~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Martin Pool
  • Date: 2007-09-19 13:12:32 UTC
  • mto: (2592.3.153 repository)
  • mto: This revision was merged to the branch mainline in revision 2933.
  • Revision ID: mbp@sourcefrog.net-20070919131232-0gtp1q90fxz10ctn
move commit_write_group to RepositoryPackCollection

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
    debug,
 
47
    delta,
 
48
    dirstate,
 
49
    errors,
 
50
    generate_ids,
 
51
    globbing,
 
52
    hashcache,
 
53
    ignores,
 
54
    merge,
 
55
    osutils,
 
56
    revision as _mod_revision,
 
57
    revisiontree,
 
58
    textui,
 
59
    trace,
 
60
    transform,
 
61
    urlutils,
 
62
    xml5,
 
63
    xml6,
 
64
    )
 
65
import bzrlib.branch
 
66
from bzrlib.transport import get_transport
 
67
import bzrlib.ui
 
68
""")
 
69
 
 
70
from bzrlib import symbol_versioning
 
71
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
72
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
73
from bzrlib.lockable_files import LockableFiles, TransportLock
 
74
from bzrlib.lockdir import LockDir
 
75
import bzrlib.mutabletree
 
76
from bzrlib.mutabletree import needs_tree_write_lock
 
77
from bzrlib.osutils import (
 
78
    file_kind,
 
79
    isdir,
 
80
    normpath,
 
81
    pathjoin,
 
82
    rand_chars,
 
83
    realpath,
 
84
    safe_unicode,
 
85
    splitpath,
 
86
    )
 
87
from bzrlib.trace import mutter, note
 
88
from bzrlib.transport.local import LocalTransport
 
89
from bzrlib.tree import InterTree
 
90
from bzrlib.progress import DummyProgress, ProgressPhase
 
91
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
92
from bzrlib.rio import RioReader, rio_file, Stanza
 
93
from bzrlib.symbol_versioning import (deprecated_passed,
 
94
        deprecated_method,
 
95
        deprecated_function,
 
96
        DEPRECATED_PARAMETER,
 
97
        )
 
98
from bzrlib.tree import Tree
 
99
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
100
 
 
101
 
 
102
# This is the Windows equivalent of ENOTDIR
 
103
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
104
# just an error code.
 
105
ERROR_PATH_NOT_FOUND = 3
 
106
ERROR_DIRECTORY = 267
 
107
 
 
108
 
 
109
class WorkingTree4(WorkingTree3):
 
110
    """This is the Format 4 working tree.
 
111
 
 
112
    This differs from WorkingTree3 by:
 
113
     - Having a consolidated internal dirstate, stored in a
 
114
       randomly-accessible sorted file on disk.
 
115
     - Not having a regular inventory attribute.  One can be synthesized 
 
116
       on demand but this is expensive and should be avoided.
 
117
 
 
118
    This is new in bzr 0.15.
 
119
    """
 
120
 
 
121
    def __init__(self, basedir,
 
122
                 branch,
 
123
                 _control_files=None,
 
124
                 _format=None,
 
125
                 _bzrdir=None):
 
126
        """Construct a WorkingTree for basedir.
 
127
 
 
128
        If the branch is not supplied, it is opened automatically.
 
129
        If the branch is supplied, it must be the branch for this basedir.
 
130
        (branch.base is not cross checked, because for remote branches that
 
131
        would be meaningless).
 
132
        """
 
133
        self._format = _format
 
134
        self.bzrdir = _bzrdir
 
135
        from bzrlib.trace import note, mutter
 
136
        assert isinstance(basedir, basestring), \
 
137
            "base directory %r is not a string" % basedir
 
138
        basedir = safe_unicode(basedir)
 
139
        mutter("opening working tree %r", basedir)
 
140
        self._branch = branch
 
141
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
142
            "branch %r is not a Branch" % self.branch
 
143
        self.basedir = realpath(basedir)
 
144
        # if branch is at our basedir and is a format 6 or less
 
145
        # assume all other formats have their own control files.
 
146
        assert isinstance(_control_files, LockableFiles), \
 
147
            "_control_files must be a LockableFiles, not %r" % _control_files
 
148
        self._control_files = _control_files
 
149
        self._dirty = None
 
150
        #-------------
 
151
        # during a read or write lock these objects are set, and are
 
152
        # None the rest of the time.
 
153
        self._dirstate = None
 
154
        self._inventory = None
 
155
        #-------------
 
156
 
 
157
    @needs_tree_write_lock
 
158
    def _add(self, files, ids, kinds):
 
159
        """See MutableTree._add."""
 
160
        state = self.current_dirstate()
 
161
        for f, file_id, kind in zip(files, ids, kinds):
 
162
            f = f.strip('/')
 
163
            assert '//' not in f
 
164
            assert '..' not in f
 
165
            if self.path2id(f):
 
166
                # special case tree root handling.
 
167
                if f == '' and self.path2id(f) == ROOT_ID:
 
168
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
169
                continue
 
170
            if file_id is None:
 
171
                file_id = generate_ids.gen_file_id(f)
 
172
            # deliberately add the file with no cached stat or sha1
 
173
            # - on the first access it will be gathered, and we can
 
174
            # always change this once tests are all passing.
 
175
            state.add(f, file_id, kind, None, '')
 
176
        self._make_dirty(reset_inventory=True)
 
177
 
 
178
    def _make_dirty(self, reset_inventory):
 
179
        """Make the tree state dirty.
 
180
 
 
181
        :param reset_inventory: True if the cached inventory should be removed
 
182
            (presuming there is one).
 
183
        """
 
184
        self._dirty = True
 
185
        if reset_inventory and self._inventory is not None:
 
186
            self._inventory = None
 
187
 
 
188
    @needs_tree_write_lock
 
189
    def add_reference(self, sub_tree):
 
190
        # use standard implementation, which calls back to self._add
 
191
        # 
 
192
        # So we don't store the reference_revision in the working dirstate,
 
193
        # it's just recorded at the moment of commit. 
 
194
        self._add_reference(sub_tree)
 
195
 
 
196
    def break_lock(self):
 
197
        """Break a lock if one is present from another instance.
 
198
 
 
199
        Uses the ui factory to ask for confirmation if the lock may be from
 
200
        an active process.
 
201
 
 
202
        This will probe the repository for its lock as well.
 
203
        """
 
204
        # if the dirstate is locked by an active process, reject the break lock
 
205
        # call.
 
206
        try:
 
207
            if self._dirstate is None:
 
208
                clear = True
 
209
            else:
 
210
                clear = False
 
211
            state = self._current_dirstate()
 
212
            if state._lock_token is not None:
 
213
                # we already have it locked. sheese, cant break our own lock.
 
214
                raise errors.LockActive(self.basedir)
 
215
            else:
 
216
                try:
 
217
                    # try for a write lock - need permission to get one anyhow
 
218
                    # to break locks.
 
219
                    state.lock_write()
 
220
                except errors.LockContention:
 
221
                    # oslocks fail when a process is still live: fail.
 
222
                    # TODO: get the locked lockdir info and give to the user to
 
223
                    # assist in debugging.
 
224
                    raise errors.LockActive(self.basedir)
 
225
                else:
 
226
                    state.unlock()
 
227
        finally:
 
228
            if clear:
 
229
                self._dirstate = None
 
230
        self._control_files.break_lock()
 
231
        self.branch.break_lock()
 
232
 
 
233
    def _comparison_data(self, entry, path):
 
234
        kind, executable, stat_value = \
 
235
            WorkingTree3._comparison_data(self, entry, path)
 
236
        # it looks like a plain directory, but it's really a reference -- see
 
237
        # also kind()
 
238
        if (self._repo_supports_tree_reference and
 
239
            kind == 'directory' and
 
240
            self._directory_is_tree_reference(path)):
 
241
            kind = 'tree-reference'
 
242
        return kind, executable, stat_value
 
243
 
 
244
    @needs_write_lock
 
245
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
246
        # mark the tree as dirty post commit - commit
 
247
        # can change the current versioned list by doing deletes.
 
248
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
249
        self._make_dirty(reset_inventory=True)
 
250
        return result
 
251
 
 
252
    def current_dirstate(self):
 
253
        """Return the current dirstate object.
 
254
 
 
255
        This is not part of the tree interface and only exposed for ease of
 
256
        testing.
 
257
 
 
258
        :raises errors.NotWriteLocked: when not in a lock.
 
259
        """
 
260
        self._must_be_locked()
 
261
        return self._current_dirstate()
 
262
 
 
263
    def _current_dirstate(self):
 
264
        """Internal function that does not check lock status.
 
265
 
 
266
        This is needed for break_lock which also needs the dirstate.
 
267
        """
 
268
        if self._dirstate is not None:
 
269
            return self._dirstate
 
270
        local_path = self.bzrdir.get_workingtree_transport(None
 
271
            ).local_abspath('dirstate')
 
272
        self._dirstate = dirstate.DirState.on_file(local_path)
 
273
        return self._dirstate
 
274
 
 
275
    def filter_unversioned_files(self, paths):
 
276
        """Filter out paths that are versioned.
 
277
 
 
278
        :return: set of paths.
 
279
        """
 
280
        # TODO: make a generic multi-bisect routine roughly that should list
 
281
        # the paths, then process one half at a time recursively, and feed the
 
282
        # results of each bisect in further still
 
283
        paths = sorted(paths)
 
284
        result = set()
 
285
        state = self.current_dirstate()
 
286
        # TODO we want a paths_to_dirblocks helper I think
 
287
        for path in paths:
 
288
            dirname, basename = os.path.split(path.encode('utf8'))
 
289
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
290
                dirname, basename, 0)
 
291
            if not path_is_versioned:
 
292
                result.add(path)
 
293
        return result
 
294
 
 
295
    def flush(self):
 
296
        """Write all cached data to disk."""
 
297
        if self._control_files._lock_mode != 'w':
 
298
            raise errors.NotWriteLocked(self)
 
299
        self.current_dirstate().save()
 
300
        self._inventory = None
 
301
        self._dirty = False
 
302
 
 
303
    @needs_tree_write_lock
 
304
    def _gather_kinds(self, files, kinds):
 
305
        """See MutableTree._gather_kinds."""
 
306
        for pos, f in enumerate(files):
 
307
            if kinds[pos] is None:
 
308
                kinds[pos] = self._kind(f)
 
309
 
 
310
    def _generate_inventory(self):
 
311
        """Create and set self.inventory from the dirstate object.
 
312
        
 
313
        This is relatively expensive: we have to walk the entire dirstate.
 
314
        Ideally we would not, and can deprecate this function.
 
315
        """
 
316
        #: uncomment to trap on inventory requests.
 
317
        # import pdb;pdb.set_trace()
 
318
        state = self.current_dirstate()
 
319
        state._read_dirblocks_if_needed()
 
320
        root_key, current_entry = self._get_entry(path='')
 
321
        current_id = root_key[2]
 
322
        assert current_entry[0][0] == 'd' # directory
 
323
        inv = Inventory(root_id=current_id)
 
324
        # Turn some things into local variables
 
325
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
326
        factory = entry_factory
 
327
        utf8_decode = cache_utf8._utf8_decode
 
328
        inv_byid = inv._byid
 
329
        # we could do this straight out of the dirstate; it might be fast
 
330
        # and should be profiled - RBC 20070216
 
331
        parent_ies = {'' : inv.root}
 
332
        for block in state._dirblocks[1:]: # skip the root
 
333
            dirname = block[0]
 
334
            try:
 
335
                parent_ie = parent_ies[dirname]
 
336
            except KeyError:
 
337
                # all the paths in this block are not versioned in this tree
 
338
                continue
 
339
            for key, entry in block[1]:
 
340
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
341
                if minikind in ('a', 'r'): # absent, relocated
 
342
                    # a parent tree only entry
 
343
                    continue
 
344
                name = key[1]
 
345
                name_unicode = utf8_decode(name)[0]
 
346
                file_id = key[2]
 
347
                kind = minikind_to_kind[minikind]
 
348
                inv_entry = factory[kind](file_id, name_unicode,
 
349
                                          parent_ie.file_id)
 
350
                if kind == 'file':
 
351
                    # This is only needed on win32, where this is the only way
 
352
                    # we know the executable bit.
 
