """See WorkingTreeFormat.get_format_description()."""
 
1244
        return "Working tree format 4"
 
1245
 
 
1246
    def initialize(self, a_bzrdir, revision_id=None):
 
1247
        """See WorkingTreeFormat.initialize().
 
1248
 
 
1249
        :param revision_id: allows creating a working tree at a different
 
1250
        revision than the branch is at.
 
1251
 
 
1252
        These trees get an initial random root id, if their repository supports
 
1253
        rich root data, TREE_ROOT otherwise.
 
1254
        """
 
1255
        revision_id = osutils.safe_revision_id(revision_id)
 
1256
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1257
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1258
        transport = a_bzrdir.get_workingtree_transport(self)
 
1259
        control_files = self._open_control_files(a_bzrdir)
 
1260
        control_files.create_lock()
 
1261
        control_files.lock_write()
 
1262
        control_files.put_utf8('format', self.get_format_string())
 
1263
        branch = a_bzrdir.open_branch()
 
1264
        if revision_id is None:
 
1265
            revision_id = branch.last_revision()
 
1266
        local_path = transport.local_abspath('dirstate')
 
1267
        # write out new dirstate (must exist when we create the tree)
 
1268
        state = dirstate.DirState.initialize(local_path)
 
1269
        state.unlock()
 
1270
        del state
 
1271
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1272
                         branch,
 
1273
                         _format=self,
 
1274
                         _bzrdir=a_bzrdir,
 
1275
                         _control_files=control_files)
 
1276
        wt._new_tree()
 
1277
        wt.lock_tree_write()
 
1278
        wt.current_dirstate()._validate()
 
1279
        try:
 
1280
            if revision_id in (None, NULL_REVISION):
 
1281
                if branch.repository.supports_rich_root():
 
1282
                    wt._set_root_id(generate_ids.gen_root_id())
 
1283
                else:
 
1284
                    wt._set_root_id(ROOT_ID)
 
1285
                wt.flush()
 
1286
                wt.current_dirstate()._validate()
 
1287
            wt.set_last_revision(revision_id)
 
1288
            wt.flush()
 
1289
            basis = wt.basis_tree()
 
1290
            basis.lock_read()
 
1291
            # if the basis has a root id we have to use that; otherwise we use
 
1292
            # a new random one
 
1293
            basis_root_id = basis.get_root_id()
 
1294
            if basis_root_id is not None:
 
1295
                wt._set_root_id(basis_root_id)
 
1296
                wt.flush()
 
1297
            transform.build_tree(basis, wt)
 
1298
            basis.unlock()
 
1299
        finally:
 
1300
            control_files.unlock()
 
1301
            wt.unlock()
 
1302
        return wt
 
1303
 
 
1304
    def _open(self, a_bzrdir, control_files):
 
1305
        """Open the tree itself.
 
1306
 
 
1307
        :param a_bzrdir: the dir for the tree.
 
1308
        :param control_files: the control files for the tree.
 
1309
        """
 
1310
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1311
                           branch=a_bzrdir.open_branch(),
 
1312
                           _format=self,
 
1313
                           _bzrdir=a_bzrdir,
 
1314
                           _control_files=control_files)
 
1315
 
 
1316
    def __get_matchingbzrdir(self):
 
1317
        # please test against something that will let us do tree references
 
1318
        return bzrdir.format_registry.make_bzrdir(
 
1319
            'dirstate-with-subtree')
 
1320
 
 
1321
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1322
 
 
1323
 
 
1324
class DirStateRevisionTree(Tree):
 
1325
    """A revision tree pulling the inventory from a dirstate."""
 
1326
 
 
1327
    def __init__(self, dirstate, revision_id, repository):
 
1328
        self._dirstate = dirstate
 
1329
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1330
        self._repository = repository
 
1331
        self._inventory = None
 
1332
        self._locked = 0
 
1333
        self._dirstate_locked = False
 
1334
 
 
1335
    def __repr__(self):
 
1336
        return "<%s of %s in %s>" % \
 
1337
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1338
 
 
1339
    def annotate_iter(self, file_id):
 
1340
        """See Tree.annotate_iter"""
 
1341
        w = self._repository.weave_store.get_weave(file_id,
 
1342
                           self._repository.get_transaction())
 
1343
        return w.annotate_iter(self.inventory[file_id].revision)
 
1344
 
 
1345
    def _comparison_data(self, entry, path):
 
1346
        """See Tree._comparison_data."""
 
1347
        if entry is None:
 
1348
            return None, False, None
 
1349
        # trust the entry as RevisionTree does, but this may not be
 
1350
        # sensible: the entry might not have come from us?
 
1351
        return entry.kind, entry.executable, None
 
1352
 
 
1353
    def _file_size(self, entry, stat_value):
 
1354
        return entry.text_size
 
1355
 
 
1356
    def filter_unversioned_files(self, paths):
 
1357
        """Filter out paths that are not versioned.
 
1358
 
 
1359
        :return: set of paths.
 
1360
        """
 
1361
        pred = self.has_filename
 
1362
        return set((p for p in paths if not pred(p)))
 
1363
 
 
1364
    def get_root_id(self):
 
1365
        return self.path2id('')
 
1366
 
 
1367
    def _get_parent_index(self):
 
1368
        """Return the index in the dirstate referenced by this tree."""
 
1369
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1370
 
 
1371
    def _get_entry(self, file_id=None, path=None):
 
1372
        """Get the dirstate row for file_id or path.
 
1373
 
 
1374
        If either file_id or path is supplied, it is used as the key to lookup.
 
1375
        If both are supplied, the fastest lookup is used, and an error is
 
1376
        raised if they do not both point at the same row.
 
1377
        
 
1378
        :param file_id: An optional unicode file_id to be looked up.
 
1379
        :param path: An optional unicode path to be looked up.
 
1380
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1381
        """
 
1382
        if file_id is None and path is None:
 
1383
            raise errors.BzrError('must supply file_id or path')
 
1384
        file_id = osutils.safe_file_id(file_id)
 
1385
        if path is not None:
 
1386
            path = path.encode('utf8')
 
1387
        parent_index = self._get_parent_index()
 
1388
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1389
 
 
1390
    def _generate_inventory(self):
 
1391
        """Create and set self.inventory from the dirstate object.
 
1392
 
 
1393
        (So this is only called the first time the inventory is requested for
 
1394
        this tree; it then remains in memory until it's out of date.)
 
1395
 
 
1396
        This is relatively expensive: we have to walk the entire dirstate.
 
1397
        """
 
1398
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1399
            'dirstate revision tree'
 
1400
        # separate call for profiling - makes it clear where the costs are.
 
