~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Martin Pool
  • Date: 2007-10-24 02:33:14 UTC
  • mto: This revision was merged to the branch mainline in revision 2933.
  • Revision ID: mbp@sourcefrog.net-20071024023314-l9mscm8wsb1bvj1f
typo

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
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    trace,
 
59
    transform,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
 
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
 
67
""")
 
68
 
 
69
from bzrlib import symbol_versioning
 
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
 
76
from bzrlib.osutils import (
 
77
    file_kind,
 
78
    isdir,
 
79
    normpath,
 
80
    pathjoin,
 
81
    rand_chars,
 
82
    realpath,
 
83
    safe_unicode,
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
 
87
from bzrlib.transport.local import LocalTransport
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
 
120
    def __init__(self, basedir,
 
121
                 branch,
 
122
                 _control_files=None,
 
123
                 _format=None,
 
124
                 _bzrdir=None):
 
125
        """Construct a WorkingTree for basedir.
 
126
 
 
127
        If the branch is not supplied, it is opened automatically.
 
128
        If the branch is supplied, it must be the branch for this basedir.
 
129
        (branch.base is not cross checked, because for remote branches that
 
130
        would be meaningless).
 
131
        """
 
132
        self._format = _format
 
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
 
136
        basedir = safe_unicode(basedir)
 
137
        mutter("opening working tree %r", basedir)
 
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
 
141
        self.basedir = realpath(basedir)
 
142
        # if branch is at our basedir and is a format 6 or less
 
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
 
146
        self._control_files = _control_files
 
147
        self._dirty = None
 
148
        #-------------
 
149
        # during a read or write lock these objects are set, and are
 
150
        # None the rest of the time.
 
151
        self._dirstate = None
 
152
        self._inventory = None
 
153
        #-------------
 
154
 
 
155
    @needs_tree_write_lock
 
156
    def _add(self, files, ids, kinds):
 
157
        """See MutableTree._add."""
 
158
        state = self.current_dirstate()
 
159
        for f, file_id, kind in zip(files, ids, kinds):
 
160
            f = f.strip('/')
 
161
            if self.path2id(f):
 
162
                # special case tree root handling.
 
163
                if f == '' and self.path2id(f) == ROOT_ID:
 
164
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
165
                continue
 
166
            if file_id is None:
 
167
                file_id = generate_ids.gen_file_id(f)
 
168
            # deliberately add the file with no cached stat or sha1
 
169
            # - on the first access it will be gathered, and we can
 
170
            # always change this once tests are all passing.
 
171
            state.add(f, file_id, kind, None, '')
 
172
        self._make_dirty(reset_inventory=True)
 
173
 
 
174
    def _make_dirty(self, reset_inventory):
 
175
        """Make the tree state dirty.
 
176
 
 
177
        :param reset_inventory: True if the cached inventory should be removed
 
178
            (presuming there is one).
 
179
        """
 
180
        self._dirty = True
 
181
        if reset_inventory and self._inventory is not None:
 
182
            self._inventory = None
 
183
 
 
184
    @needs_tree_write_lock
 
185
    def add_reference(self, sub_tree):
 
186
        # use standard implementation, which calls back to self._add
 
187
        # 
 
188
        # So we don't store the reference_revision in the working dirstate,
 
189
        # it's just recorded at the moment of commit. 
 
190
        self._add_reference(sub_tree)
 
191
 
 
192
    def break_lock(self):
 
193
        """Break a lock if one is present from another instance.
 
194
 
 
195
        Uses the ui factory to ask for confirmation if the lock may be from
 
196
        an active process.
 
197
 
 
198
        This will probe the repository for its lock as well.
 
199
        """
 
200
        # if the dirstate is locked by an active process, reject the break lock
 
201
        # call.
 
202
        try:
 
203
            if self._dirstate is None:
 
204
                clear = True
 
205
            else:
 
206
                clear = False
 
207
            state = self._current_dirstate()
 
208
            if state._lock_token is not None:
 
209
                # we already have it locked. sheese, cant break our own lock.
 