353
                    inv_entry.executable = executable
 
354
                    # not strictly needed: working tree
 
355
                    #inv_entry.text_size = size
 
356
                    #inv_entry.text_sha1 = sha1
 
357
                elif kind == 'directory':
 
358
                    # add this entry to the parent map.
 
359
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
360
                elif kind == 'tree-reference':
 
361
                    assert self._repo_supports_tree_reference, \
 
362
                        "repository of %r " \
 
363
                        "doesn't support tree references " \
 
364
                        "required by entry %r" \
 
365
                        % (self, name)
 
366
                    inv_entry.reference_revision = link_or_sha1 or None
 
367
                elif kind != 'symlink':
 
368
                    raise AssertionError("unknown kind %r" % kind)
 
369
                # These checks cost us around 40ms on a 55k entry tree
 
370
                assert file_id not in inv_byid, ('file_id %s already in'
 
371
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
372
                assert name_unicode not in parent_ie.children
 
373
                inv_byid[file_id] = inv_entry
 
374
                parent_ie.children[name_unicode] = inv_entry
 
375
        self._inventory = inv
 
376
 
 
377
    def _get_entry(self, file_id=None, path=None):
 
378
        """Get the dirstate row for file_id or path.
 
379
 
 
380
        If either file_id or path is supplied, it is used as the key to lookup.
 
381
        If both are supplied, the fastest lookup is used, and an error is
 
382
        raised if they do not both point at the same row.
 
383
        
 
384
        :param file_id: An optional unicode file_id to be looked up.
 
385
        :param path: An optional unicode path to be looked up.
 
386
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
387
        """
 
388
        if file_id is None and path is None:
 
389
            raise errors.BzrError('must supply file_id or path')
 
390
        state = self.current_dirstate()
 
391
        if path is not None:
 
392
            path = path.encode('utf8')
 
393
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
394
 
 
395
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
396
        # check file id is valid unconditionally.
 
397
        entry = self._get_entry(file_id=file_id, path=path)
 
398
        if entry[0] is None:
 
399
            raise errors.NoSuchId(self, file_id)
 
400
        if path is None:
 
401
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
402
 
 
403
        file_abspath = self.abspath(path)
 
404
        state = self.current_dirstate()
 
405
        if stat_value is None:
 
406
            try:
 
407
                stat_value = os.lstat(file_abspath)
 
408
            except OSError, e:
 
409
                if e.errno == errno.ENOENT:
 
410
                    return None
 
411
                else:
 
412
                    raise
 
413
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
414
                                          stat_value=stat_value)
 
415
        if entry[1][0][0] == 'f':
 
416
            return link_or_sha1
 
417
        return None
 
418
 
 
419
    def _get_inventory(self):
 
420
        """Get the inventory for the tree. This is only valid within a lock."""
 
421
        if 'evil' in debug.debug_flags:
 
422
            trace.mutter_callsite(2,
 
423
                "accessing .inventory forces a size of tree translation.")
 
424
        if self._inventory is not None:
 
425
            return self._inventory
 
426
        self._must_be_locked()
 
427
        self._generate_inventory()
 
428
        return self._inventory
 
429
 
 
430
    inventory = property(_get_inventory,
 
431
                         doc="Inventory of this Tree")
 
432
 
 
433
    @needs_read_lock
 
434
    def get_parent_ids(self):
 
435
        """See Tree.get_parent_ids.
 
436
        
 
437
        This implementation requests the ids list from the dirstate file.
 
438
        """
 
439
        return self.current_dirstate().get_parent_ids()
 
440
 
 
441
    def get_reference_revision(self, file_id, path=None):
 
442
        # referenced tree's revision is whatever's currently there
 
443
        return self.get_nested_tree(file_id, path).last_revision()
 
444
 
 
445
    def get_nested_tree(self, file_id, path=None):
 
446
        if path is None:
 
447
            path = self.id2path(file_id)
 
448
        # else: check file_id is at path?
 
449
        return WorkingTree.open(self.abspath(path))
 
450
 
 
451
    @needs_read_lock
 
452
    def get_root_id(self):
 
453
        """Return the id of this trees root"""
 
454
        return self._get_entry(path='')[0][2]
 
455
 
 
456
    def has_id(self, file_id):
 
457
        state = self.current_dirstate()
 
458
        file_id = osutils.safe_file_id(file_id)
 
459
        row, parents = self._get_entry(file_id=file_id)
 
460
        if row is None:
 
461
            return False
 
462
        return osutils.lexists(pathjoin(
 
463
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
464
 
 
465
    @needs_read_lock
 
466
    def id2path(self, file_id):
 
467
        "Convert a file-id to a path."
 
468
        file_id = osutils.safe_file_id(file_id)
 
469
        state = self.current_dirstate()
 
470
        entry = self._get_entry(file_id=file_id)
 
471
        if entry == (None, None):
 
472
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
473
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
474
        return path_utf8.decode('utf8')
 
475
 
 
476
    if not osutils.supports_executable():
 
477
        def is_executable(self, file_id, path=None):
 
478
            """Test if a file is executable or not.
 
479
 
 
480
            Note: The caller is expected to take a read-lock before calling this.
 
481
            """
 
482
            file_id = osutils.safe_file_id(file_id)
 
483
            entry = self._get_entry(file_id=file_id, path=path)
 
484
            if entry == (None, None):
 
485
                return False
 
486
            return entry[1][0][3]
 
487
    else:
 
488
        def is_executable(self, file_id, path=None):
 
489
            """Test if a file is executable or not.
 
490
 
 
491
            Note: The caller is expected to take a read-lock before calling this.
 
492
            """
 
493
            if not path:
 
494
                file_id = osutils.safe_file_id(file_id)
 
495
                path = self.id2path(file_id)
 
496
            mode = os.lstat(self.abspath(path)).st_mode
 
497
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
498
 
 
499
    @needs_read_lock
 
500
    def __iter__(self):
 
501
        """Iterate through file_ids for this tree.
 
502
 
 
503
        file_ids are in a WorkingTree if they are in the working inventory
 
504
        and the working file exists.
 
505
        """
 
506
        result = []
 
507
        for key, tree_details in self.current_dirstate()._iter_entries():
 
508
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
509
                # not relevant to the working tree
 
510
                continue
 
511
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
512
            if osutils.lexists(path):
 
513
                result.append(key[2])
 
514
        return iter(result)
 
515
 
 
516
    def iter_references(self):
 
517
        for key, tree_details in self.current_dirstate()._iter_entries():
 
518
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
519
                # not relevant to the working tree
 
520
                continue
 
521
            if not key[1]:
 
522
                # the root is not a reference.
 
523
                continue
 
524
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
525
            try:
 
526
                if self._kind(path) == 'tree-reference':
 
527
                    yield path, key[2]
 
528
            except errors.NoSuchFile:
 
529
                # path is missing on disk.
 
530
                continue
 
531
 
 
532
    def kind(self, file_id):
 
533
        """Return the kind of a file.
 
534
 
 
535
        This is always the actual kind that's on disk, regardless of what it
 
536
        was added as.
 
537
 
 
538
        Note: The caller is expected to take a read-lock before calling this.
 
539
        """
 
540
        relpath = self.id2path(file_id)
 
541
        assert relpath != None, \
 
542
            "path for id {%s} is None!" % file_id
 
543
        return self._kind(relpath)
 
544
 
 
545
    def _kind(self, relpath):
 
546
        abspath = self.abspath(relpath)
 
547
        kind = file_kind(abspath)
 
548
        if (self._repo_supports_tree_reference and
 
549
            kind == 'directory' and
 
550
            self._directory_is_tree_reference(relpath)):
 
551
            kind = 'tree-reference'
 
552
        return kind
 
553
 
 
554
    @needs_read_lock
 
555
    def _last_revision(self):
 
556
        """See Mutable.last_revision."""
 
557
        parent_ids = self.current_dirstate().get_parent_ids()
 
558
        if parent_ids:
 
559
            return parent_ids[0]
 
560
        else:
 
561
            return _mod_revision.NULL_REVISION
 
562
 
 
563
    def lock_read(self):
 
564
        """See Branch.lock_read, and WorkingTree.unlock."""
 
565
        self.branch.lock_read()
 
566
        try:
 
567
            self._control_files.lock_read()
 
568
            try:
 
569
                state = self.current_dirstate()
 
570
                if not state._lock_token:
 
571
                    state.lock_read()
 
572
                # set our support for tree references from the repository in
 
573
                # use.
 
574
                self._repo_supports_tree_reference = getattr(
 
575
                    self.branch.repository._format, "supports_tree_reference",
 
576
                    False)
 
577
            except:
 
578
                self._control_files.unlock()
 
579
                raise
 
580
        except:
 
581
            self.branch.unlock()
 
582
            raise
 
583
 
 
584
    def _lock_self_write(self):
 
585
        """This should be called after the branch is locked."""
 
586
        try:
 
587
            self._control_files.lock_write()
 
588
            try:
 
589
                state = self.current_dirstate()
 
590
                if not state._lock_token:
 
591
                    state.lock_write()
 
592
                # set our support for tree references from the repository in
 
593
                # use.
 
594
                self._repo_supports_tree_reference = getattr(
 
595
                    self.branch.repository._format, "supports_tree_reference",
 
596
                    False)
 
597
            except:
 
598
                self._control_files.unlock()
 
599
                raise
 
600
        except:
 
601
            self.branch.unlock()
 
602
            raise
 
603
 
 
604
    def lock_tree_write(self):
 
605
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
606
        self.branch.lock_read()
 
607
        self._lock_self_write()
 
608
 
 
609
    def lock_write(self):
 
610
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
611
        self.branch.lock_write()
 
612
        self._lock_self_write()
 
613
 
 
614
    @needs_tree_write_lock
 
615
    def move(self, from_paths, to_dir, after=False):
 
616
        """See WorkingTree.move()."""
 
617
        result = []
 
618
        if not from_paths:
 
619
            return result
 
620
 
 
621
        state = self.current_dirstate()
 
622
 
 
623
        assert not isinstance(from_paths, basestring)
 
624
        to_dir_utf8 = to_dir.encode('utf8')
 
625
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
626
        id_index = state._get_id_index()
 
627
        # check destination directory
 
628
        # get the details for it
 
629
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
630
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
631
        if not entry_present:
 
632
            raise errors.BzrMoveFailedError('', to_dir,
 
633
                errors.NotVersionedError(to_dir))
 
634
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
635
        # get a handle on the block itself.
 
636
        to_block_index = state._ensure_block(
 
637
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
638
        to_block = state._dirblocks[to_block_index]
 
639
        to_abs = self.abspath(to_dir)
 
640
        if not isdir(to_abs):
 
641
            raise errors.BzrMoveFailedError('',to_dir,
 
642
                errors.NotADirectory(to_abs))
 
643
 
 
644
        if to_entry[1][0][0] != 'd':
 
645
            raise errors.BzrMoveFailedError('',to_dir,
 
646
                errors.NotADirectory(to_abs))
 
647
 
 
648
        if self._inventory is not None:
 
649
            update_inventory = True
 
650
            inv = self.inventory
 
651
            to_dir_ie = inv[to_dir_id]
 
652
            to_dir_id = to_entry[0][2]
 
653
        else:
 
654
            update_inventory = False
 
655
 
 
656
        rollbacks = []
 
657
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
658
                     fingerprint, packed_stat, size,
 
659
                     to_block, to_key, to_path_utf8):
 
660
            state._make_absent(old_entry)
 
661
            from_key = old_entry[0]
 
662
            rollbacks.append(
 
663
                lambda:state.update_minimal(from_key,
 
664
                    minikind,
 
665
                    executable=executable,
 
666
                    fingerprint=fingerprint,
 
667
                    packed_stat=packed_stat,
 
668
                    size=size,
 
669
                    path_utf8=from_path_utf8))
 
670
            state.update_minimal(to_key,
 
671
                    minikind,
 
672
                    executable=executable,
 
673
                    fingerprint=fingerprint,
 
674
                    packed_stat=packed_stat,
 
675
                    size=size,
 
676
                    path_utf8=to_path_utf8)
 
677
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
678
            new_entry = to_block[1][added_entry_index]
 
679
            rollbacks.append(lambda:state._make_absent(new_entry))
 
680
 
 
681
        for from_rel in from_paths:
 
682
            # from_rel is 'pathinroot/foo/bar'
 
683
            from_rel_utf8 = from_rel.encode('utf8')
 
684
            from_dirname, from_tail = osutils.split(from_rel)
 
685
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
686
            from_entry = self._get_entry(path=from_rel)
 
687
            if from_entry == (None, None):
 
688
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
689
                    errors.NotVersionedError(path=str(from_rel)))
 
690
 
 
691
            from_id = from_entry[0][2]
 
692
            to_rel = pathjoin(to_dir, from_tail)
 
693
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
694
            item_to_entry = self._get_entry(path=to_rel)
 
695
            if item_to_entry != (None, None):
 
696
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
697
                    "Target is already versioned.")
 