1401
        self._dirstate._read_dirblocks_if_needed()
 
1402
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1403
            'parent %s has disappeared from %s' % (
 
1404
            self._revision_id, self._dirstate.get_parent_ids())
 
1405
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1406
        # This is identical now to the WorkingTree _generate_inventory except
 
1407
        # for the tree index use.
 
1408
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1409
        current_id = root_key[2]
 
1410
        assert current_entry[parent_index][0] == 'd'
 
1411
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1412
        inv.root.revision = current_entry[parent_index][4]
 
1413
        # Turn some things into local variables
 
1414
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1415
        factory = entry_factory
 
1416
        utf8_decode = cache_utf8._utf8_decode
 
1417
        inv_byid = inv._byid
 
1418
        # we could do this straight out of the dirstate; it might be fast
 
1419
        # and should be profiled - RBC 20070216
 
1420
        parent_ies = {'' : inv.root}
 
1421
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1422
            dirname = block[0]
 
1423
            try:
 
1424
                parent_ie = parent_ies[dirname]
 
1425
            except KeyError:
 
1426
                # all the paths in this block are not versioned in this tree
 
1427
                continue
 
1428
            for key, entry in block[1]:
 
1429
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1430
                if minikind in ('a', 'r'): # absent, relocated
 
1431
                    # not this tree
 
1432
                    continue
 
1433
                name = key[1]
 
1434
                name_unicode = utf8_decode(name)[0]
 
1435
                file_id = key[2]
 
1436
                kind = minikind_to_kind[minikind]
 
1437
                inv_entry = factory[kind](file_id, name_unicode,
 
1438
                                          parent_ie.file_id)
 
1439
                inv_entry.revision = revid
 
1440
                if kind == 'file':
 
1441
                    inv_entry.executable = executable
 
1442
                    inv_entry.text_size = size
 
1443
                    inv_entry.text_sha1 = fingerprint
 
1444
                elif kind == 'directory':
 
1445
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1446
                elif kind == 'symlink':
 
1447
                    inv_entry.executable = False
 
1448
                    inv_entry.text_size = None
 
1449
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1450
                elif kind == 'tree-reference':
 
1451
                    inv_entry.reference_revision = fingerprint or None
 
1452
                else:
 
1453
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1454
                            % entry)
 
1455
                # These checks cost us around 40ms on a 55k entry tree
 
1456
                assert file_id not in inv_byid
 
1457
                assert name_unicode not in parent_ie.children
 
1458
                inv_byid[file_id] = inv_entry
 
1459
                parent_ie.children[name_unicode] = inv_entry
 
1460
        self._inventory = inv
 
1461
 
 
1462
    def get_file_mtime(self, file_id, path=None):
 
1463
        """Return the modification time for this record.
 
1464
 
 
1465
        We return the timestamp of the last-changed revision.
 
1466
        """
 
1467
        # Make sure the file exists
 
1468
        entry = self._get_entry(file_id, path=path)
 
1469
        if entry == (None, None): # do we raise?
 
1470
            return None
 
1471
        parent_index = self._get_parent_index()
 
1472
        last_changed_revision = entry[1][parent_index][4]
 
1473
        return self._repository.get_revision(last_changed_revision).timestamp
 
1474
 
 
1475
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1476
        entry = self._get_entry(file_id=file_id, path=path)
 
1477
        parent_index = self._get_parent_index()
 
1478
        parent_details = entry[1][parent_index]
 
1479
        if parent_details[0] == 'f':
 
1480
            return parent_details[1]
 
1481
        return None
 
1482
 
 
1483
    def get_file(self, file_id):
 
1484
        return StringIO(self.get_file_text(file_id))
 
1485
 
 
1486
    def get_file_lines(self, file_id):
 
1487
        ie = self.inventory[file_id]
 
1488
        return self._repository.weave_store.get_weave(file_id,
 
1489
                self._repository.get_transaction()).get_lines(ie.revision)
 
1490
 
 
1491
    def get_file_size(self, file_id):
 
1492
        return self.inventory[file_id].text_size
 
1493
 
 
1494
    def get_file_text(self, file_id):
 
1495
        return ''.join(self.get_file_lines(file_id))
 
1496
 
 
1497
    def get_reference_revision(self, file_id, path=None):
 
1498
        return self.inventory[file_id].reference_revision
 
1499
 
 
1500
    def get_symlink_target(self, file_id):
 
1501
        entry = self._get_entry(file_id=file_id)
 
1502
        parent_index = self._get_parent_index()
 
1503
        if entry[1][parent_index][0] != 'l':
 
1504
            return None
 
1505
        else:
 
1506
            # At present, none of the tree implementations supports non-ascii
 
1507
            # symlink targets. So we will just assume that the dirstate path is
 
1508
            # correct.
 
1509
            return entry[1][parent_index][1]
 
1510
 
 
1511
    def get_revision_id(self):
 
1512
        """Return the revision id for this tree."""
 
1513
        return self._revision_id
 
1514
 
 
1515
    def _get_inventory(self):
 
1516
        if self._inventory is not None:
 
1517
            return self._inventory
 
1518
        self._must_be_locked()
 
1519
        self._generate_inventory()
 
1520
        return self._inventory
 
1521
 
 
1522
    inventory = property(_get_inventory,
 
1523
                         doc="Inventory of this Tree")
 
1524
 
 
1525
    def get_parent_ids(self):
 
1526
        """The parents of a tree in the dirstate are not cached."""
 
1527
        return self._repository.get_revision(self._revision_id).parent_ids
 
1528
 
 
1529
    def has_filename(self, filename):
 
1530
        return bool(self.path2id(filename))
 
1531
 
 
1532
    def kind(self, file_id):
 
1533
        return self.inventory[file_id].kind
 
1534
 
 
1535
    def is_executable(self, file_id, path=None):
 
1536
        ie = self.inventory[file_id]
 
1537
        if ie.kind != "file":
 
1538
            return None
 
1539
        return ie.executable
 
1540
 
 
1541
    def list_files(self, include_root=False):
 
1542
        # We use a standard implementation, because DirStateRevisionTree is
 
1543
        # dealing with one of the parents of the current state
 
1544
        inv = self._get_inventory()
 
1545
        entries = inv.iter_entries()
 
1546
        if self.inventory.root is not None and not include_root:
 
1547
            entries.next()
 
1548
        for path, entry in entries:
 
1549
            yield path, 'V', entry.kind, entry.file_id, entry
 
1550
 
 
1551
    def lock_read(self):
 
1552
        """Lock the tree for a set of operations."""
 