210
                raise errors.LockActive(self.basedir)
 
211
            else:
 
212
                try:
 
213
                    # try for a write lock - need permission to get one anyhow
 
214
                    # to break locks.
 
215
                    state.lock_write()
 
216
                except errors.LockContention:
 
217
                    # oslocks fail when a process is still live: fail.
 
218
                    # TODO: get the locked lockdir info and give to the user to
 
219
                    # assist in debugging.
 
220
                    raise errors.LockActive(self.basedir)
 
221
                else:
 
222
                    state.unlock()
 
223
        finally:
 
224
            if clear:
 
225
                self._dirstate = None
 
226
        self._control_files.break_lock()
 
227
        self.branch.break_lock()
 
228
 
 
229
    def _comparison_data(self, entry, path):
 
230
        kind, executable, stat_value = \
 
231
            WorkingTree3._comparison_data(self, entry, path)
 
232
        # it looks like a plain directory, but it's really a reference -- see
 
233
        # also kind()
 
234
        if (self._repo_supports_tree_reference and
 
235
            kind == 'directory' and
 
236
            self._directory_is_tree_reference(path)):
 
237
            kind = 'tree-reference'
 
238
        return kind, executable, stat_value
 
239
 
 
240
    @needs_write_lock
 
241
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
242
        # mark the tree as dirty post commit - commit
 
243
        # can change the current versioned list by doing deletes.
 
244
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
245
        self._make_dirty(reset_inventory=True)
 
246
        return result
 
247
 
 
248
    def current_dirstate(self):
 
249
        """Return the current dirstate object.
 
250
 
 
251
        This is not part of the tree interface and only exposed for ease of
 
252
        testing.
 
253
 
 
254
        :raises errors.NotWriteLocked: when not in a lock.
 
255
        """
 
256
        self._must_be_locked()
 
257
        return self._current_dirstate()
 
258
 
 
259
    def _current_dirstate(self):
 
260
        """Internal function that does not check lock status.
 
261
 
 
262
        This is needed for break_lock which also needs the dirstate.
 
263
        """
 
264
        if self._dirstate is not None:
 
265
            return self._dirstate
 
266
        local_path = self.bzrdir.get_workingtree_transport(None
 
267
            ).local_abspath('dirstate')
 
268
        self._dirstate = dirstate.DirState.on_file(local_path)
 
269
        return self._dirstate
 
270
 
 
271
    def filter_unversioned_files(self, paths):
 
272
        """Filter out paths that are versioned.
 
273
 
 
274
        :return: set of paths.
 
275
        """
 
276
        # TODO: make a generic multi-bisect routine roughly that should list
 
277
        # the paths, then process one half at a time recursively, and feed the
 
278
        # results of each bisect in further still
 
279
        paths = sorted(paths)
 
280
        result = set()
 
281
        state = self.current_dirstate()
 
282
        # TODO we want a paths_to_dirblocks helper I think
 
283
        for path in paths:
 
284
            dirname, basename = os.path.split(path.encode('utf8'))
 
285
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
286
                dirname, basename, 0)
 
287
            if not path_is_versioned:
 
288
                result.add(path)
 
289
        return result
 
290
 
 
291
    def flush(self):
 
292
        """Write all cached data to disk."""
 
293
        if self._control_files._lock_mode != 'w':
 
294
            raise errors.NotWriteLocked(self)
 
295
        self.current_dirstate().save()
 
296
        self._inventory = None
 
297
        self._dirty = False
 
298
 
 
299
    @needs_tree_write_lock
 
300
    def _gather_kinds(self, files, kinds):
 
301
        """See MutableTree._gather_kinds."""
 
302
        for pos, f in enumerate(files):
 
303
            if kinds[pos] is None:
 
304
                kinds[pos] = self._kind(f)
 
305
 
 
306
    def _generate_inventory(self):
 
307
        """Create and set self.inventory from the dirstate object.
 