698
 
 
699
            if from_rel == to_rel:
 
700
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
701
                    "Source and target are identical.")
 
702
 
 
703
            from_missing = not self.has_filename(from_rel)
 
704
            to_missing = not self.has_filename(to_rel)
 
705
            if after:
 
706
                move_file = False
 
707
            else:
 
708
                move_file = True
 
709
            if to_missing:
 
710
                if not move_file:
 
711
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
712
                        errors.NoSuchFile(path=to_rel,
 
713
                        extra="New file has not been created yet"))
 
714
                elif from_missing:
 
715
                    # neither path exists
 
716
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
717
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
718
            else:
 
719
                if from_missing: # implicitly just update our path mapping
 
720
                    move_file = False
 
721
                elif not after:
 
722
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
723
                        extra="(Use --after to update the Bazaar id)")
 
724
 
 
725
            rollbacks = []
 
726
            def rollback_rename():
 
727
                """A single rename has failed, roll it back."""
 
728
                # roll back everything, even if we encounter trouble doing one
 
729
                # of them.
 
730
                #
 
731
                # TODO: at least log the other exceptions rather than just
 
732
                # losing them mbp 20070307
 
733
                exc_info = None
 
734
                for rollback in reversed(rollbacks):
 
735
                    try:
 
736
                        rollback()
 
737
                    except Exception, e:
 
738
                        exc_info = sys.exc_info()
 
739
                if exc_info:
 
740
                    raise exc_info[0], exc_info[1], exc_info[2]
 
741
 
 
742
            # perform the disk move first - its the most likely failure point.
 
743
            if move_file:
 
744
                from_rel_abs = self.abspath(from_rel)
 
745
                to_rel_abs = self.abspath(to_rel)
 
746
                try:
 
747
                    osutils.rename(from_rel_abs, to_rel_abs)
 
748
                except OSError, e:
 
749
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
750
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
751
            try:
 
752
                # perform the rename in the inventory next if needed: its easy
 
753
                # to rollback
 
754
                if update_inventory:
 
755
                    # rename the entry
 
756
                    from_entry = inv[from_id]
 
757
                    current_parent = from_entry.parent_id
 
758
                    inv.rename(from_id, to_dir_id, from_tail)
 
759
                    rollbacks.append(
 
760
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
761
                # finally do the rename in the dirstate, which is a little
 
762
                # tricky to rollback, but least likely to need it.
 
763
                old_block_index, old_entry_index, dir_present, file_present = \
 
764
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
765
                old_block = state._dirblocks[old_block_index][1]
 
766
                old_entry = old_block[old_entry_index]
 
767
                from_key, old_entry_details = old_entry
 
768
                cur_details = old_entry_details[0]
 
769
                # remove the old row
 
770
                to_key = ((to_block[0],) + from_key[1:3])
 
771
                minikind = cur_details[0]
 
772
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
773
                         minikind=minikind,
 
774
                         executable=cur_details[3],
 
775
                         fingerprint=cur_details[1],
 
776
                         packed_stat=cur_details[4],
 
777
                         size=cur_details[2],
 
778
                         to_block=to_block,
 
779
                         to_key=to_key,
 
780
                         to_path_utf8=to_rel_utf8)
 
781
 
 
782
                if minikind == 'd':
 
783
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
784
                        """Recursively update all entries in this dirblock."""
 
785
                        assert from_dir != '', "renaming root not supported"
 
786
                        from_key = (from_dir, '')
 
787
                        from_block_idx, present = \
 
788
                            state._find_block_index_from_key(from_key)
 
789
                        if not present:
 
790
                            # This is the old record, if it isn't present, then
 
791
                            # there is theoretically nothing to update.
 
792
                            # (Unless it isn't present because of lazy loading,
 
793
                            # but we don't do that yet)
 
794
                            return
 
795
                        from_block = state._dirblocks[from_block_idx]
 
796
                        to_block_index, to_entry_index, _, _ = \
 
797
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
798
                        to_block_index = state._ensure_block(
 
799
                            to_block_index, to_entry_index, to_dir_utf8)
 
800
                        to_block = state._dirblocks[to_block_index]
 
801
 
 
802
                        # Grab a copy since move_one may update the list.
 
803
                        for entry in from_block[1][:]:
 
804
                            assert entry[0][0] == from_dir
 
805
                            cur_details = entry[1][0]
 
806
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
807
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
808
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
809
                            minikind = cur_details[0]
 
810
                            if minikind in 'ar':
 
811
                                # Deleted children of a renamed directory
 
812
                                # Do not need to be updated.
 
813
                                # Children that have been renamed out of this
 
814
                                # directory should also not be updated
 
815
                                continue
 
816
                            move_one(entry, from_path_utf8=from_path_utf8,
 
817
                                     minikind=minikind,
 
818
                                     executable=cur_details[3],
 
819
                                     fingerprint=cur_details[1],
 
820
                                     packed_stat=cur_details[4],
 
821
                                     size=cur_details[2],
 
822
                                     to_block=to_block,
 
823
                                     to_key=to_key,
 
824
                                     to_path_utf8=to_path_utf8)
 
825
                            if minikind == 'd':
 
826
                                # We need to move all the children of this
 
827
                                # entry
 
828
                                update_dirblock(from_path_utf8, to_key,
 
829
                                                to_path_utf8)
 
830
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
831
            except:
 
832
                rollback_rename()
 
833
                raise
 
834
            result.append((from_rel, to_rel))
 
835
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
836
            self._make_dirty(reset_inventory=False)
 
837
 
 
838
        return result
 
839
 
 
840
    def _must_be_locked(self):
 
841
        if not self._control_files._lock_count:
 
842
            raise errors.ObjectNotLocked(self)
 
843
 
 
844
    def _new_tree(self):
 
845
        """Initialize the state in this tree to be a new tree."""
 
846
        self._dirty = True
 
847
 
 
848
    @needs_read_lock
 
849
    def path2id(self, path):
 
850
        """Return the id for path in this tree."""
 
851
        path = path.strip('/')
 
852
        entry = self._get_entry(path=path)
 
853
        if entry == (None, None):
 
854
            return None
 
855
        return entry[0][2]
 
856
 
 
857
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
858
        """See Tree.paths2ids().
 
859
 
 
860
        This specialisation fast-paths the case where all the trees are in the
 
861
        dirstate.
 
862
        """
 
863
        if paths is None:
 
864
            return None
 
865
        parents = self.get_parent_ids()
 
866
        for tree in trees:
 
867
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
868
                parents):
 
869
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
870
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
871
        # -- make all paths utf8 --
 
872
        paths_utf8 = set()
 
873
        for path in paths:
 
874
            paths_utf8.add(path.encode('utf8'))
 
875
        paths = paths_utf8
 
876
        # -- paths is now a utf8 path set --
 
877
        # -- get the state object and prepare it.
 
878
        state = self.current_dirstate()
 
879
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
880
            and '' not in paths):
 
881
            paths2ids = self._paths2ids_using_bisect
 
882
        else:
 
883
            paths2ids = self._paths2ids_in_memory
 
884
        return paths2ids(paths, search_indexes,
 
885
                         require_versioned=require_versioned)
 
886
 
 
887
    def _paths2ids_in_memory(self, paths, search_indexes,
 
888
                             require_versioned=True):
 
889
        state = self.current_dirstate()
 
890
        state._read_dirblocks_if_needed()
 
891
        def _entries_for_path(path):
 
892
            """Return a list with all the entries that match path for all ids.
 
893
            """
 
894
            dirname, basename = os.path.split(path)
 
895
            key = (dirname, basename, '')
 
896
            block_index, present = state._find_block_index_from_key(key)
 
897
            if not present:
 
898
                # the block which should contain path is absent.
 
899
                return []
 
900
            result = []
 
901
            block = state._dirblocks[block_index][1]
 
902
            entry_index, _ = state._find_entry_index(key, block)
 
903
            # we may need to look at multiple entries at this path: walk while the paths match.
 
904
            while (entry_index < len(block) and
 
905
                block[entry_index][0][0:2] == key[0:2]):
 
906
                result.append(block[entry_index])
 
907
                entry_index += 1
 
908
            return result
 
909
        if require_versioned:
 
910
            # -- check all supplied paths are versioned in a search tree. --
 
911
            all_versioned = True
 
912
            for path in paths:
 
913
                path_entries = _entries_for_path(path)
 
914
                if not path_entries:
 
915
                    # this specified path is not present at all: error
 
916
                    all_versioned = False
 
917
                    break
 
918
                found_versioned = False
 
919
                # for each id at this path
 
920
                for entry in path_entries:
 
921
                    # for each tree.
 
922
                    for index in search_indexes:
 
923
                        if entry[1][index][0] != 'a': # absent
 
924
                            found_versioned = True
 
925
                            # all good: found a versioned cell
 
926
                            break
 
927
                if not found_versioned:
 
928
                    # none of the indexes was not 'absent' at all ids for this
 
929
                    # path.
 
930
                    all_versioned = False
 
931
                    break
 
932
            if not all_versioned:
 
933
                raise errors.PathsNotVersionedError(paths)
 
934
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
935
        search_paths = set()
 
936
        for path in paths:
 
937
            other_paths = paths.difference(set([path]))
 
938
            if not osutils.is_inside_any(other_paths, path):
 
939
                # this is a top level path, we must check it.
 
940
                search_paths.add(path)
 
941
        # sketch: 
 
942
        # for all search_indexs in each path at or under each element of
 
943
        # search_paths, if the detail is relocated: add the id, and add the
 
944
        # relocated path as one to search if its not searched already. If the
 
945
        # detail is not relocated, add the id.
 
946
        searched_paths = set()
 
947
        found_ids = set()
 
948
        def _process_entry(entry):
 
949
            """Look at search_indexes within entry.
 
950
 
 
951
            If a specific tree's details are relocated, add the relocation
 
952
            target to search_paths if not searched already. If it is absent, do
 
953
            nothing. Otherwise add the id to found_ids.
 
954
            """
 
955
            for index in search_indexes:
 
956
                if entry[1][index][0] == 'r': # relocated
 
957
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
958
                        search_paths.add(entry[1][index][1])
 
959
                elif entry[1][index][0] != 'a': # absent
 
960
                    found_ids.add(entry[0][2])
 
961
        while search_paths:
 
962
            current_root = search_paths.pop()
 
963
            searched_paths.add(current_root)
 
964
            # process the entries for this containing directory: the rest will be
 
965
            # found by their parents recursively.
 
966
            root_entries = _entries_for_path(current_root)
 
967
            if not root_entries:
 
968
                # this specified path is not present at all, skip it.
 
969
                continue
 
970
            for entry in root_entries:
 
971
                _process_entry(entry)
 
972
            initial_key = (current_root, '', '')
 
973
            block_index, _ = state._find_block_index_from_key(initial_key)
 
974
            while (block_index < len(state._dirblocks) and
 
975
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
976
                for entry in state._dirblocks[block_index][1]:
 
977
                    _process_entry(entry)
 
978
                block_index += 1
 
979
        return found_ids
 
980
 
 
981
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
982
                                require_versioned=True):
 
983
        state = self.current_dirstate()
 
984
        found_ids = set()
 
985
 
 
986
        split_paths = sorted(osutils.split(p) for p in paths)
 
987
        found = state._bisect_recursive(split_paths)
 
988
 
 
989
        if require_versioned:
 
990
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
991
            for dir_name in split_paths:
 
992
                if dir_name not in found_dir_names:
 
993
                    raise errors.PathsNotVersionedError(paths)
 
994
 
 
995
        for dir_name_id, trees_info in found.iteritems():
 
996
            for index in search_indexes:
 
997
                if trees_info[index][0] not in ('r', 'a'):
 
998
                    found_ids.add(dir_name_id[2])
 
999
        return found_ids
 
1000
 
 
1001
    def read_working_inventory(self):
 
1002
        """Read the working inventory.
 
