/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: John Arbash Meinel
  • Date: 2007-03-09 00:52:52 UTC
  • mto: This revision was merged to the branch mainline in revision 2351.
  • Revision ID: john@arbash-meinel.com-20070309005252-4heez38fmg2qsfqy
Rather than setting the executable bit for every file, properly create the working inventory to include it.

Show diffs side-by-side

added added

removed removed

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