1553
        if not self._locked:
 
1554
            self._repository.lock_read()
 
1555
            if self._dirstate._lock_token is None:
 
1556
                self._dirstate.lock_read()
 
1557
                self._dirstate_locked = True
 
1558
        self._locked += 1
 
1559
 
 
1560
    def _must_be_locked(self):
 
1561
        if not self._locked:
 
1562
            raise errors.ObjectNotLocked(self)
 
1563
 
 
1564
    @needs_read_lock
 
1565
    def path2id(self, path):
 
1566
        """Return the id for path in this tree."""
 
1567
        # lookup by path: faster than splitting and walking the ivnentory.
 
1568
        entry = self._get_entry(path=path)
 
1569
        if entry == (None, None):
 
1570
            return None
 
1571
        return entry[0][2]
 
1572
 
 
1573
    def unlock(self):
 
1574
        """Unlock, freeing any cache memory used during the lock."""
 
1575
        # outside of a lock, the inventory is suspect: release it.
 
1576
        self._locked -=1
 
1577
        if not self._locked:
 
1578
            self._inventory = None
 
1579
            self._locked = 0
 
1580
            if self._dirstate_locked:
 
1581
                self._dirstate.unlock()
 
1582
                self._dirstate_locked = False
 
1583
            self._repository.unlock()
 
1584
 
 
1585
    def walkdirs(self, prefix=""):
 
1586
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1587
        # implementation based on an inventory.  
 
1588
        # This should be cleaned up to use the much faster Dirstate code
 
1589
        # So for now, we just build up the parent inventory, and extract
 
1590
        # it the same way RevisionTree does.
 
1591
        _directory = 'directory'
 
1592
        inv = self._get_inventory()
 
1593
        top_id = inv.path2id(prefix)
 
1594
        if top_id is None:
 
1595
            pending = []
 
1596
        else:
 
1597
            pending = [(prefix, top_id)]
 
1598
        while pending:
 
1599
            dirblock = []
 
1600
            relpath, file_id = pending.pop()
 
1601
            # 0 - relpath, 1- file-id
 
1602
            if relpath:
 
1603
                relroot = relpath + '/'
 
1604
            else:
 
1605
                relroot = ""
 
1606
            # FIXME: stash the node in pending
 
1607
            entry = inv[file_id]
 
1608
            for name, child in entry.sorted_children():
 
1609
                toppath = relroot + name
 
1610
                dirblock.append((toppath, name, child.kind, None,
 
1611
                    child.file_id, child.kind
 
1612
                    ))
 
1613
            yield (relpath, entry.file_id), dirblock
 
1614
            # push the user specified dirs from dirblock
 
1615
            for dir in reversed(dirblock):
 
1616
                if dir[2] == _directory:
 
1617
                    pending.append((dir[0], dir[4]))
 
1618
 
 
1619
 
 
1620
class InterDirStateTree(InterTree):
 
1621
    """Fast path optimiser for changes_from with dirstate trees.
 
1622
    
 
1623
    This is used only when both trees are in the dirstate working file, and 
 
1624
    the source is any parent within the dirstate, and the destination is 
 
1625
    the current working tree of the same dirstate.
 
1626
    """
 
1627
    # this could be generalized to allow comparisons between any trees in the
 
1628
    # dirstate, and possibly between trees stored in different dirstates.
 
1629
 
 
1630
    def __init__(self, source, target):
 
1631
        super(InterDirStateTree, self).__init__(source, target)
 
1632
        if not InterDirStateTree.is_compatible(source, target):
 
1633
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1634
 
 
1635
    @staticmethod
 
1636
    def make_source_parent_tree(source, target):
 
1637
        """Change the source tree into a parent of the target."""
 
1638
        revid = source.commit('record tree')
 
1639
        target.branch.repository.fetch(source.branch.repository, revid)
 
1640
        target.set_parent_ids([revid])
 
1641
        return target.basis_tree(), target
 
1642
 
 
1643
    _matching_from_tree_format = WorkingTreeFormat4()
 
1644
    _matching_to_tree_format = WorkingTreeFormat4()
 
1645
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1646
 
 
1647
    def _iter_changes(self, include_unchanged=False,
 
1648
                      specific_files=None, pb=None, extra_trees=[],
 
1649
                      require_versioned=True, want_unversioned=False):
 
1650
        """Return the changes from source to target.
 
1651
 
 
1652
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1653
            for details.
 
1654
        :param specific_files: An optional list of file paths to restrict the
 
1655
            comparison to. When mapping filenames to ids, all matches in all
 
1656
            trees (including optional extra_trees) are used, and all children of
 
1657
            matched directories are included.
 
1658
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1659
            unchanged entries in the result.
 
1660
        :param extra_trees: An optional list of additional trees to use when
 
1661
            mapping the contents of specific_files (paths) to file_ids.
 
1662
        :param require_versioned: If True, all files in specific_files must be
 
1663
            versioned in one of source, target, extra_trees or
 
1664
            PathsNotVersionedError is raised.
 
1665
        :param want_unversioned: Should unversioned files be returned in the
 
1666
            output. An unversioned file is defined as one with (False, False)
 
1667
            for the versioned pair.
 
1668
        """
 
1669
        utf8_decode_or_none = cache_utf8._utf8_decode_with_None
 
1670
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1671
        # NB: show_status depends on being able to pass in non-versioned files
 
1672
        # and report them as unknown
 
1673
        # TODO: handle extra trees in the dirstate.
 
1674
        # TODO: handle comparisons as an empty tree as a different special
 
1675
        # case? mbp 20070226
 
1676
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1677
            # we can't fast-path these cases (yet)
 
1678
            for f in super(InterDirStateTree, self)._iter_changes(
 
1679
                include_unchanged, specific_files, pb, extra_trees,
 
1680
                require_versioned, want_unversioned=want_unversioned):
 
1681
                yield f
 
1682
            return
 
1683
        parent_ids = self.target.get_parent_ids()
 
1684
        assert (self.source._revision_id in parent_ids), \
 
1685
                "revision {%s} is not stored in {%s}, but %s " \
 
1686
                "can only be used for trees stored in the dirstate" \
 
1687
                % (self.source._revision_id, self.target, self._iter_changes)
 
1688
        target_index = 0
 
1689
        if self.source._revision_id == NULL_REVISION:
 
1690
            source_index = None
 
1691
            indices = (target_index,)
 
1692
        else:
 
1693
            assert (self.source._revision_id in parent_ids), \
 
1694
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1695
                self.source._revision_id, parent_ids)
 
1696
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1697
            indices = (source_index,target_index)
 
1698
        # -- make all specific_files utf8 --
 
1699
        if specific_files:
 
1700
            specific_files_utf8 = set()
 
1701
            for path in specific_files:
 
1702
                specific_files_utf8.add(path.encode('utf8'))
 
1703
            specific_files = specific_files_utf8
 
1704
        else:
 
1705
            specific_files = set([''])
 
1706
        # -- specific_files is now a utf8 path set --
 
1707
        # -- get the state object and prepare it.
 