1003
        
 
1004
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1005
        """
 
1006
        return self.inventory
 
1007
 
 
1008
    @needs_read_lock
 
1009
    def revision_tree(self, revision_id):
 
1010
        """See Tree.revision_tree.
 
1011
 
 
1012
        WorkingTree4 supplies revision_trees for any basis tree.
 
1013
        """
 
1014
        revision_id = osutils.safe_revision_id(revision_id)
 
1015
        dirstate = self.current_dirstate()
 
1016
        parent_ids = dirstate.get_parent_ids()
 
1017
        if revision_id not in parent_ids:
 
1018
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1019
        if revision_id in dirstate.get_ghosts():
 
1020
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1021
        return DirStateRevisionTree(dirstate, revision_id,
 
1022
            self.branch.repository)
 
1023
 
 
1024
    @needs_tree_write_lock
 
1025
    def set_last_revision(self, new_revision):
 
1026
        """Change the last revision in the working tree."""
 
1027
        new_revision = osutils.safe_revision_id(new_revision)
 
1028
        parents = self.get_parent_ids()
 
1029
        if new_revision in (NULL_REVISION, None):
 
1030
            assert len(parents) < 2, (
 
1031
                "setting the last parent to none with a pending merge is "
 
1032
                "unsupported.")
 
1033
            self.set_parent_ids([])
 
1034
        else:
 
1035
            self.set_parent_ids([new_revision] + parents[1:],
 
1036
                allow_leftmost_as_ghost=True)
 
1037
 
 
1038
    @needs_tree_write_lock
 
1039
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1040
        """Set the parent ids to revision_ids.
 
1041
        
 
1042
        See also set_parent_trees. This api will try to retrieve the tree data
 
1043
        for each element of revision_ids from the trees repository. If you have
 
1044
        tree data already available, it is more efficient to use
 
1045
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1046
        an easier API to use.
 
1047
 
 
1048
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1049
            working tree. Any of these may be ghosts.
 
1050
        """
 
1051
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1052
        trees = []
 
1053
        for revision_id in revision_ids:
 
1054
            try:
 
1055
                revtree = self.branch.repository.revision_tree(revision_id)
 
1056
                # TODO: jam 20070213 KnitVersionedFile raises
 
1057
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1058
                #       given revision_id is not present. Should Repository be
 
1059
                #       catching it and re-raising NoSuchRevision?
 
1060
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1061
                revtree = None
 
1062
            trees.append((revision_id, revtree))
 
1063
        self.set_parent_trees(trees,
 
1064
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1065
 
 
1066
    @needs_tree_write_lock
 
1067
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1068
        """Set the parents of the working tree.
 
1069
 
 
1070
        :param parents_list: A list of (revision_id, tree) tuples.
 
1071
            If tree is None, then that element is treated as an unreachable
 
1072
            parent tree - i.e. a ghost.
 
1073
        """
 
1074
        dirstate = self.current_dirstate()
 
1075
        if len(parents_list) > 0:
 
1076
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1077
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1078
        real_trees = []
 
1079
        ghosts = []
 
1080
        # convert absent trees to the null tree, which we convert back to
 
1081
        # missing on access.
 
1082
        for rev_id, tree in parents_list:
 
1083
            rev_id = osutils.safe_revision_id(rev_id)
 
1084
            _mod_revision.check_not_reserved_id(rev_id)
 
1085
            if tree is not None:
 
1086
                real_trees.append((rev_id, tree))
 
1087
            else:
 
1088
                real_trees.append((rev_id,
 
1089
                    self.branch.repository.revision_tree(None)))
 
1090
                ghosts.append(rev_id)
 
1091
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1092
        self._make_dirty(reset_inventory=False)
 
1093
 
 
1094
    def _set_root_id(self, file_id):
 
1095
        """See WorkingTree.set_root_id."""
 
1096
        state = self.current_dirstate()
 
1097
        state.set_path_id('', file_id)
 
1098
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1099
            self._make_dirty(reset_inventory=True)
 
1100
 
 
1101
    @needs_read_lock
 
1102
    def supports_tree_reference(self):
 
1103
        return self._repo_supports_tree_reference
 
1104
 
 
1105
    def unlock(self):
 
1106
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1107
        # do non-implementation specific cleanup
 
1108
        self._cleanup()
 
1109
 
 
1110
        if self._control_files._lock_count == 1:
 
1111
            # eventually we should do signature checking during read locks for
 
1112
            # dirstate updates.
 
1113
            if self._control_files._lock_mode == 'w':
 
1114
                if self._dirty:
 
1115
                    self.flush()
 
1116
            if self._dirstate is not None:
 
1117
                # This is a no-op if there are no modifications.
 
1118
                self._dirstate.save()
 
1119
                self._dirstate.unlock()
 
1120
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1121
            #       point. Instead, it could check if the header has been
 
1122
            #       modified when it is locked, and if not, it can hang on to
 
1123
            #       the data it has in memory.
 
1124
            self._dirstate = None
 
1125
            self._inventory = None
 
1126
        # reverse order of locking.
 
1127
        try:
 
1128
            return self._control_files.unlock()
 
1129
        finally:
 
1130
            self.branch.unlock()
 
1131
 
 
1132
    @needs_tree_write_lock
 
1133
    def unversion(self, file_ids):
 
1134
        """Remove the file ids in file_ids from the current versioned set.
 
1135
 
 
1136
        When a file_id is unversioned, all of its children are automatically
 
1137
        unversioned.
 
1138
 
 
1139
        :param file_ids: The file ids to stop versioning.
 
1140
        :raises: NoSuchId if any fileid is not currently versioned.
 
1141
        """
 
1142
        if not file_ids:
 
1143
            return
 
1144
        state = self.current_dirstate()
 
1145
        state._read_dirblocks_if_needed()
 
1146
        ids_to_unversion = set()
 
1147
        for file_id in file_ids:
 
1148
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1149
        paths_to_unversion = set()
 
1150
        # sketch:
 
1151
        # check if the root is to be unversioned, if so, assert for now.
 
1152
        # walk the state marking unversioned things as absent.
 
1153
        # if there are any un-unversioned ids at the end, raise
 
1154
        for key, details in state._dirblocks[0][1]:
 
1155
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1156
                key[2] in ids_to_unversion):
 
1157
                # I haven't written the code to unversion / yet - it should be
 
1158
                # supported.
 
1159
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1160
        block_index = 0
 
1161
        while block_index < len(state._dirblocks):
 
1162
            # process one directory at a time.
 
1163
            block = state._dirblocks[block_index]
 
1164
            # first check: is the path one to remove - it or its children
 
1165
            delete_block = False
 
1166
            for path in paths_to_unversion:
 
1167
                if (block[0].startswith(path) and
 
1168
                    (len(block[0]) == len(path) or
 
1169
                     block[0][len(path)] == '/')):
 
1170
                    # this entire block should be deleted - its the block for a
 
1171
                    # path to unversion; or the child of one
 
1172
                    delete_block = True
 
1173
                    break
 
1174
            # TODO: trim paths_to_unversion as we pass by paths
 
1175
            if delete_block:
 
1176
                # this block is to be deleted: process it.
 
1177
                # TODO: we can special case the no-parents case and
 
1178
                # just forget the whole block.
 
1179
                entry_index = 0
 
1180
                while entry_index < len(block[1]):
 
1181
                    # Mark this file id as having been removed
 
1182
                    entry = block[1][entry_index]
 
1183
                    ids_to_unversion.discard(entry[0][2])
 
1184
                    if (entry[1][0][0] == 'a'
 
1185
                        or not state._make_absent(entry)):
 
1186
                        entry_index += 1
 
1187
                # go to the next block. (At the moment we dont delete empty
 
1188
                # dirblocks)
 
1189
                block_index += 1
 
1190
                continue
 
1191
            entry_index = 0
 
1192
            while entry_index < len(block[1]):
 
1193
                entry = block[1][entry_index]
 
1194
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1195
                    # ^ some parent row.
 
1196
                    entry[0][2] not in ids_to_unversion):
 
1197
                    # ^ not an id to unversion
 
1198
                    entry_index += 1
 
1199
                    continue
 
1200
                if entry[1][0][0] == 'd':
 
1201
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1202
                if not state._make_absent(entry):
 
1203
                    entry_index += 1
 
1204
                # we have unversioned this id
 
1205
                ids_to_unversion.remove(entry[0][2])
 
1206
            block_index += 1
 
1207
        if ids_to_unversion:
 
1208
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1209
        self._make_dirty(reset_inventory=False)
 
1210
        # have to change the legacy inventory too.
 
1211
        if self._inventory is not None:
 
1212
            for file_id in file_ids:
 
1213
                self._inventory.remove_recursive_id(file_id)
 
1214
 
 
1215
    @needs_read_lock
 
1216
    def _validate(self):
 
1217
        self._dirstate._validate()
 
1218
 
 
1219
    @needs_tree_write_lock
 
1220
    def _write_inventory(self, inv):
 
1221
        """Write inventory as the current inventory."""
 
1222
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1223
        self.current_dirstate().set_state_from_inventory(inv)
 
1224
        self._make_dirty(reset_inventory=False)
 
1225
        if self._inventory is not None:
 
1226
            self._inventory = inv
 
1227
        self.flush()
 
1228
 
 
1229
 
 
1230
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1231
    """The first consolidated dirstate working tree format.
 
1232
 
 
1233
    This format:
 
1234
        - exists within a metadir controlling .bzr
 
1235
        - includes an explicit version marker for the workingtree control
 
1236
          files, separate from the BzrDir format
 
1237
        - modifies the hash cache format
 
1238
        - is new in bzr 0.15
 
1239
        - uses a LockDir to guard access to it.
 
1240
    """
 
1241
 
 
1242
    upgrade_recommended = False
 
1243
 
 
1244
    def get_format_string(self):
 
1245
        """See WorkingTreeFormat.get_format_string()."""
 
1246
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1247
 
 
1248
    def get_format_description(self):
 
1249
        """See WorkingTreeFormat.get_format_description()."""
 
1250
        return "Working tree format 4"
 
1251
 
 
1252
    def initialize(self, a_bzrdir, revision_id=None):
 
1253
        """See WorkingTreeFormat.initialize().
 
1254
 
 
1255
        :param revision_id: allows creating a working tree at a different
 
1256
        revision than the branch is at.
 
1257
 
 
1258
        These trees get an initial random root id, if their repository supports
 
1259
        rich root data, TREE_ROOT otherwise.
 
1260
        """
 
1261
        revision_id = osutils.safe_revision_id(revision_id)
 
1262
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1263
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1264
        transport = a_bzrdir.get_workingtree_transport(self)
 
1265
        control_files = self._open_control_files(a_bzrdir)
 
1266
        control_files.create_lock()
 
1267
        control_files.lock_write()
 
1268
        control_files.put_utf8('format', self.get_format_string())
 
1269
        branch = a_bzrdir.open_branch()
 
1270
        if revision_id is None:
 
1271
            revision_id = branch.last_revision()
 
1272
        local_path = transport.local_abspath('dirstate')
 
1273
        # write out new dirstate (must exist when we create the tree)
 
1274
        state = dirstate.DirState.initialize(local_path)
 
1275
        state.unlock()
 
1276
        del state
 
1277
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1278
                         branch,
 
1279
                         _format=self,
 
1280
                         _bzrdir=a_bzrdir,
 
1281
                         _control_files=control_files)
 
1282
        wt._new_tree()
 
1283
        wt.lock_tree_write()
 
1284
        try:
 
1285
            if revision_id in (None, NULL_REVISION):
 
1286
                if branch.repository.supports_rich_root():
 
1287
                    wt._set_root_id(generate_ids.gen_root_id())
 
1288
                else:
 
1289
                    wt._set_root_id(ROOT_ID)
 
1290
                wt.flush()
 
1291
            wt.set_last_revision(revision_id)
 
1292
            wt.flush()
 
1293
            basis = wt.basis_tree()
 
1294
            basis.lock_read()
 
1295
            # if the basis has a root id we have to use that; otherwise we use
 
1296
            # a new random one
 
1297
            basis_root_id = basis.get_root_id()
 
1298
            if basis_root_id is not None:
 
1299
                wt._set_root_id(basis_root_id)
 
1300
                wt.flush()
 
1301
            transform.build_tree(basis, wt)
 
1302
            basis.unlock()
 
1303
        finally:
 
1304
            control_files.unlock()
 
1305
            wt.unlock()
 
1306
        return wt
 
1307
 
 
1308
    def _open(self, a_bzrdir, control_files):
 
1309
        """Open the tree itself.
 