308
        
 
309
        This is relatively expensive: we have to walk the entire dirstate.
 
310
        Ideally we would not, and can deprecate this function.
 
311
        """
 
312
        #: uncomment to trap on inventory requests.
 
313
        # import pdb;pdb.set_trace()
 
314
        state = self.current_dirstate()
 
315
        state._read_dirblocks_if_needed()
 
316
        root_key, current_entry = self._get_entry(path='')
 
317
        current_id = root_key[2]
 
318
        assert current_entry[0][0] == 'd' # directory
 
319
        inv = Inventory(root_id=current_id)
 
320
        # Turn some things into local variables
 
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
322
        factory = entry_factory
 
323
        utf8_decode = cache_utf8._utf8_decode
 
324
        inv_byid = inv._byid
 
325
        # we could do this straight out of the dirstate; it might be fast
 
326
        # and should be profiled - RBC 20070216
 
327
        parent_ies = {'' : inv.root}
 
328
        for block in state._dirblocks[1:]: # skip the root
 
329
            dirname = block[0]
 
330
            try:
 
331
                parent_ie = parent_ies[dirname]
 
332
            except KeyError:
 
333
                # all the paths in this block are not versioned in this tree
 
334
                continue
 
335
            for key, entry in block[1]:
 
336
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
337
                if minikind in ('a', 'r'): # absent, relocated
 
338
                    # a parent tree only entry
 
339
                    continue
 
340
                name = key[1]
 
341
                name_unicode = utf8_decode(name)[0]
 
342
                file_id = key[2]
 
343
                kind = minikind_to_kind[minikind]
 
344
                inv_entry = factory[kind](file_id, name_unicode,
 
345
                                          parent_ie.file_id)
 
346
                if kind == 'file':
 
347
                    # This is only needed on win32, where this is the only way
 
348
                    # we know the executable bit.
 
349
                    inv_entry.executable = executable
 
350
                    # not strictly needed: working tree
 
351
                    #inv_entry.text_size = size
 
352
                    #inv_entry.text_sha1 = sha1
 
353
                elif kind == 'directory':
 
354
                    # add this entry to the parent map.
 
355
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
356
                elif kind == 'tree-reference':
 
357
                    assert self._repo_supports_tree_reference, \
 
358
                        "repository of %r " \
 
359
                        "doesn't support tree references " \
 
360
                        "required by entry %r" \
 
361
                        % (self, name)
 
362
                    inv_entry.reference_revision = link_or_sha1 or None
 
363
                elif kind != 'symlink':
 
364
                    raise AssertionError("unknown kind %r" % kind)
 
365
                # These checks cost us around 40ms on a 55k entry tree
 
366
                assert file_id not in inv_byid, ('file_id %s already in'
 
367
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                assert name_unicode not in parent_ie.children
 
369
                inv_byid[file_id] = inv_entry
 
370
                parent_ie.children[name_unicode] = inv_entry
 
371
        self._inventory = inv
 
372
 
 
373
    def _get_entry(self, file_id=None, path=None):
 
374
        """Get the dirstate row for file_id or path.
 
375
 
 
376
        If either file_id or path is supplied, it is used as the key to lookup.
 
377
        If both are supplied, the fastest lookup is used, and an error is
 
378
        raised if they do not both point at the same row.
 
379
        
 
380
        :param file_id: An optional unicode file_id to be looked up.
 
381
        :param path: An optional unicode path to be looked up.
 
382
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
383
        """
 
384
        if file_id is None and path is None:
 
385
            raise errors.BzrError('must supply file_id or path')
 
386
        state = self.current_dirstate()
 
387
        if path is not None:
 
388
            path = path.encode('utf8')
 
389
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
390
 
 
391
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
392
        # check file id is valid unconditionally.
 