1708
        state = self.target.current_dirstate()
 
1709
        state._read_dirblocks_if_needed()
 
1710
        def _entries_for_path(path):
 
1711
            """Return a list with all the entries that match path for all ids.
 
1712
            """
 
1713
            dirname, basename = os.path.split(path)
 
1714
            key = (dirname, basename, '')
 
1715
            block_index, present = state._find_block_index_from_key(key)
 
1716
            if not present:
 
1717
                # the block which should contain path is absent.
 
1718
                return []
 
1719
            result = []
 
1720
            block = state._dirblocks[block_index][1]
 
1721
            entry_index, _ = state._find_entry_index(key, block)
 
1722
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1723
            while (entry_index < len(block) and
 
1724
                block[entry_index][0][0:2] == key[0:2]):
 
1725
                result.append(block[entry_index])
 
1726
                entry_index += 1
 
1727
            return result
 
1728
        if require_versioned:
 
1729
            # -- check all supplied paths are versioned in a search tree. --
 
1730
            all_versioned = True
 
1731
            for path in specific_files:
 
1732
                path_entries = _entries_for_path(path)
 
1733
                if not path_entries:
 
1734
                    # this specified path is not present at all: error
 
1735
                    all_versioned = False
 
1736
                    break
 
1737
                found_versioned = False
 
1738
                # for each id at this path
 
1739
                for entry in path_entries:
 
1740
                    # for each tree.
 
1741
                    for index in indices:
 
1742
                        if entry[1][index][0] != 'a': # absent
 
1743
                            found_versioned = True
 
1744
                            # all good: found a versioned cell
 
1745
                            break
 
1746
                if not found_versioned:
 
1747
                    # none of the indexes was not 'absent' at all ids for this
 
1748
                    # path.
 
1749
                    all_versioned = False
 
1750
                    break
 
1751
            if not all_versioned:
 
1752
                raise errors.PathsNotVersionedError(specific_files)
 
1753
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1754
        search_specific_files = set()
 
1755
        for path in specific_files:
 
1756
            other_specific_files = specific_files.difference(set([path]))
 
1757
            if not osutils.is_inside_any(other_specific_files, path):
 
1758
                # this is a top level path, we must check it.
 
1759
                search_specific_files.add(path)
 
1760
        # sketch: 
 
1761
        # compare source_index and target_index at or under each element of search_specific_files.
 
1762
        # follow the following comparison table. Note that we only want to do diff operations when
 
1763
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1764
        # for the target.
 
1765
        # cases:
 
1766
        # 
 
1767
        # Source | Target | disk | action
 
1768
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1769
        #        |        |      | diff check on source-target
 
1770
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1771
        #        |        |      | ???
 
1772
        #   r    |  a     |      | add source to search
 
1773
        #   r    |  a     |  a   | 
 
1774
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1775
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1776
        #   a    | fdlt   |      | add new id
 
1777
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1778
        #   a    |  a     |      | not present in either tree, skip
 
1779
        #   a    |  a     |  a   | not present in any tree, skip
 
1780
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1781
        #        |        |      | may not be selected by the users list of paths.
 
1782
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1783
        #        |        |      | may not be selected by the users list of paths.
 
1784
        #  fdlt  | fdlt   |      | content in both: diff them
 
1785
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1786
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1787
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1788
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1789
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1790
        #        |        |      | this id at the other path.
 
1791
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1792
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1793
        #        |        |      | this id at the other path.
 
1794
 
 
1795
        # for all search_indexs in each path at or under each element of
 
1796
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1797
        # relocated path as one to search if its not searched already. If the
 
1798
        # detail is not relocated, add the id.
 
1799
        searched_specific_files = set()
 
1800
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1801
        # Using a list so that we can access the values and change them in
 
1802
        # nested scope. Each one is [path, file_id, entry]
 
1803
        last_source_parent = [None, None, None]
 
1804
        last_target_parent = [None, None, None]
 
1805
 
 
1806
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1807
 
 
1808
        def _process_entry(entry, path_info):
 
1809
            """Compare an entry and real disk to generate delta information.
 
1810
 
 
1811
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1812
                the path of entry. If None, then the path is considered absent.
 
1813
                (Perhaps we should pass in a concrete entry for this ?)
 
1814
                Basename is returned as a utf8 string because we expect this
 
1815
                tuple will be ignored, and don't want to take the time to
 
1816
                decode.
 
1817
            """
 
1818
            if source_index is None:
 
1819
                source_details = NULL_PARENT_DETAILS
 
1820
            else:
 
1821
                source_details = entry[1][source_index]
 
1822
            target_details = entry[1][target_index]
 
1823
            target_minikind = target_details[0]
 
1824
            if path_info is not None and target_minikind in 'fdlt':
 
1825
                assert target_index == 0
 
1826
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1827
                                                  stat_value=path_info[3])
 
1828
                # The entry may have been modified by update_entry
 
1829
                target_details = entry[1][target_index]
 
1830
                target_minikind = target_details[0]
 
1831
            else:
 
1832
                link_or_sha1 = None
 
1833
            source_minikind = source_details[0]
 
1834
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1835
                # claimed content in both: diff
 
1836
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1837
                #        |        |      | diff check on source-target
 
1838
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1839
                #        |        |      | ???
 
1840
                if source_minikind in 'r':
 
1841
                    # add the source to the search path to find any children it
 
1842
                    # has.  TODO ? : only add if it is a container ?
 
1843
                    if not osutils.is_inside_any(searched_specific_files,
 
1844
                                                 source_details[1]):
 
1845
                        search_specific_files.add(source_details[1])
 
1846
                    # generate the old path; this is needed for stating later
 
1847
                    # as well.
 
1848
                    old_path = source_details[1]
 
1849
                    old_dirname, old_basename = os.path.split(old_path)
 
1850
                    path = pathjoin(entry[0][0], entry[0][1])
 
1851
                    old_entry = state._get_entry(source_index,
 
1852
                                                 path_utf8=old_path)
 
1853
                    # update the source details variable to be the real
 
1854
                    # location.
 
1855
                    source_details = old_entry[1][source_index]
 
1856
                    source_minikind = source_details[0]
 
1857
                else:
 
1858
                    old_dirname = entry[0][0]
 
1859
                    old_basename = entry[0][1]
 
1860
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1861
                if path_info is None:
 
1862
                    # the file is missing on disk, show as removed.
 