1310
 
 
1311
        :param a_bzrdir: the dir for the tree.
 
1312
        :param control_files: the control files for the tree.
 
1313
        """
 
1314
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1315
                           branch=a_bzrdir.open_branch(),
 
1316
                           _format=self,
 
1317
                           _bzrdir=a_bzrdir,
 
1318
                           _control_files=control_files)
 
1319
 
 
1320
    def __get_matchingbzrdir(self):
 
1321
        # please test against something that will let us do tree references
 
1322
        return bzrdir.format_registry.make_bzrdir(
 
1323
            'dirstate-with-subtree')
 
1324
 
 
1325
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1326
 
 
1327
 
 
1328
class DirStateRevisionTree(Tree):
 
1329
    """A revision tree pulling the inventory from a dirstate."""
 
1330
 
 
1331
    def __init__(self, dirstate, revision_id, repository):
 
1332
        self._dirstate = dirstate
 
1333
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1334
        self._repository = repository
 
1335
        self._inventory = None
 
1336
        self._locked = 0
 
1337
        self._dirstate_locked = False
 
1338
 
 
1339
    def __repr__(self):
 
1340
        return "<%s of %s in %s>" % \
 
1341
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1342
 
 
1343
    def annotate_iter(self, file_id,
 
1344
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1345
        """See Tree.annotate_iter"""
 
1346
        w = self._get_weave(file_id)
 
1347
        return w.annotate_iter(self.inventory[file_id].revision)
 
1348
 
 
1349
    def _get_ancestors(self, default_revision):
 
1350
        return set(self._repository.get_ancestry(self._revision_id,
 
1351
                                                 topo_sorted=False))
 
1352
    def _comparison_data(self, entry, path):
 
1353
        """See Tree._comparison_data."""
 
1354
        if entry is None:
 
1355
            return None, False, None
 
1356
        # trust the entry as RevisionTree does, but this may not be
 
1357
        # sensible: the entry might not have come from us?
 
1358
        return entry.kind, entry.executable, None
 
1359
 
 
1360
    def _file_size(self, entry, stat_value):
 
1361
        return entry.text_size
 
1362
 
 
1363
    def filter_unversioned_files(self, paths):
 
1364
        """Filter out paths that are not versioned.
 
1365
 
 
1366
        :return: set of paths.
 
1367
        """
 
1368
        pred = self.has_filename
 
1369
        return set((p for p in paths if not pred(p)))
 
1370
 
 
1371
    def get_root_id(self):
 
1372
        return self.path2id('')
 
1373
 
 
1374
    def _get_parent_index(self):
 
1375
        """Return the index in the dirstate referenced by this tree."""
 
1376
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1377
 
 
1378
    def _get_entry(self, file_id=None, path=None):
 
1379
        """Get the dirstate row for file_id or path.
 
1380
 
 
1381
        If either file_id or path is supplied, it is used as the key to lookup.
 
1382
        If both are supplied, the fastest lookup is used, and an error is
 
1383
        raised if they do not both point at the same row.
 
1384
        
 
1385
        :param file_id: An optional unicode file_id to be looked up.
 
1386
        :param path: An optional unicode path to be looked up.
 
1387
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1388
        """
 
1389
        if file_id is None and path is None:
 
1390
            raise errors.BzrError('must supply file_id or path')
 
1391
        file_id = osutils.safe_file_id(file_id)
 
1392
        if path is not None:
 
1393
            path = path.encode('utf8')
 
1394
        parent_index = self._get_parent_index()
 
1395
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1396
 
 
1397
    def _generate_inventory(self):
 
1398
        """Create and set self.inventory from the dirstate object.
 
1399
 
 
1400
        (So this is only called the first time the inventory is requested for
 
1401
        this tree; it then remains in memory until it's out of date.)
 
1402
 
 
1403
        This is relatively expensive: we have to walk the entire dirstate.
 
1404
        """
 
1405
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1406
            'dirstate revision tree'
 
1407
        # separate call for profiling - makes it clear where the costs are.
 
1408
        self._dirstate._read_dirblocks_if_needed()
 
1409
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1410
            'parent %s has disappeared from %s' % (
 
1411
            self._revision_id, self._dirstate.get_parent_ids())
 
1412
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1413
        # This is identical now to the WorkingTree _generate_inventory except
 
1414
        # for the tree index use.
 
1415
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1416
        current_id = root_key[2]
 
1417
        assert current_entry[parent_index][0] == 'd'
 
1418
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1419
        inv.root.revision = current_entry[parent_index][4]
 
1420
        # Turn some things into local variables
 
1421
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1422
        factory = entry_factory
 
1423
        utf8_decode = cache_utf8._utf8_decode
 
1424
        inv_byid = inv._byid
 
1425
        # we could do this straight out of the dirstate; it might be fast
 
1426
        # and should be profiled - RBC 20070216
 
1427
        parent_ies = {'' : inv.root}
 
1428
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1429
            dirname = block[0]
 
1430
            try:
 
1431
                parent_ie = parent_ies[dirname]
 
1432
            except KeyError:
 
1433
                # all the paths in this block are not versioned in this tree
 
1434
                continue
 
1435
            for key, entry in block[1]:
 
1436
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1437
                if minikind in ('a', 'r'): # absent, relocated
 
1438
                    # not this tree
 
1439
                    continue
 
1440
                name = key[1]
 
1441
                name_unicode = utf8_decode(name)[0]
 
1442
                file_id = key[2]
 
1443
                kind = minikind_to_kind[minikind]
 
1444
                inv_entry = factory[kind](file_id, name_unicode,
 
1445
                                          parent_ie.file_id)
 
1446
                inv_entry.revision = revid
 
1447
                if kind == 'file':
 
1448
                    inv_entry.executable = executable
 
1449
                    inv_entry.text_size = size
 
1450
                    inv_entry.text_sha1 = fingerprint
 
1451
                elif kind == 'directory':
 
1452
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1453
                elif kind == 'symlink':
 
1454
                    inv_entry.executable = False
 
1455
                    inv_entry.text_size = None
 
1456
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1457
                elif kind == 'tree-reference':
 
1458
                    inv_entry.reference_revision = fingerprint or None
 
1459
                else:
 
1460
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1461
                            % entry)
 
1462
                # These checks cost us around 40ms on a 55k entry tree
 
1463
                assert file_id not in inv_byid
 
1464
                assert name_unicode not in parent_ie.children
 
1465
                inv_byid[file_id] = inv_entry
 
1466
                parent_ie.children[name_unicode] = inv_entry
 
1467
        self._inventory = inv
 
1468
 
 
1469
    def get_file_mtime(self, file_id, path=None):
 
1470
        """Return the modification time for this record.
 
1471
 
 
1472
        We return the timestamp of the last-changed revision.
 
1473
        """
 
1474
        # Make sure the file exists
 
1475
        entry = self._get_entry(file_id, path=path)
 
1476
        if entry == (None, None): # do we raise?
 
1477
            return None
 
1478
        parent_index = self._get_parent_index()
 
1479
        last_changed_revision = entry[1][parent_index][4]
 
1480
        return self._repository.get_revision(last_changed_revision).timestamp
 
1481
 
 
1482
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1483
        entry = self._get_entry(file_id=file_id, path=path)
 
1484
        parent_index = self._get_parent_index()
 
1485
        parent_details = entry[1][parent_index]
 
1486
        if parent_details[0] == 'f':
 
1487
            return parent_details[1]
 
1488
        return None
 
1489
 
 
1490
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1491
    def get_weave(self, file_id):
 
1492
        return self._get_weave(file_id)
 
1493
 
 
1494
    def _get_weave(self, file_id):
 
1495
        return self._repository.weave_store.get_weave(file_id,
 
1496
                self._repository.get_transaction())
 
1497
 
 
1498
    def get_file(self, file_id, path=None):
 
1499
        return StringIO(self.get_file_text(file_id))
 
1500
 
 
1501
    def get_file_lines(self, file_id):
 
1502
        ie = self.inventory[file_id]
 
1503
        return self._get_weave(file_id).get_lines(ie.revision)
 
1504
 
 
1505
    def get_file_size(self, file_id):
 
1506
        return self.inventory[file_id].text_size
 
1507
 
 
1508
    def get_file_text(self, file_id):
 
1509
        return ''.join(self.get_file_lines(file_id))
 
1510
 
 
1511
    def get_reference_revision(self, file_id, path=None):
 
1512
        return self.inventory[file_id].reference_revision
 
1513
 
 
1514
    def iter_files_bytes(self, desired_files):
 
1515
        """See Tree.iter_files_bytes.
 
1516
 
 
1517
        This version is implemented on top of Repository.iter_files_bytes"""
 
1518
        parent_index = self._get_parent_index()
 
1519
        repo_desired_files = []
 
1520
        for file_id, identifier in desired_files:
 
1521
            entry = self._get_entry(file_id)
 
1522
            if entry == (None, None):
 
1523
                raise errors.NoSuchId(self, file_id)
 
1524
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1525
                                       identifier))
 
1526
        return self._repository.iter_files_bytes(repo_desired_files)
 
1527
 
 
1528
    def get_symlink_target(self, file_id):
 
1529
        entry = self._get_entry(file_id=file_id)
 
1530
        parent_index = self._get_parent_index()
 
1531
        if entry[1][parent_index][0] != 'l':
 
1532
            return None
 
1533
        else:
 
1534
            # At present, none of the tree implementations supports non-ascii
 
1535
            # symlink targets. So we will just assume that the dirstate path is
 
1536
            # correct.
 
1537
            return entry[1][parent_index][1]
 
1538
 
 
1539
    def get_revision_id(self):
 
1540
        """Return the revision id for this tree."""
 
1541
        return self._revision_id
 
1542
 
 
1543
    def _get_inventory(self):
 
1544
        if self._inventory is not None:
 
1545
            return self._inventory
 
1546
        self._must_be_locked()
 
1547
        self._generate_inventory()
 
1548
        return self._inventory
 
1549
 
 
1550
    inventory = property(_get_inventory,
 
1551
                         doc="Inventory of this Tree")
 
1552
 
 
1553
    def get_parent_ids(self):
 
1554
        """The parents of a tree in the dirstate are not cached."""
 
1555
        return self._repository.get_revision(self._revision_id).parent_ids
 
1556
 
 
1557
    def has_filename(self, filename):
 
1558
        return bool(self.path2id(filename))
 
1559
 
 
1560
    def kind(self, file_id):
 
1561
        return self.inventory[file_id].kind
 
1562
 
 
1563
    def path_content_summary(self, path):
 
1564
        """See Tree.path_content_summary."""
 
1565
        id = self.inventory.path2id(path)
 
1566
        if id is None:
 
1567
            return ('missing', None, None, None)
 
1568
        entry = self._inventory[id]
 
1569
        kind = entry.kind
 
1570
        if kind == 'file':
 
1571
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1572
        elif kind == 'symlink':
 
1573
            return (kind, None, None, entry.symlink_target)
 
1574
        else:
 
1575
            return (kind, None, None, None)
 
1576
 
 
1577
    def is_executable(self, file_id, path=None):
 
1578
        ie = self.inventory[file_id]
 
1579
        if ie.kind != "file":
 
1580
            return None
 
1581
        return ie.executable
 
1582
 
 
1583
    def list_files(self, include_root=False):
 
1584
        # We use a standard implementation, because DirStateRevisionTree is
 
1585
        # dealing with one of the parents of the current state
 
1586
        inv = self._get_inventory()
 
1587
        entries = inv.iter_entries()
 
1588
        if self.inventory.root is not None and not include_root:
 
1589
            entries.next()
 
1590
        for path, entry in entries:
 
1591
            yield path, 'V', entry.kind, entry.file_id, entry
 
1592
 
 
1593
    def lock_read(self):
 
1594
        """Lock the tree for a set of operations."""
 