393
        entry = self._get_entry(file_id=file_id, path=path)
 
394
        if entry[0] is None:
 
395
            raise errors.NoSuchId(self, file_id)
 
396
        if path is None:
 
397
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
398
 
 
399
        file_abspath = self.abspath(path)
 
400
        state = self.current_dirstate()
 
401
        if stat_value is None:
 
402
            try:
 
403
                stat_value = os.lstat(file_abspath)
 
404
            except OSError, e:
 
405
                if e.errno == errno.ENOENT:
 
406
                    return None
 
407
                else:
 
408
                    raise
 
409
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
410
                                          stat_value=stat_value)
 
411
        if entry[1][0][0] == 'f':
 
412
            return link_or_sha1
 
413
        return None
 
414
 
 
415
    def _get_inventory(self):
 
416
        """Get the inventory for the tree. This is only valid within a lock."""
 
417
        if 'evil' in debug.debug_flags:
 
418
            trace.mutter_callsite(2,
 
419
                "accessing .inventory forces a size of tree translation.")
 
420
        if self._inventory is not None:
 
421
            return self._inventory
 
422
        self._must_be_locked()
 
423
        self._generate_inventory()
 
424
        return self._inventory
 
425
 
 
426
    inventory = property(_get_inventory,
 
427
                         doc="Inventory of this Tree")
 
428
 
 
429
    @needs_read_lock
 
430
    def get_parent_ids(self):
 
431
        """See Tree.get_parent_ids.
 
432
        
 
433
        This implementation requests the ids list from the dirstate file.
 
434
        """
 
435
        return self.current_dirstate().get_parent_ids()
 
436
 
 
437
    def get_reference_revision(self, file_id, path=None):
 
438
        # referenced tree's revision is whatever's currently there
 
439
        return self.get_nested_tree(file_id, path).last_revision()
 
440
 
 
441
    def get_nested_tree(self, file_id, path=None):
 
442
        if path is None:
 
443
            path = self.id2path(file_id)
 
444
        # else: check file_id is at path?
 
445
        return WorkingTree.open(self.abspath(path))
 
446
 
 
447
    @needs_read_lock
 
448
    def get_root_id(self):
 
449
        """Return the id of this trees root"""
 
450
        return self._get_entry(path='')[0][2]
 
451
 
 
452
    def has_id(self, file_id):
 
453
        state = self.current_dirstate()
 
454
        row, parents = self._get_entry(file_id=file_id)
 
455
        if row is None:
 
456
            return False
 
457
        return osutils.lexists(pathjoin(
 
458
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
459
 
 
460
    @needs_read_lock
 
461
    def id2path(self, file_id):
 
462
        "Convert a file-id to a path."
 
463
        state = self.current_dirstate()
 
464
        entry = self._get_entry(file_id=file_id)
 
465
        if entry == (None, None):
 
466
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
467
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
468
        return path_utf8.decode('utf8')
 
469
 
 
470
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
471
        entry = self._get_entry(path=path)
 
472
        if entry == (None, None):
 
473
            return False # Missing entries are not executable
 
474
        return entry[1][0][3] # Executable?
 
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
            entry = self._get_entry(file_id=file_id, path=path)
 
483
            if entry == (None, None):
 
484
                return False
 
485
            return entry[1][0][3]
 
486
 
 
487
        _is_executable_from_path_and_stat = \
 
488
            _is_executable_from_path_and_stat_from_basis
 
489
    else:
 
490
        def is_executable(self, file_id, path=None):
 
491
            """Test if a file is executable or not.
 
492
 
 
493
            Note: The caller is expected to take a read-lock before calling this.
 
494
            """
 
495
            if not path:
 
496
                path = self.id2path(file_id)
 
497
            mode = os.lstat(self.abspath(path)).st_mode
 
498
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
499
 
 
500
    @needs_read_lock
 
501
    def __iter__(self):
 
502
        """Iterate through file_ids for this tree.
 
503
 
 
504
        file_ids are in a WorkingTree if they are in the working inventory
 
505
        and the working file exists.
 
506
        """
 
507
        result = []
 
508
        for key, tree_details in self.current_dirstate()._iter_entries():
 
509
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
510
                # not relevant to the working tree
 
511
                continue
 
512
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
513
            if osutils.lexists(path):
 
514
                result.append(key[2])
 
515
        return iter(result)
 
516
 
 
517
    def iter_references(self):
 
518
        for key, tree_details in self.current_dirstate()._iter_entries():
 
519
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
520
                # not relevant to the working tree
 
521
                continue
 
522
            if not key[1]:
 
523
                # the root is not a reference.
 