1863
                    content_change = True
 
1864
                    target_kind = None
 
1865
                    target_exec = False
 
1866
                else:
 
1867
                    # source and target are both versioned and disk file is present.
 
1868
                    target_kind = path_info[2]
 
1869
                    if target_kind == 'directory':
 
1870
                        if source_minikind != 'd':
 
1871
                            content_change = True
 
1872
                        else:
 
1873
                            # directories have no fingerprint
 
1874
                            content_change = False
 
1875
                        target_exec = False
 
1876
                    elif target_kind == 'file':
 
1877
                        if source_minikind != 'f':
 
1878
                            content_change = True
 
1879
                        else:
 
1880
                            # We could check the size, but we already have the
 
1881
                            # sha1 hash.
 
1882
                            content_change = (link_or_sha1 != source_details[1])
 
1883
                        # Target details is updated at update_entry time
 
1884
                        if use_filesystem_for_exec:
 
1885
                            # We don't need S_ISREG here, because we are sure
 
1886
                            # we are dealing with a file.
 
1887
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1888
                        else:
 
1889
                            target_exec = target_details[3]
 
1890
                    elif target_kind == 'symlink':
 
1891
                        if source_minikind != 'l':
 
1892
                            content_change = True
 
1893
                        else:
 
1894
                            content_change = (link_or_sha1 != source_details[1])
 
1895
                        target_exec = False
 
1896
                    elif target_kind == 'tree-reference':
 
1897
                        if source_minikind != 't':
 
1898
                            content_change = True
 
1899
                        else:
 
1900
                            content_change = False
 
1901
                        target_exec = False
 
1902
                    else:
 
1903
                        raise Exception, "unknown kind %s" % path_info[2]
 
1904
                # parent id is the entry for the path in the target tree
 
1905
                if old_dirname == last_source_parent[0]:
 
1906
                    source_parent_id = last_source_parent[1]
 
1907
                else:
 
1908
                    source_parent_entry = state._get_entry(source_index,
 
1909
                                                           path_utf8=old_dirname)
 
1910
                    source_parent_id = source_parent_entry[0][2]
 
1911
                    if source_parent_id == entry[0][2]:
 
1912
                        # This is the root, so the parent is None
 
1913
                        source_parent_id = None
 
1914
                    else:
 
1915
                        last_source_parent[0] = old_dirname
 
1916
                        last_source_parent[1] = source_parent_id
 
1917
                        last_source_parent[2] = source_parent_entry
 
1918
                new_dirname = entry[0][0]
 
1919
                if new_dirname == last_target_parent[0]:
 
1920
                    target_parent_id = last_target_parent[1]
 
1921
                else:
 
1922
                    # TODO: We don't always need to do the lookup, because the
 
1923
                    #       parent entry will be the same as the source entry.
 
1924
                    target_parent_entry = state._get_entry(target_index,
 
1925
                                                           path_utf8=new_dirname)
 
1926
                    assert target_parent_entry != (None, None), (
 
1927
                        "Could not find target parent in wt: %s\nparent of: %s"
 
1928
                        % (new_dirname, entry))
 
1929
                    target_parent_id = target_parent_entry[0][2]
 
1930
                    if target_parent_id == entry[0][2]:
 
1931
                        # This is the root, so the parent is None
 
1932
                        target_parent_id = None
 
1933
                    else:
 
1934
                        last_target_parent[0] = new_dirname
 
1935
                        last_target_parent[1] = target_parent_id
 
1936
                        last_target_parent[2] = target_parent_entry
 
1937
 
 
1938
                source_exec = source_details[3]
 
1939
                return ((entry[0][2], (old_path, path), content_change,
 
1940
                        (True, True),
 
1941
                        (source_parent_id, target_parent_id),
 
1942
                        (old_basename, entry[0][1]),
 
1943
                        (_minikind_to_kind[source_minikind], target_kind),
 
1944
                        (source_exec, target_exec)),)
 
1945
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1946
                # looks like a new file
 
1947
                if path_info is not None:
 
1948
                    path = pathjoin(entry[0][0], entry[0][1])
 
1949
                    # parent id is the entry for the path in the target tree
 
1950
                    # TODO: these are the same for an entire directory: cache em.
 
1951
                    parent_id = state._get_entry(target_index,
 
1952
                                                 path_utf8=entry[0][0])[0][2]
 
1953
                    if parent_id == entry[0][2]:
 
1954
                        parent_id = None
 
1955
                    if use_filesystem_for_exec:
 
1956
                        # We need S_ISREG here, because we aren't sure if this
 
1957
                        # is a file or not.
 
1958
                        target_exec = bool(
 
1959
                            stat.S_ISREG(path_info[3].st_mode)
 
1960
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1961
                    else:
 
1962
                        target_exec = target_details[3]
 
1963
                    return ((entry[0][2], (None, path), True,
 
1964
                            (False, True),
 
1965
                            (None, parent_id),
 
1966
                            (None, entry[0][1]),
 
1967
                            (None, path_info[2]),
 
1968
                            (None, target_exec)),)
 
1969
                else:
 
1970
                    # but its not on disk: we deliberately treat this as just
 
1971
                    # never-present. (Why ?! - RBC 20070224)
 
1972
                    pass
 
1973
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1974
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1975
                # if its still on disk, *and* theres no other entry at this
 
1976
                # path [we dont know this in this routine at the moment -
 
1977
                # perhaps we should change this - then it would be an unknown.
 
1978
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1979
                # parent id is the entry for the path in the target tree
 
1980
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1981
                if parent_id == entry[0][2]:
 
1982
                    parent_id = None
 
1983
                return ((entry[0][2], (old_path, None), True,
 
1984
                        (True, False),
 
1985
                        (parent_id, None),
 
1986
                        (entry[0][1], None),
 
1987
                        (_minikind_to_kind[source_minikind], None),
 
1988
                        (source_details[3], None)),)
 
1989
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1990
                # a rename; could be a true rename, or a rename inherited from
 
1991
                # a renamed parent. TODO: handle this efficiently. Its not
 
1992
                # common case to rename dirs though, so a correct but slow
 
1993
                # implementation will do.
 
1994
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1995
                    search_specific_files.add(target_details[1])
 
1996
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
1997
                # neither of the selected trees contain this file,
 
1998
                # so skip over it. This is not currently directly tested, but
 
1999
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2000
                pass
 
2001
            else:
 
2002
                raise AssertionError("don't know how to compare "
 
2003
                    "source_minikind=%r, target_minikind=%r"
 
2004
                    % (source_minikind, target_minikind))
 
2005
                ## import pdb;pdb.set_trace()
 
2006
            return ()
 
2007
 
 
2008
        while search_specific_files:
 
2009
            # TODO: the pending list should be lexically sorted?  the
 
2010
            # interface doesn't require it.
 