1595
        if not self._locked:
 
1596
            self._repository.lock_read()
 
1597
            if self._dirstate._lock_token is None:
 
1598
                self._dirstate.lock_read()
 
1599
                self._dirstate_locked = True
 
1600
        self._locked += 1
 
1601
 
 
1602
    def _must_be_locked(self):
 
1603
        if not self._locked:
 
1604
            raise errors.ObjectNotLocked(self)
 
1605
 
 
1606
    @needs_read_lock
 
1607
    def path2id(self, path):
 
1608
        """Return the id for path in this tree."""
 
1609
        # lookup by path: faster than splitting and walking the ivnentory.
 
1610
        entry = self._get_entry(path=path)
 
1611
        if entry == (None, None):
 
1612
            return None
 
1613
        return entry[0][2]
 
1614
 
 
1615
    def unlock(self):
 
1616
        """Unlock, freeing any cache memory used during the lock."""
 
1617
        # outside of a lock, the inventory is suspect: release it.
 
1618
        self._locked -=1
 
1619
        if not self._locked:
 
1620
            self._inventory = None
 
1621
            self._locked = 0
 
1622
            if self._dirstate_locked:
 
1623
                self._dirstate.unlock()
 
1624
                self._dirstate_locked = False
 
1625
            self._repository.unlock()
 
1626
 
 
1627
    def walkdirs(self, prefix=""):
 
1628
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1629
        # implementation based on an inventory.  
 
1630
        # This should be cleaned up to use the much faster Dirstate code
 
1631
        # So for now, we just build up the parent inventory, and extract
 
1632
        # it the same way RevisionTree does.
 
1633
        _directory = 'directory'
 
1634
        inv = self._get_inventory()
 
1635
        top_id = inv.path2id(prefix)
 
1636
        if top_id is None:
 
1637
            pending = []
 
1638
        else:
 
1639
            pending = [(prefix, top_id)]
 
1640
        while pending:
 
1641
            dirblock = []
 
1642
            relpath, file_id = pending.pop()
 
1643
            # 0 - relpath, 1- file-id
 
1644
            if relpath:
 
1645
                relroot = relpath + '/'
 
1646
            else:
 
1647
                relroot = ""
 
1648
            # FIXME: stash the node in pending
 
1649
            entry = inv[file_id]
 
1650
            for name, child in entry.sorted_children():
 
1651
                toppath = relroot + name
 
1652
                dirblock.append((toppath, name, child.kind, None,
 
1653
                    child.file_id, child.kind
 
1654
                    ))
 
1655
            yield (relpath, entry.file_id), dirblock
 
1656
            # push the user specified dirs from dirblock
 
1657
            for dir in reversed(dirblock):
 
1658
                if dir[2] == _directory:
 
1659
                    pending.append((dir[0], dir[4]))
 
1660
 
 
1661
 
 
1662
class InterDirStateTree(InterTree):
 
1663
    """Fast path optimiser for changes_from with dirstate trees.
 
1664
    
 
1665
    This is used only when both trees are in the dirstate working file, and 
 
1666
    the source is any parent within the dirstate, and the destination is 
 
1667
    the current working tree of the same dirstate.
 
1668
    """
 
1669
    # this could be generalized to allow comparisons between any trees in the
 
1670
    # dirstate, and possibly between trees stored in different dirstates.
 
1671
 
 
1672
    def __init__(self, source, target):
 
1673
        super(InterDirStateTree, self).__init__(source, target)
 
1674
        if not InterDirStateTree.is_compatible(source, target):
 
1675
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1676
 
 
1677
    @staticmethod
 
1678
    def make_source_parent_tree(source, target):
 
1679
        """Change the source tree into a parent of the target."""
 
1680
        revid = source.commit('record tree')
 
1681
        target.branch.repository.fetch(source.branch.repository, revid)
 
1682
        target.set_parent_ids([revid])
 
1683
        return target.basis_tree(), target
 
1684
 
 
1685
    _matching_from_tree_format = WorkingTreeFormat4()
 
1686
    _matching_to_tree_format = WorkingTreeFormat4()
 
1687
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1688
 
 
1689
    def _iter_changes(self, include_unchanged=False,
 
1690
                      specific_files=None, pb=None, extra_trees=[],
 
1691
                      require_versioned=True, want_unversioned=False):
 
1692
        """Return the changes from source to target.
 
1693
 
 
1694
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1695
            for details.
 
1696
        :param specific_files: An optional list of file paths to restrict the
 
1697
            comparison to. When mapping filenames to ids, all matches in all
 
1698
            trees (including optional extra_trees) are used, and all children of
 
1699
            matched directories are included.
 
1700
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1701
            unchanged entries in the result.
 
1702
        :param extra_trees: An optional list of additional trees to use when
 
1703
            mapping the contents of specific_files (paths) to file_ids.
 
1704
        :param require_versioned: If True, all files in specific_files must be
 
1705
            versioned in one of source, target, extra_trees or
 
1706
            PathsNotVersionedError is raised.
 
1707
        :param want_unversioned: Should unversioned files be returned in the
 
1708
            output. An unversioned file is defined as one with (False, False)
 
1709
            for the versioned pair.
 
1710
        """
 
1711
        utf8_decode = cache_utf8._utf8_decode
 
1712
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1713
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1714
        # NB: show_status depends on being able to pass in non-versioned files
 
1715
        # and report them as unknown
 
1716
        # TODO: handle extra trees in the dirstate.
 
1717
        # TODO: handle comparisons as an empty tree as a different special
 
1718
        # case? mbp 20070226
 
1719
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1720
            or specific_files == []):
 
1721
            # we can't fast-path these cases (yet)
 
1722
            for f in super(InterDirStateTree, self)._iter_changes(
 
1723
                include_unchanged, specific_files, pb, extra_trees,
 
1724
                require_versioned, want_unversioned=want_unversioned):
 
1725
                yield f
 
1726
            return
 
1727
        parent_ids = self.target.get_parent_ids()
 
1728
        assert (self.source._revision_id in parent_ids), \
 
1729
                "revision {%s} is not stored in {%s}, but %s " \
 
1730
                "can only be used for trees stored in the dirstate" \
 
1731
                % (self.source._revision_id, self.target, self._iter_changes)
 
1732
        target_index = 0
 
1733
        if self.source._revision_id == NULL_REVISION:
 
1734
            source_index = None
 
1735
            indices = (target_index,)
 
1736
        else:
 
1737
            assert (self.source._revision_id in parent_ids), \
 
1738
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1739
                self.source._revision_id, parent_ids)
 
1740
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1741
            indices = (source_index,target_index)
 
1742
        # -- make all specific_files utf8 --
 
1743
        if specific_files:
 
1744
            specific_files_utf8 = set()
 
1745
            for path in specific_files:
 
1746
                specific_files_utf8.add(path.encode('utf8'))
 
1747
            specific_files = specific_files_utf8
 
1748
        else:
 
1749
            specific_files = set([''])
 
1750
        # -- specific_files is now a utf8 path set --
 
1751
        # -- get the state object and prepare it.
 
1752
        state = self.target.current_dirstate()
 
1753
        state._read_dirblocks_if_needed()
 
1754
        def _entries_for_path(path):
 
1755
            """Return a list with all the entries that match path for all ids.
 
1756
            """
 
1757
            dirname, basename = os.path.split(path)
 
1758
            key = (dirname, basename, '')
 
1759
            block_index, present = state._find_block_index_from_key(key)
 
1760
            if not present:
 
1761
                # the block which should contain path is absent.
 
1762
                return []
 
1763
            result = []
 
1764
            block = state._dirblocks[block_index][1]
 
1765
            entry_index, _ = state._find_entry_index(key, block)
 
1766
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1767
            while (entry_index < len(block) and
 
1768
                block[entry_index][0][0:2] == key[0:2]):
 
1769
                result.append(block[entry_index])
 
1770
                entry_index += 1
 
1771
            return result
 
1772
        if require_versioned:
 
1773
            # -- check all supplied paths are versioned in a search tree. --
 
1774
            all_versioned = True
 
1775
            for path in specific_files:
 
1776
                path_entries = _entries_for_path(path)
 
1777
                if not path_entries:
 
1778
                    # this specified path is not present at all: error
 
1779
                    all_versioned = False
 
1780
                    break
 
1781
                found_versioned = False
 
1782
                # for each id at this path
 
1783
                for entry in path_entries:
 
1784
                    # for each tree.
 
1785
                    for index in indices:
 
1786
                        if entry[1][index][0] != 'a': # absent
 
1787
                            found_versioned = True
 
1788
                            # all good: found a versioned cell
 
1789
                            break
 
1790
                if not found_versioned:
 
1791
                    # none of the indexes was not 'absent' at all ids for this
 
1792
                    # path.
 
1793
                    all_versioned = False
 
1794
                    break
 
1795
            if not all_versioned:
 
1796
                raise errors.PathsNotVersionedError(specific_files)
 
1797
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1798
        search_specific_files = set()
 
1799
        for path in specific_files:
 
1800
            other_specific_files = specific_files.difference(set([path]))
 
1801
            if not osutils.is_inside_any(other_specific_files, path):
 
1802
                # this is a top level path, we must check it.
 
1803
                search_specific_files.add(path)
 
1804
        # sketch: 
 
1805
        # compare source_index and target_index at or under each element of search_specific_files.
 
1806
        # follow the following comparison table. Note that we only want to do diff operations when
 
1807
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1808
        # for the target.
 
1809
        # cases:
 
1810
        # 
 
1811
        # Source | Target | disk | action
 
1812
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1813
        #        |        |      | diff check on source-target
 
1814
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1815
        #        |        |      | ???
 
1816
        #   r    |  a     |      | add source to search
 
1817
        #   r    |  a     |  a   | 
 
1818
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1819
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1820
        #   a    | fdlt   |      | add new id
 
1821
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1822
        #   a    |  a     |      | not present in either tree, skip
 
1823
        #   a    |  a     |  a   | not present in any tree, skip
 
1824
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1825
        #        |        |      | may not be selected by the users list of paths.
 
1826
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1827
        #        |        |      | may not be selected by the users list of paths.
 
1828
        #  fdlt  | fdlt   |      | content in both: diff them
 
1829
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1830
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1831
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1832
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1833
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1834
        #        |        |      | this id at the other path.
 
1835
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1836
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1837
        #        |        |      | this id at the other path.
 
1838
 
 
1839
        # for all search_indexs in each path at or under each element of
 
1840
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1841
        # relocated path as one to search if its not searched already. If the
 
1842
        # detail is not relocated, add the id.
 
1843
        searched_specific_files = set()
 
1844
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1845
        # Using a list so that we can access the values and change them in
 
1846
        # nested scope. Each one is [path, file_id, entry]
 
1847
        last_source_parent = [None, None]
 
1848
        last_target_parent = [None, None]
 
1849
 
 
1850
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1851
 
 
1852
        # Just a sentry, so that _process_entry can say that this
 
1853
        # record is handled, but isn't interesting to process (unchanged)
 
1854
        uninteresting = object()
 
1855
 
 
1856
 
 
1857
        old_dirname_to_file_id = {}
 
1858
        new_dirname_to_file_id = {}
 
1859
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1860
        #       keeping a cache of directories that we have seen.
 
1861
 
 
1862
        def _process_entry(entry, path_info):
 
1863
            """Compare an entry and real disk to generate delta information.
 
1864
 
 
1865
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1866
                the path of entry. If None, then the path is considered absent.
 
1867
                (Perhaps we should pass in a concrete entry for this ?)
 