524
                continue
 
525
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
526
            try:
 
527
                if self._kind(path) == 'tree-reference':
 
528
                    yield path, key[2]
 
529
            except errors.NoSuchFile:
 
530
                # path is missing on disk.
 
531
                continue
 
532
 
 
533
    def kind(self, file_id):
 
534
        """Return the kind of a file.
 
535
 
 
536
        This is always the actual kind that's on disk, regardless of what it
 
537
        was added as.
 
538
 
 
539
        Note: The caller is expected to take a read-lock before calling this.
 
540
        """
 
541
        relpath = self.id2path(file_id)
 
542
        assert relpath != None, \
 
543
            "path for id {%s} is None!" % file_id
 
544
        return self._kind(relpath)
 
545
 
 
546
    def _kind(self, relpath):
 
547
        abspath = self.abspath(relpath)
 
548
        kind = file_kind(abspath)
 
549
        if (self._repo_supports_tree_reference and
 
550
            kind == 'directory' and
 
551
            self._directory_is_tree_reference(relpath)):
 
552
            kind = 'tree-reference'
 
553
        return kind
 
554
 
 
555
    @needs_read_lock
 
556
    def _last_revision(self):
 
557
        """See Mutable.last_revision."""
 
558
        parent_ids = self.current_dirstate().get_parent_ids()
 
559
        if parent_ids:
 
560
            return parent_ids[0]
 
561
        else:
 
562
            return _mod_revision.NULL_REVISION
 
563
 
 
564
    def lock_read(self):
 
565
        """See Branch.lock_read, and WorkingTree.unlock."""
 
566
        self.branch.lock_read()
 
567
        try:
 
568
            self._control_files.lock_read()
 
569
            try:
 
570
                state = self.current_dirstate()
 
571
                if not state._lock_token:
 
572
                    state.lock_read()
 
573
                # set our support for tree references from the repository in
 
574
                # use.
 
575
                self._repo_supports_tree_reference = getattr(
 
576
                    self.branch.repository._format, "supports_tree_reference",
 
577
                    False)
 
578
            except:
 
579
                self._control_files.unlock()
 
580
                raise
 
581
        except:
 
582
            self.branch.unlock()
 
583
            raise
 
584
 
 
585
    def _lock_self_write(self):
 
586
        """This should be called after the branch is locked."""
 
587
        try:
 
588
            self._control_files.lock_write()
 
589
            try:
 
590
                state = self.current_dirstate()
 
591
                if not state._lock_token:
 
592
                    state.lock_write()
 
593
                # set our support for tree references from the repository in
 
594
                # use.
 
595
                self._repo_supports_tree_reference = getattr(
 
596
                    self.branch.repository._format, "supports_tree_reference",
 
597
                    False)
 
598
            except:
 
599
                self._control_files.unlock()
 
600
                raise
 
601
        except:
 
602
            self.branch.unlock()
 
603
            raise
 
604
 
 
605
    def lock_tree_write(self):
 
606
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
607
        self.branch.lock_read()
 
608
        self._lock_self_write()
 
609
 
 
610
    def lock_write(self):
 
611
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
612
        self.branch.lock_write()
 
613
        self._lock_self_write()
 
614
 
 
615
    @needs_tree_write_lock
 
616
    def move(self, from_paths, to_dir, after=False):
 
617
        """See WorkingTree.move()."""
 
618
        result = []
 
619
        if not from_paths:
 
620
            return result
 
621
 
 
622
        state = self.current_dirstate()
 
623
 
 
624
        assert not isinstance(from_paths, basestring)
 
625
        to_dir_utf8 = to_dir.encode('utf8')
 
626
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
627
        id_index = state._get_id_index()
 
628
        # check destination directory
 
629
        # get the details for it
 
630
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
631
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
632
        if not entry_present:
 
633
            raise errors.BzrMoveFailedError('', to_dir,
 
634
                errors.NotVersionedError(to_dir))
 
635
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
636
        # get a handle on the block itself.
 