2011
            current_root = search_specific_files.pop()
 
2012
            current_root_unicode = current_root.decode('utf8')
 
2013
            searched_specific_files.add(current_root)
 
2014
            # process the entries for this containing directory: the rest will be
 
2015
            # found by their parents recursively.
 
2016
            root_entries = _entries_for_path(current_root)
 
2017
            root_abspath = self.target.abspath(current_root_unicode)
 
2018
            try:
 
2019
                root_stat = os.lstat(root_abspath)
 
2020
            except OSError, e:
 
2021
                if e.errno == errno.ENOENT:
 
2022
                    # the path does not exist: let _process_entry know that.
 
2023
                    root_dir_info = None
 
2024
                else:
 
2025
                    # some other random error: hand it up.
 
2026
                    raise
 
2027
            else:
 
2028
                root_dir_info = ('', current_root,
 
2029
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2030
                    root_abspath)
 
2031
                if root_dir_info[2] == 'directory':
 
2032
                    if self.target._directory_is_tree_reference(
 
2033
                        current_root.decode('utf8')):
 
2034
                        root_dir_info = root_dir_info[:2] + \
 
2035
                            ('tree-reference',) + root_dir_info[3:]
 
2036
 
 
2037
            if not root_entries and not root_dir_info:
 
2038
                # this specified path is not present at all, skip it.
 
2039
                continue
 
2040
            path_handled = False
 
2041
            for entry in root_entries:
 
2042
                for result in _process_entry(entry, root_dir_info):
 
2043
                    # this check should probably be outside the loop: one
 
2044
                    # 'iterate two trees' api, and then _iter_changes filters
 
2045
                    # unchanged pairs. - RBC 20070226
 
2046
                    path_handled = True
 
2047
                    if (include_unchanged
 
2048
                        or result[2]                    # content change
 
2049
                        or result[3][0] != result[3][1] # versioned status
 
2050
                        or result[4][0] != result[4][1] # parent id
 
2051
                        or result[5][0] != result[5][1] # name
 
2052
                        or result[6][0] != result[6][1] # kind
 
2053
                        or result[7][0] != result[7][1] # executable
 
2054
                        ):
 
2055
                        yield (result[0],
 
2056
                               (utf8_decode_or_none(result[1][0]),
 
2057
                                utf8_decode_or_none(result[1][1])),
 
2058
                               result[2],
 
2059
                               result[3],
 
2060
                               result[4],
 
2061
                               (utf8_decode_or_none(result[5][0]),
 
2062
                                utf8_decode_or_none(result[5][1])),
 
2063
                               result[6],
 
2064
                               result[7],
 
2065
                              )
 
2066
            if want_unversioned and not path_handled and root_dir_info:
 
2067
                new_executable = bool(
 
2068
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2069
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2070
                yield (None,
 
2071
                       (None, current_root_unicode),
 
2072
                       True,
 
2073
                       (False, False),
 
2074
                       (None, None),
 
2075
                       (None, splitpath(current_root_unicode)[-1]),
 
2076
                       (None, root_dir_info[2]),
 
2077
                       (None, new_executable)
 
2078
                      )
 
2079
            initial_key = (current_root, '', '')
 
2080
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2081
            if block_index == 0:
 
2082
                # we have processed the total root already, but because the
 
2083
                # initial key matched it we should skip it here.
 
2084
                block_index +=1
 
2085
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2086
                current_dir_info = None
 
2087
            else:
 
2088
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2089
                try:
 
2090
                    current_dir_info = dir_iterator.next()
 
2091
                except OSError, e:
 
2092
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2093
                    # python 2.5 has e.errno == EINVAL,
 
2094
                    #            and e.winerror == ERROR_DIRECTORY
 
2095
                    e_winerror = getattr(e, 'winerror', None)
 
2096
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2097
                    # there may be directories in the inventory even though
 
2098
                    # this path is not a file on disk: so mark it as end of
 
2099
                    # iterator
 
2100
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2101
                        current_dir_info = None
 
2102
                    elif (sys.platform == 'win32'
 
2103
                          and (e.errno in win_errors
 
2104
                               or e_winerror in win_errors)):
 
2105
                        current_dir_info = None
 
2106
                    else:
 
2107
                        raise
 
2108
                else:
 
2109
                    if current_dir_info[0][0] == '':
 
2110
                        # remove .bzr from iteration
 
2111
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2112
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2113
                        del current_dir_info[1][bzr_index]
 
2114
            # walk until both the directory listing and the versioned metadata
 
2115
            # are exhausted. 
 
2116
            if (block_index < len(state._dirblocks) and
 
2117
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2118
                current_block = state._dirblocks[block_index]
 
2119
            else:
 
2120
                current_block = None
 
2121
            while (current_dir_info is not None or
 
2122
                   current_block is not None):
 
2123
                if (current_dir_info and current_block
 
2124
                    and current_dir_info[0][0] != current_block[0]):
 
2125
                    if current_dir_info[0][0].split('/') < current_block[0].split('/'):
 
2126
                        # filesystem data refers to paths not covered by the dirblock.
 
2127
                        # this has two possibilities:
 
2128
                        # A) it is versioned but empty, so there is no block for it
 
2129
                        # B) it is not versioned.
 
2130
 
 
2131
                        # if (A) then we need to recurse into it to check for
 
2132
                        # new unknown files or directories.
 
2133
                        # if (B) then we should ignore it, because we don't
 
2134
                        # recurse into unknown directories.
 
2135
                        path_index = 0
 
2136
                        while path_index < len(current_dir_info[1]):
 
2137
                                current_path_info = current_dir_info[1][path_index]
 
2138
                                if want_unversioned:
 
2139
                                    if current_path_info[2] == 'directory':
 
2140
                                        if self.target._directory_is_tree_reference(
 
2141
                                            current_path_info[0].decode('utf8')):
 
2142
                                            current_path_info = current_path_info[:2] + \
 
2143
                                                ('tree-reference',) + current_path_info[3:]
 
2144
                                    new_executable = bool(
 
2145
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2146
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2147
                                    yield (None,
 
2148
                                        (None, utf8_decode_or_none(current_path_info[0])),
 
2149
                                        True,
 
2150
                                        (False, False),
 
2151
                                        (None, None),
 
2152
                                        (None, utf8_decode_or_none(current_path_info[1])),
 
2153
                                        (None, current_path_info[2]),
 
2154
                                        (None, new_executable))
 
2155
                                # dont descend into this unversioned path if it is
 
2156
                                # a dir
 
2157
                                if current_path_info[2] in ('directory',
 
2158
                                                            'tree-reference'):
 
2159
                                    del current_dir_info[1][path_index]
 
2160
                                    path_index -= 1
 
2161
                                path_index += 1
 
2162
 
 
2163
                        # This dir info has been handled, go to the next
 
2164
                        try:
 
2165
                            current_dir_info = dir_iterator.next()
 
2166
                        except StopIteration:
 
2167
                            current_dir_info = None
 
2168
                    else:
 
2169
                        # We have a dirblock entry for this location, but there
 
2170
                        # is no filesystem path for this. This is most likely
 
2171
                        # because a directory was removed from the disk.
 