1868
                Basename is returned as a utf8 string because we expect this
 
1869
                tuple will be ignored, and don't want to take the time to
 
1870
                decode.
 
1871
            :return: None if these don't match
 
1872
                     A tuple of information about the change, or
 
1873
                     the object 'uninteresting' if these match, but are
 
1874
                     basically identical.
 
1875
            """
 
1876
            if source_index is None:
 
1877
                source_details = NULL_PARENT_DETAILS
 
1878
            else:
 
1879
                source_details = entry[1][source_index]
 
1880
            target_details = entry[1][target_index]
 
1881
            target_minikind = target_details[0]
 
1882
            if path_info is not None and target_minikind in 'fdlt':
 
1883
                assert target_index == 0
 
1884
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1885
                                                  stat_value=path_info[3])
 
1886
                # The entry may have been modified by update_entry
 
1887
                target_details = entry[1][target_index]
 
1888
                target_minikind = target_details[0]
 
1889
            else:
 
1890
                link_or_sha1 = None
 
1891
            file_id = entry[0][2]
 
1892
            source_minikind = source_details[0]
 
1893
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1894
                # claimed content in both: diff
 
1895
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1896
                #        |        |      | diff check on source-target
 
1897
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1898
                #        |        |      | ???
 
1899
                if source_minikind in 'r':
 
1900
                    # add the source to the search path to find any children it
 
1901
                    # has.  TODO ? : only add if it is a container ?
 
1902
                    if not osutils.is_inside_any(searched_specific_files,
 
1903
                                                 source_details[1]):
 
1904
                        search_specific_files.add(source_details[1])
 
1905
                    # generate the old path; this is needed for stating later
 
1906
                    # as well.
 
1907
                    old_path = source_details[1]
 
1908
                    old_dirname, old_basename = os.path.split(old_path)
 
1909
                    path = pathjoin(entry[0][0], entry[0][1])
 
1910
                    old_entry = state._get_entry(source_index,
 
1911
                                                 path_utf8=old_path)
 
1912
                    # update the source details variable to be the real
 
1913
                    # location.
 
1914
                    source_details = old_entry[1][source_index]
 
1915
                    source_minikind = source_details[0]
 
1916
                else:
 
1917
                    old_dirname = entry[0][0]
 
1918
                    old_basename = entry[0][1]
 
1919
                    old_path = path = None
 
1920
                if path_info is None:
 
1921
                    # the file is missing on disk, show as removed.
 
1922
                    content_change = True
 
1923
                    target_kind = None
 
1924
                    target_exec = False
 
1925
                else:
 
1926
                    # source and target are both versioned and disk file is present.
 
1927
                    target_kind = path_info[2]
 
1928
                    if target_kind == 'directory':
 
1929
                        if path is None:
 
1930
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1931
                        new_dirname_to_file_id[path] = file_id
 
1932
                        if source_minikind != 'd':
 
1933
                            content_change = True
 
1934
                        else:
 
1935
                            # directories have no fingerprint
 
1936
                            content_change = False
 
1937
                        target_exec = False
 
1938
                    elif target_kind == 'file':
 
1939
                        if source_minikind != 'f':
 
1940
                            content_change = True
 
1941
                        else:
 
1942
                            # We could check the size, but we already have the
 
1943
                            # sha1 hash.
 
1944
                            content_change = (link_or_sha1 != source_details[1])
 
1945
                        # Target details is updated at update_entry time
 
1946
                        if use_filesystem_for_exec:
 
1947
                            # We don't need S_ISREG here, because we are sure
 
1948
                            # we are dealing with a file.
 
1949
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1950
                        else:
 
1951
                            target_exec = target_details[3]
 
1952
                    elif target_kind == 'symlink':
 
1953
                        if source_minikind != 'l':
 
1954
                            content_change = True
 
1955
                        else:
 
1956
                            content_change = (link_or_sha1 != source_details[1])
 
1957
                        target_exec = False
 
1958
                    elif target_kind == 'tree-reference':
 
1959
                        if source_minikind != 't':
 
1960
                            content_change = True
 
1961
                        else:
 
1962
                            content_change = False
 
1963
                        target_exec = False
 
1964
                    else:
 
1965
                        raise Exception, "unknown kind %s" % path_info[2]
 
1966
                if source_minikind == 'd':
 
1967
                    if path is None:
 
1968
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1969
                    old_dirname_to_file_id[old_path] = file_id
 
1970
                # parent id is the entry for the path in the target tree
 
1971
                if old_dirname == last_source_parent[0]:
 
1972
                    source_parent_id = last_source_parent[1]
 
1973
                else:
 
1974
                    try:
 
1975
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1976
                    except KeyError:
 
1977
                        source_parent_entry = state._get_entry(source_index,
 
1978
                                                               path_utf8=old_dirname)
 
1979
                        source_parent_id = source_parent_entry[0][2]
 
1980
                    if source_parent_id == entry[0][2]:
 
1981
                        # This is the root, so the parent is None
 
1982
                        source_parent_id = None
 
1983
                    else:
 
1984
                        last_source_parent[0] = old_dirname
 
1985
                        last_source_parent[1] = source_parent_id
 
1986
                new_dirname = entry[0][0]
 
1987
                if new_dirname == last_target_parent[0]:
 
1988
                    target_parent_id = last_target_parent[1]
 
1989
                else:
 
1990
                    try:
 
1991
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1992
                    except KeyError:
 
1993
                        # TODO: We don't always need to do the lookup, because the
 
1994
                        #       parent entry will be the same as the source entry.
 
1995
                        target_parent_entry = state._get_entry(target_index,
 
1996
                                                               path_utf8=new_dirname)
 
1997
                        assert target_parent_entry != (None, None), (
 
1998
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1999
                            % (new_dirname, entry))
 
2000
                        target_parent_id = target_parent_entry[0][2]
 
2001
                    if target_parent_id == entry[0][2]:
 
2002
                        # This is the root, so the parent is None
 
2003
                        target_parent_id = None
 
2004
                    else:
 
2005
                        last_target_parent[0] = new_dirname
 
2006
                        last_target_parent[1] = target_parent_id
 
2007
 
 
2008
                source_exec = source_details[3]
 
2009
                if (include_unchanged
 
2010
                    or content_change
 
2011
                    or source_parent_id != target_parent_id
 
2012
                    or old_basename != entry[0][1]
 
2013
                    or source_exec != target_exec
 
2014
                    ):
 
2015
                    if old_path is None:
 
2016
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2017
                        old_path_u = utf8_decode(old_path)[0]
 
2018
                        path_u = old_path_u
 
2019
                    else:
 
2020
                        old_path_u = utf8_decode(old_path)[0]
 
2021
                        if old_path == path:
 
2022
                            path_u = old_path_u
 
2023
                        else:
 
2024
                            path_u = utf8_decode(path)[0]
 
2025
                    source_kind = _minikind_to_kind[source_minikind]
 
2026
                    return (entry[0][2],
 
2027
                           (old_path_u, path_u),
 
2028
                           content_change,
 
2029
                           (True, True),
 
2030
                           (source_parent_id, target_parent_id),
 
2031
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2032
                           (source_kind, target_kind),
 
2033
                           (source_exec, target_exec))
 
2034
                else:
 
2035
                    return uninteresting
 
2036
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2037
                # looks like a new file
 
2038
                if path_info is not None:
 
2039
                    path = pathjoin(entry[0][0], entry[0][1])
 
2040
                    # parent id is the entry for the path in the target tree
 
2041
                    # TODO: these are the same for an entire directory: cache em.
 
2042
                    parent_id = state._get_entry(target_index,
 
2043
                                                 path_utf8=entry[0][0])[0][2]
 
2044
                    if parent_id == entry[0][2]:
 
2045
                        parent_id = None
 
2046
                    if use_filesystem_for_exec:
 
2047
                        # We need S_ISREG here, because we aren't sure if this
 
2048
                        # is a file or not.
 
2049
                        target_exec = bool(
 
2050
                            stat.S_ISREG(path_info[3].st_mode)
 
2051
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2052
                    else:
 
2053
                        target_exec = target_details[3]
 
2054
                    return (entry[0][2],
 
2055
                           (None, utf8_decode(path)[0]),
 
2056
                           True,
 
2057
                           (False, True),
 
2058
                           (None, parent_id),
 
2059
                           (None, utf8_decode(entry[0][1])[0]),
 
2060
                           (None, path_info[2]),
 
2061
                           (None, target_exec))
 
2062
                else:
 
2063
                    # but its not on disk: we deliberately treat this as just
 
2064
                    # never-present. (Why ?! - RBC 20070224)
 
2065
                    pass
 
2066
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2067
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2068
                # if its still on disk, *and* theres no other entry at this
 
2069
                # path [we dont know this in this routine at the moment -
 
2070
                # perhaps we should change this - then it would be an unknown.
 
2071
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2072
                # parent id is the entry for the path in the target tree
 
2073
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2074
                if parent_id == entry[0][2]:
 
2075
                    parent_id = None
 
2076
                return (entry[0][2],
 
2077
                       (utf8_decode(old_path)[0], None),
 
2078
                       True,
 
2079
                       (True, False),
 
2080
                       (parent_id, None),
 
2081
                       (utf8_decode(entry[0][1])[0], None),
 
2082
                       (_minikind_to_kind[source_minikind], None),
 
2083
                       (source_details[3], None))
 
2084
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2085
                # a rename; could be a true rename, or a rename inherited from
 
2086
                # a renamed parent. TODO: handle this efficiently. Its not
 
2087
                # common case to rename dirs though, so a correct but slow
 
2088
                # implementation will do.
 
2089
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2090
                    search_specific_files.add(target_details[1])
 
2091
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2092
                # neither of the selected trees contain this file,
 
2093
                # so skip over it. This is not currently directly tested, but
 
2094
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2095
                pass
 
2096
            else:
 
2097
                raise AssertionError("don't know how to compare "
 
2098
                    "source_minikind=%r, target_minikind=%r"
 
2099
                    % (source_minikind, target_minikind))
 
2100
                ## import pdb;pdb.set_trace()
 
2101
            return None
 
2102
 
 
2103
        while search_specific_files:
 
2104
            # TODO: the pending list should be lexically sorted?  the
 
2105
            # interface doesn't require it.
 
2106
            current_root = search_specific_files.pop()
 
2107
            current_root_unicode = current_root.decode('utf8')
 
2108
            searched_specific_files.add(current_root)
 
2109
            # process the entries for this containing directory: the rest will be
 
2110
            # found by their parents recursively.
 
2111
            root_entries = _entries_for_path(current_root)
 
2112
            root_abspath = self.target.abspath(current_root_unicode)
 
2113
            try:
 
2114
                root_stat = os.lstat(root_abspath)
 
2115
            except OSError, e:
 
2116
                if e.errno == errno.ENOENT:
 
2117
                    # the path does not exist: let _process_entry know that.
 
2118
                    root_dir_info = None
 
2119
                else:
 
2120
                    # some other random error: hand it up.
 
2121
                    raise
 
2122
            else:
 
2123
                root_dir_info = ('', current_root,
 
2124
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2125
                    root_abspath)
 
2126
                if root_dir_info[2] == 'directory':
 
2127
                    if self.target._directory_is_tree_reference(
 
2128
                        current_root.decode('utf8')):
 
2129
                        root_dir_info = root_dir_info[:2] + \
 
2130
                            ('tree-reference',) + root_dir_info[3:]
 
2131
 
 
2132
            if not root_entries and not root_dir_info:
 
2133
                # this specified path is not present at all, skip it.
 
2134
                continue
 
2135
            path_handled = False
 
2136
            for entry in root_entries:
 
2137
                result = _process_entry(entry, root_dir_info)
 
2138
                if result is not None:
 
2139
                    path_handled = True
 
2140
                    if result is not uninteresting:
 
2141
                        yield result
 
2142
            if want_unversioned and not path_handled and root_dir_info:
 
2143
                new_executable = bool(
 
2144
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2145
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2146
                yield (None,
 
2147
                       (None, current_root_unicode),
 
2148
                       True,
 
2149
                       (False, False),
 
2150
                       (None, None),
 
2151
                       (None, splitpath(current_root_unicode)[-1]),
 
2152
                       (None, root_dir_info[2]),
 
2153
                       (None, new_executable)
 
2154
                      )
 
2155
            initial_key = (current_root, '', '')
 
2156
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2157
            if block_index == 0:
 
2158
                # we have processed the total root already, but because the
 
2159
                # initial key matched it we should skip it here.
 