637
        to_block_index = state._ensure_block(
 
638
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
639
        to_block = state._dirblocks[to_block_index]
 
640
        to_abs = self.abspath(to_dir)
 
641
        if not isdir(to_abs):
 
642
            raise errors.BzrMoveFailedError('',to_dir,
 
643
                errors.NotADirectory(to_abs))
 
644
 
 
645
        if to_entry[1][0][0] != 'd':
 
646
            raise errors.BzrMoveFailedError('',to_dir,
 
647
                errors.NotADirectory(to_abs))
 
648
 
 
649
        if self._inventory is not None:
 
650
            update_inventory = True
 
651
            inv = self.inventory
 
652
            to_dir_ie = inv[to_dir_id]
 
653
            to_dir_id = to_entry[0][2]
 
654
        else:
 
655
            update_inventory = False
 
656
 
 
657
        rollbacks = []
 
658
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
659
                     fingerprint, packed_stat, size,
 
660
                     to_block, to_key, to_path_utf8):
 
661
            state._make_absent(old_entry)
 
662
            from_key = old_entry[0]
 
663
            rollbacks.append(
 
664
                lambda:state.update_minimal(from_key,
 
665
                    minikind,
 
666
                    executable=executable,
 
667
                    fingerprint=fingerprint,
 
668
                    packed_stat=packed_stat,
 
669
                    size=size,
 
670
                    path_utf8=from_path_utf8))
 
671
            state.update_minimal(to_key,
 
672
                    minikind,
 
673
                    executable=executable,
 
674
                    fingerprint=fingerprint,
 
675
                    packed_stat=packed_stat,
 
676
                    size=size,
 
677
                    path_utf8=to_path_utf8)
 
678
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
679
            new_entry = to_block[1][added_entry_index]
 
680
            rollbacks.append(lambda:state._make_absent(new_entry))
 
681
 
 
682
        for from_rel in from_paths:
 
683
            # from_rel is 'pathinroot/foo/bar'
 
684
            from_rel_utf8 = from_rel.encode('utf8')
 
685
            from_dirname, from_tail = osutils.split(from_rel)
 
686
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
687
            from_entry = self._get_entry(path=from_rel)
 
688
            if from_entry == (None, None):
 
689
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
690
                    errors.NotVersionedError(path=str(from_rel)))
 
691
 
 
692
            from_id = from_entry[0][2]
 
693
            to_rel = pathjoin(to_dir, from_tail)
 
694
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
695
            item_to_entry = self._get_entry(path=to_rel)
 
696
            if item_to_entry != (None, None):
 
697
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
698
                    "Target is already versioned.")
 
699
 
 
700
            if from_rel == to_rel:
 
701
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
702
                    "Source and target are identical.")
 
703
 
 
704
            from_missing = not self.has_filename(from_rel)
 
705
            to_missing = not self.has_filename(to_rel)
 
706
            if after:
 
707
                move_file = False
 
708
            else:
 
709
                move_file = True
 
710
            if to_missing:
 
711
                if not move_file:
 
712
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
713
                        errors.NoSuchFile(path=to_rel,
 
714
                        extra="New file has not been created yet"))
 
715
                elif from_missing:
 
716
                    # neither path exists
 
717
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
718
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
719
            else:
 
720
                if from_missing: # implicitly just update our path mapping
 
721
                    move_file = False
 
722
                elif not after:
 
723
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
724
                        extra="(Use --after to update the Bazaar id)")
 
725
 
 
726
            rollbacks = []
 
727
            def rollback_rename():
 
728
                """A single rename has failed, roll it back."""
 
729
                # roll back everything, even if we encounter trouble doing one
 
730
                # of them.
 