2172
                        # We don't have to report the missing directory,
 
2173
                        # because that should have already been handled, but we
 
2174
                        # need to handle all of the files that are contained
 
2175
                        # within.
 
2176
                        for current_entry in current_block[1]:
 
2177
                            # entry referring to file not present on disk.
 
2178
                            # advance the entry only, after processing.
 
2179
                            for result in _process_entry(current_entry, None):
 
2180
                                # this check should probably be outside the loop: one
 
2181
                                # 'iterate two trees' api, and then _iter_changes filters
 
2182
                                # unchanged pairs. - RBC 20070226
 
2183
                                if (include_unchanged
 
2184
                                    or result[2]                    # content change
 
2185
                                    or result[3][0] != result[3][1] # versioned status
 
2186
                                    or result[4][0] != result[4][1] # parent id
 
2187
                                    or result[5][0] != result[5][1] # name
 
2188
                                    or result[6][0] != result[6][1] # kind
 
2189
                                    or result[7][0] != result[7][1] # executable
 
2190
                                    ):
 
2191
                                    yield (result[0],
 
2192
                                           (utf8_decode_or_none(result[1][0]),
 
2193
                                            utf8_decode_or_none(result[1][1])),
 
2194
                                           result[2],
 
2195
                                           result[3],
 
2196
                                           result[4],
 
2197
                                           (utf8_decode_or_none(result[5][0]),
 
2198
                                            utf8_decode_or_none(result[5][1])),
 
2199
                                           result[6],
 
2200
                                           result[7],
 
2201
                                          )
 
2202
                        block_index +=1
 
2203
                        if (block_index < len(state._dirblocks) and
 
2204
                            osutils.is_inside(current_root,
 
2205
                                              state._dirblocks[block_index][0])):
 
2206
                            current_block = state._dirblocks[block_index]
 
2207
                        else:
 
2208
                            current_block = None
 
2209
                    continue
 
2210
                entry_index = 0
 
2211
                if current_block and entry_index < len(current_block[1]):
 
2212
                    current_entry = current_block[1][entry_index]
 
2213
                else:
 
2214
                    current_entry = None
 
2215
                advance_entry = True
 
2216
                path_index = 0
 
2217
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2218
                    current_path_info = current_dir_info[1][path_index]
 
2219
                    if current_path_info[2] == 'directory':
 
2220
                        if self.target._directory_is_tree_reference(
 
2221
                            current_path_info[0].decode('utf8')):
 
2222
                            current_path_info = current_path_info[:2] + \
 
2223
                                ('tree-reference',) + current_path_info[3:]
 
2224
                else:
 
2225
                    current_path_info = None
 
2226
                advance_path = True
 
2227
                path_handled = False
 
2228
                while (current_entry is not None or
 
2229
                    current_path_info is not None):
 
2230
                    if current_entry is None:
 
2231
                        # the check for path_handled when the path is adnvaced
 
2232
                        # will yield this path if needed.
 
2233
                        pass
 
2234
                    elif current_path_info is None:
 
2235
                        # no path is fine: the per entry code will handle it.
 
2236
                        for result in _process_entry(current_entry, current_path_info):
 
2237
                            # this check should probably be outside the loop: one
 
2238
                            # 'iterate two trees' api, and then _iter_changes filters
 
2239
                            # unchanged pairs. - RBC 20070226
 
2240
                            if (include_unchanged
 
2241
                                or result[2]                    # content change
 
2242
                                or result[3][0] != result[3][1] # versioned status
 
2243
                                or result[4][0] != result[4][1] # parent id
 
2244
                                or result[5][0] != result[5][1] # name
 
2245
                                or result[6][0] != result[6][1] # kind
 
2246
                                or result[7][0] != result[7][1] # executable
 
2247
                                ):
 
2248
                                yield (result[0],
 
2249
                                       (utf8_decode_or_none(result[1][0]),
 
2250
                                        utf8_decode_or_none(result[1][1])),
 
2251
                                       result[2],
 
2252
                                       result[3],
 
2253
                                       result[4],
 
2254
                                       (utf8_decode_or_none(result[5][0]),
 
2255
                                        utf8_decode_or_none(result[5][1])),
 
2256
                                       result[6],
 
2257
                                       result[7],
 
2258
                                      )
 
2259
                    elif (current_entry[0][1] != current_path_info[1]
 
2260
                          or current_entry[1][target_index][0] in 'ar'):
 
2261
                        # The current path on disk doesn't match the dirblock
 
2262
                        # record. Either the dirblock is marked as absent, or
 
2263
                        # the file on disk is not present at all in the
 
2264
                        # dirblock. Either way, report about the dirblock
 
2265
                        # entry, and let other code handle the filesystem one.
 
2266
                        if current_path_info[1].split('/') < current_entry[0][1].split('/'):
 
2267
                            # extra file on disk: pass for now, but only
 
2268
                            # increment the path, not the entry
 
2269
                            advance_entry = False
 
2270
                        else:
 
2271
                            # entry referring to file not present on disk.
 
2272
                            # advance the entry only, after processing.
 