2160
                block_index +=1
 
2161
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2162
                current_dir_info = None
 
2163
            else:
 
2164
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2165
                try:
 
2166
                    current_dir_info = dir_iterator.next()
 
2167
                except OSError, e:
 
2168
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2169
                    # python 2.5 has e.errno == EINVAL,
 
2170
                    #            and e.winerror == ERROR_DIRECTORY
 
2171
                    e_winerror = getattr(e, 'winerror', None)
 
2172
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2173
                    # there may be directories in the inventory even though
 
2174
                    # this path is not a file on disk: so mark it as end of
 
2175
                    # iterator
 
2176
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2177
                        current_dir_info = None
 
2178
                    elif (sys.platform == 'win32'
 
2179
                          and (e.errno in win_errors
 
2180
                               or e_winerror in win_errors)):
 
2181
                        current_dir_info = None
 
2182
                    else:
 
2183
                        raise
 
2184
                else:
 
2185
                    if current_dir_info[0][0] == '':
 
2186
                        # remove .bzr from iteration
 
2187
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2188
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2189
                        del current_dir_info[1][bzr_index]
 
2190
            # walk until both the directory listing and the versioned metadata
 
2191
            # are exhausted. 
 
2192
            if (block_index < len(state._dirblocks) and
 
2193
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2194
                current_block = state._dirblocks[block_index]
 
2195
            else:
 
2196
                current_block = None
 
2197
            while (current_dir_info is not None or
 
2198
                   current_block is not None):
 
2199
                if (current_dir_info and current_block
 
2200
                    and current_dir_info[0][0] != current_block[0]):
 
2201
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2202
                        # filesystem data refers to paths not covered by the dirblock.
 
2203
                        # this has two possibilities:
 
2204
                        # A) it is versioned but empty, so there is no block for it
 
2205
                        # B) it is not versioned.
 
2206
 
 
2207
                        # if (A) then we need to recurse into it to check for
 
2208
                        # new unknown files or directories.
 
2209
                        # if (B) then we should ignore it, because we don't
 
2210
                        # recurse into unknown directories.
 
2211
                        path_index = 0
 
2212
                        while path_index < len(current_dir_info[1]):
 
2213
                                current_path_info = current_dir_info[1][path_index]
 
2214
                                if want_unversioned:
 
2215
                                    if current_path_info[2] == 'directory':
 
2216
                                        if self.target._directory_is_tree_reference(
 
2217
                                            current_path_info[0].decode('utf8')):
 
2218
                                            current_path_info = current_path_info[:2] + \
 
2219
                                                ('tree-reference',) + current_path_info[3:]
 
2220
                                    new_executable = bool(
 
2221
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2222
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2223
                                    yield (None,
 
2224
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2225
                                        True,
 
2226
                                        (False, False),
 
2227
                                        (None, None),
 
2228
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2229
                                        (None, current_path_info[2]),
 
2230
                                        (None, new_executable))
 
2231
                                # dont descend into this unversioned path if it is
 
2232
                                # a dir
 
2233
                                if current_path_info[2] in ('directory',
 
2234
                                                            'tree-reference'):
 
2235
                                    del current_dir_info[1][path_index]
 
2236
                                    path_index -= 1
 
2237
                                path_index += 1
 
2238
 
 
2239
                        # This dir info has been handled, go to the next
 
2240
                        try:
 
2241
                            current_dir_info = dir_iterator.next()
 
2242
                        except StopIteration:
 
2243
                            current_dir_info = None
 
2244
                    else:
 
2245
                        # We have a dirblock entry for this location, but there
 
2246
                        # is no filesystem path for this. This is most likely
 
2247
                        # because a directory was removed from the disk.
 
2248
                        # We don't have to report the missing directory,
 
2249
                        # because that should have already been handled, but we
 
2250
                        # need to handle all of the files that are contained
 
2251
                        # within.
 
2252
                        for current_entry in current_block[1]:
 
2253
                            # entry referring to file not present on disk.
 
2254
                            # advance the entry only, after processing.
 
2255
                            result = _process_entry(current_entry, None)
 
2256
                            if result is not None:
 
2257
                                if result is not uninteresting:
 
2258
                                    yield result
 
2259
                        block_index +=1
 
2260
                        if (block_index < len(state._dirblocks) and
 
2261
                            osutils.is_inside(current_root,
 
2262
                                              state._dirblocks[block_index][0])):
 
2263
                            current_block = state._dirblocks[block_index]
 
2264
                        else:
 
2265
                            current_block = None
 
2266
                    continue
 
2267
                entry_index = 0
 
2268
                if current_block and entry_index < len(current_block[1]):
 
2269
                    current_entry = current_block[1][entry_index]
 
2270
                else:
 
2271
                    current_entry = None
 
2272
                advance_entry = True
 
2273
                path_index = 0
 
2274
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2275
                    current_path_info = current_dir_info[1][path_index]
 
2276
                    if current_path_info[2] == 'directory':
 
2277
                        if self.target._directory_is_tree_reference(
 
2278
                            current_path_info[0].decode('utf8')):
 
2279
                            current_path_info = current_path_info[:2] + \
 
2280
                                ('tree-reference',) + current_path_info[3:]
 
2281
                else:
 
2282
                    current_path_info = None
 
2283
                advance_path = True
 
2284
                path_handled = False
 
2285
                while (current_entry is not None or
 
2286
                    current_path_info is not None):
 
2287
                    if current_entry is None:
 
2288
                        # the check for path_handled when the path is adnvaced
 
2289
                        # will yield this path if needed.
 
2290
                        pass
 
2291
                    elif current_path_info is None:
 
2292
                        # no path is fine: the per entry code will handle it.
 
2293
                        result = _process_entry(current_entry, current_path_info)
 
2294
                        if result is not None:
 
2295
                            if result is not uninteresting:
 
2296
                                yield result
 
2297
                    elif (current_entry[0][1] != current_path_info[1]
 
2298
                          or current_entry[1][target_index][0] in 'ar'):
 
2299
                        # The current path on disk doesn't match the dirblock
 
2300
                        # record. Either the dirblock is marked as absent, or
 
2301
                        # the file on disk is not present at all in the
 
2302
                        # dirblock. Either way, report about the dirblock
 
2303
                        # entry, and let other code handle the filesystem one.
 
2304
 
 
2305
                        # Compare the basename for these files to determine
 
2306
                        # which comes first
 
2307
                        if current_path_info[1] < current_entry[0][1]:
 
2308
                            # extra file on disk: pass for now, but only
 
2309
                            # increment the path, not the entry
 
2310
                            advance_entry = False
 
2311
                        else:
 
2312
                            # entry referring to file not present on disk.
 
2313
                            # advance the entry only, after processing.
 
2314
                            result = _process_entry(current_entry, None)
 
2315
                            if result is not None:
 
2316
                                if result is not uninteresting:
 
2317
                                    yield result
 
2318
                            advance_path = False
 
2319
                    else:
 
2320
                        result = _process_entry(current_entry, current_path_info)
 
2321
                        if result is not None:
 
2322
                            path_handled = True
 
2323
                            if result is not uninteresting:
 
2324
                                yield result
 
2325
                    if advance_entry and current_entry is not None:
 
2326
                        entry_index += 1
 
2327
                        if entry_index < len(current_block[1]):
 
2328
                            current_entry = current_block[1][entry_index]
 
2329
                        else:
 
2330
                            current_entry = None
 
2331
                    else:
 
2332
                        advance_entry = True # reset the advance flaga
 
2333
                    if advance_path and current_path_info is not None:
 
2334
                        if not path_handled:
 
2335
                            # unversioned in all regards
 
2336
                            if want_unversioned:
 
2337
                                new_executable = bool(
 
2338
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2339
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2340
                                yield (None,
 
2341
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2342
                                    True,
 
2343
                                    (False, False),
 
2344
                                    (None, None),
 
2345
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2346
                                    (None, current_path_info[2]),
 
2347
                                    (None, new_executable))
 
2348
                            # dont descend into this unversioned path if it is
 
2349
                            # a dir
 
2350
                            if current_path_info[2] in ('directory'):
 
2351
                                del current_dir_info[1][path_index]
 
2352
                                path_index -= 1
 
2353
                        # dont descend the disk iterator into any tree 
 
2354
                        # paths.
 
2355
                        if current_path_info[2] == 'tree-reference':
 
2356
                            del current_dir_info[1][path_index]
 
2357
                            path_index -= 1
 
2358
                        path_index += 1
 
2359
                        if path_index < len(current_dir_info[1]):
 
2360
                            current_path_info = current_dir_info[1][path_index]
 
2361
                            if current_path_info[2] == 'directory':
 
2362
                                if self.target._directory_is_tree_reference(
 
2363
                                    current_path_info[0].decode('utf8')):
 
2364
                                    current_path_info = current_path_info[:2] + \
 
2365
                                        ('tree-reference',) + current_path_info[3:]
 
2366
                        else:
 
2367
                            current_path_info = None
 
2368
                        path_handled = False
 
2369
                    else:
 
2370
                        advance_path = True # reset the advance flagg.
 
2371
                if current_block is not None:
 
2372
                    block_index += 1
 
2373
                    if (block_index < len(state._dirblocks) and
 
2374
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2375
                        current_block = state._dirblocks[block_index]
 
2376
                    else:
 
2377
                        current_block = None
 
2378
                if current_dir_info is not None:
 
2379
                    try:
 
2380
                        current_dir_info = dir_iterator.next()
 
2381
                    except StopIteration:
 
2382
                        current_dir_info = None
 
2383
 
 
2384
 
 
2385
    @staticmethod
 
2386
    def is_compatible(source, target):
 
2387
        # the target must be a dirstate working tree
 
2388
        if not isinstance(target, WorkingTree4):
 
2389
            return False
 
2390
        # the source must be a revtreee or dirstate rev tree.
 
2391
        if not isinstance(source,
 
2392
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2393
            return False
 
2394
        # the source revid must be in the target dirstate
 
2395
        if not (source._revision_id == NULL_REVISION or
 
2396
            source._revision_id in target.get_parent_ids()):
 
2397
            # TODO: what about ghosts? it may well need to 
 
2398
            # check for them explicitly.
 
2399
            return False
 
2400
        return True
 
2401
 
 
2402
InterTree.register_optimiser(InterDirStateTree)
 
2403
 
 
2404
 
 
2405
class Converter3to4(object):
 
2406
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2407
 
 
2408
    def __init__(self):
 
2409
        self.target_format = WorkingTreeFormat4()
 
2410
 
 
2411
    def convert(self, tree):
 
2412
        # lock the control files not the tree, so that we dont get tree
 
2413
        # on-unlock behaviours, and so that noone else diddles with the 
 
2414
        # tree during upgrade.
 
2415
        tree._control_files.lock_write()
 
2416
        try:
 
2417
            tree.read_working_inventory()
 
2418
            self.create_dirstate_data(tree)
 
2419
            self.update_format(tree)
 
2420
            self.remove_xml_files(tree)
 
2421
        finally:
 
2422
            tree._control_files.unlock()
 
2423
 
 
2424
    def create_dirstate_data(self, tree):
 
2425
        """Create the dirstate based data for tree."""
 
2426
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2427
            ).local_abspath('dirstate')
 
2428
        state = dirstate.DirState.from_tree(tree, local_path)
 
2429
        state.save()
 
2430
        state.unlock()
 
2431
 
 
2432
    def remove_xml_files(self, tree):
 
2433
        """Remove the oldformat 3 data."""
 
2434
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2435
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2436
            'pending-merges', 'stat-cache']:
 
2437
            try:
 
2438
                transport.delete(path)
 
2439
            except errors.NoSuchFile:
 
2440
                # some files are optional - just deal.
 
2441
                pass
 
2442
 
 
2443
    def update_format(self, tree):
 
2444
        """Change the format marker."""
 
2445
        tree._control_files.put_utf8('format',
 
2446
            self.target_format.get_format_string())