2273
                            for result in _process_entry(current_entry, None):
 
2274
                                # this check should probably be outside the loop: one
 
2275
                                # 'iterate two trees' api, and then _iter_changes filters
 
2276
                                # unchanged pairs. - RBC 20070226
 
2277
                                if (include_unchanged
 
2278
                                    or result[2]                    # content change
 
2279
                                    or result[3][0] != result[3][1] # versioned status
 
2280
                                    or result[4][0] != result[4][1] # parent id
 
2281
                                    or result[5][0] != result[5][1] # name
 
2282
                                    or result[6][0] != result[6][1] # kind
 
2283
                                    or result[7][0] != result[7][1] # executable
 
2284
                                    ):
 
2285
                                    yield (result[0],
 
2286
                                           (utf8_decode_or_none(result[1][0]),
 
2287
                                            utf8_decode_or_none(result[1][1])),
 
2288
                                           result[2],
 
2289
                                           result[3],
 
2290
                                           result[4],
 
2291
                                           (utf8_decode_or_none(result[5][0]),
 
2292
                                            utf8_decode_or_none(result[5][1])),
 
2293
                                           result[6],
 
2294
                                           result[7],
 
2295
                                          )
 
2296
                            advance_path = False
 
2297
                    else:
 
2298
                        for result in _process_entry(current_entry, current_path_info):
 
2299
                            # this check should probably be outside the loop: one
 
2300
                            # 'iterate two trees' api, and then _iter_changes filters
 
2301
                            # unchanged pairs. - RBC 20070226
 
2302
                            path_handled = True
 
2303
                            if (include_unchanged
 
2304
                                or result[2]                    # content change
 
2305
                                or result[3][0] != result[3][1] # versioned status
 
2306
                                or result[4][0] != result[4][1] # parent id
 
2307
                                or result[5][0] != result[5][1] # name
 
2308
                                or result[6][0] != result[6][1] # kind
 
2309
                                or result[7][0] != result[7][1] # executable
 
2310
                                ):
 
2311
                                yield (result[0],
 
2312
                                       (utf8_decode_or_none(result[1][0]),
 
2313
                                        utf8_decode_or_none(result[1][1])),
 
2314
                                       result[2],
 
2315
                                       result[3],
 
2316
                                       result[4],
 
2317
                                       (utf8_decode_or_none(result[5][0]),
 
2318
                                        utf8_decode_or_none(result[5][1])),
 
2319
                                       result[6],
 
2320
                                       result[7],
 
2321
                                      )
 
2322
                    if advance_entry and current_entry is not None:
 
2323
                        entry_index += 1
 
2324
                        if entry_index < len(current_block[1]):
 
2325
                            current_entry = current_block[1][entry_index]
 
2326
                        else:
 
2327
                            current_entry = None
 
2328
                    else:
 
2329
                        advance_entry = True # reset the advance flaga
 
2330
                    if advance_path and current_path_info is not None:
 
2331
                        if not path_handled:
 
2332
                            # unversioned in all regards
 
2333
                            if want_unversioned:
 
2334
                                new_executable = bool(
 
2335
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2336
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2337
                                yield (None,
 
2338
                                    (None, utf8_decode_or_none(current_path_info[0])),
 
2339
                                    True,
 
2340
                                    (False, False),
 
2341
                                    (None, None),
 
2342
                                    (None, utf8_decode_or_none(current_path_info[1])),
 
2343
                                    (None, current_path_info[2]),
 
2344
                                    (None, new_executable))
 
2345
                            # dont descend into this unversioned path if it is
 
2346
                            # a dir
 
2347
                            if current_path_info[2] in ('directory'):
 
2348
                                del current_dir_info[1][path_index]
 
2349
                                path_index -= 1
 
2350
                        # dont descend the disk iterator into any tree 
 
2351
                        # paths.
 
2352
                        if current_path_info[2] == 'tree-reference':
 
2353
                            del current_dir_info[1][path_index]
 
2354
                            path_index -= 1
 
2355
                        path_index += 1
 
2356
                        if path_index < len(current_dir_info[1]):
 
2357
                            current_path_info = current_dir_info[1][path_index]
 
2358
                            if current_path_info[2] == 'directory':
 
2359
                                if self.target._directory_is_tree_reference(
 
2360
                                    current_path_info[0].decode('utf8')):
 
2361
                                    current_path_info = current_path_info[:2] + \
 
2362
                                        ('tree-reference',) + current_path_info[3:]
 
2363
                        else:
 
2364
                            current_path_info = None
 
2365
                        path_handled = False
 
2366
                    else:
 
2367
                        advance_path = True # reset the advance flagg.
 
2368
                if current_block is not None:
 
2369
                    block_index += 1
 
2370
                    if (block_index < len(state._dirblocks) and
 
2371
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2372
                        current_block = state._dirblocks[block_index]
 
2373
                    else:
 
2374
                        current_block = None
 
2375
                if current_dir_info is not None:
 
2376
                    try:
 
2377
                        current_dir_info = dir_iterator.next()
 
2378
                    except StopIteration:
 
2379
                        current_dir_info = None
 
2380
 
 
2381
 
 
2382
    @staticmethod
 
2383
    def is_compatible(source, target):
 
2384
        # the target must be a dirstate working tree
 
2385
        if not isinstance(target, WorkingTree4):
 
2386
            return False
 
2387
        # the source must be a revtreee or dirstate rev tree.
 
2388
        if not isinstance(source,
 
2389
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2390
            return False
 
2391
        # the source revid must be in the target dirstate
 
2392
        if not (source._revision_id == NULL_REVISION or
 
2393
            source._revision_id in target.get_parent_ids()):
 
2394
            # TODO: what about ghosts? it may well need to 
 
2395
            # check for them explicitly.
 
2396
            return False
 
2397
        return True
 
2398
 
 
2399
InterTree.register_optimiser(InterDirStateTree)
 
2400
 
 
2401
 
 
2402
class Converter3to4(object):
 
2403
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2404
 
 
2405
    def __init__(self):
 
2406
        self.target_format = WorkingTreeFormat4()
 
2407
 
 
2408
    def convert(self, tree):
 
2409
        # lock the control files not the tree, so that we dont get tree
 
2410
        # on-unlock behaviours, and so that noone else diddles with the 
 
2411
        # tree during upgrade.
 
2412
        tree._control_files.lock_write()
 
2413
        try:
 
2414
            tree.read_working_inventory()
 
2415
            self.create_dirstate_data(tree)
 
2416
            self.update_format(tree)
 
2417
            self.remove_xml_files(tree)
 
2418
        finally:
 
2419
            tree._control_files.unlock()
 
2420
 
 
2421
    def create_dirstate_data(self, tree):
 
2422
        """Create the dirstate based data for tree."""
 
2423
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2424
            ).local_abspath('dirstate')
 
2425
        state = dirstate.DirState.from_tree(tree, local_path)
 
2426
        state.save()
 
2427
        state.unlock()
 
2428
 
 
2429
    def remove_xml_files(self, tree):
 
2430
        """Remove the oldformat 3 data."""
 
2431
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2432
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2433
            'pending-merges', 'stat-cache']:
 
2434
            try:
 
2435
                transport.delete(path)
 
2436
            except errors.NoSuchFile:
 
2437
                # some files are optional - just deal.
 
2438
                pass
 
2439
 
 
2440
    def update_format(self, tree):
 
2441
        """Change the format marker."""
 
2442
        tree._control_files.put_utf8('format',
 
2443
            self.target_format.get_format_string())