731
                #
 
732
                # TODO: at least log the other exceptions rather than just
 
733
                # losing them mbp 20070307
 
734
                exc_info = None
 
735
                for rollback in reversed(rollbacks):
 
736
                    try:
 
737
                        rollback()
 
738
                    except Exception, e:
 
739
                        exc_info = sys.exc_info()
 
740
                if exc_info:
 
741
                    raise exc_info[0], exc_info[1], exc_info[2]
 
742
 
 
743
            # perform the disk move first - its the most likely failure point.
 
744
            if move_file:
 
745
                from_rel_abs = self.abspath(from_rel)
 
746
                to_rel_abs = self.abspath(to_rel)
 
747
                try:
 
748
                    osutils.rename(from_rel_abs, to_rel_abs)
 
749
                except OSError, e:
 
750
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
751
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
752
            try:
 
753
                # perform the rename in the inventory next if needed: its easy
 
754
                # to rollback
 
755
                if update_inventory:
 
756
                    # rename the entry
 
757
                    from_entry = inv[from_id]
 
758
                    current_parent = from_entry.parent_id
 
759
                    inv.rename(from_id, to_dir_id, from_tail)
 
760
                    rollbacks.append(
 
761
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
762
                # finally do the rename in the dirstate, which is a little
 
763
                # tricky to rollback, but least likely to need it.
 
764
                old_block_index, old_entry_index, dir_present, file_present = \
 
765
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
766
                old_block = state._dirblocks[old_block_index][1]
 
767
                old_entry = old_block[old_entry_index]
 
768
                from_key, old_entry_details = old_entry
 
769
                cur_details = old_entry_details[0]
 
770
                # remove the old row
 
771
                to_key = ((to_block[0],) + from_key[1:3])
 
772
                minikind = cur_details[0]
 
773
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
774
                         minikind=minikind,
 
775
                         executable=cur_details[3],
 
776
                         fingerprint=cur_details[1],
 
777
                         packed_stat=cur_details[4],
 
778
                         size=cur_details[2],
 
779
                         to_block=to_block,
 
780
                         to_key=to_key,
 
781
                         to_path_utf8=to_rel_utf8)
 
782
 
 
783
                if minikind == 'd':
 
784
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
785
                        """Recursively update all entries in this dirblock."""
 
786
                        assert from_dir != '', "renaming root not supported"
 
787
                        from_key = (from_dir, '')
 
788
                        from_block_idx, present = \
 
789
                            state._find_block_index_from_key(from_key)
 
790
                        if not present:
 
791
                            # This is the old record, if it isn't present, then
 
792
                            # there is theoretically nothing to update.
 
793
                            # (Unless it isn't present because of lazy loading,
 
794
                            # but we don't do that yet)
 
795
                            return
 
796
                        from_block = state._dirblocks[from_block_idx]
 
797
                        to_block_index, to_entry_index, _, _ = \
 
798
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
799
                        to_block_index = state._ensure_block(
 
800
                            to_block_index, to_entry_index, to_dir_utf8)
 
801
                        to_block = state._dirblocks[to_block_index]
 
802
 
 
803
                        # Grab a copy since move_one may update the list.
 
804
                        for entry in from_block[1][:]:
 
805
                            assert entry[0][0] == from_dir
 
806
                            cur_details = entry[1][0]
 
807
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
808
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
809
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
810
                            minikind = cur_details[0]
 
811
                            if minikind in 'ar':
 
812
                                # Deleted children of a renamed directory
 
813
                                # Do not need to be updated.
 
814
                                # Children that have been renamed out of this
 
815
                                # directory should also not be updated
 
816
                                continue
 
817
                            move_one(entry, from_path_utf8=from_path_utf8,
 
818
                                     minikind=minikind,
 
819
                                     executable=cur_details[3],
 
820
                                     fingerprint=cur_details[1],
 
821
                                     packed_stat=cur_details[4],
 
822
                                     size=cur_details[2],
 
823
                                     to_block=to_block,
 
824
                                     to_key=to_key,
 
825
                                     to_path_utf8=to_path_utf8)
 
826
                            if minikind == 'd':
 
827
                                # We need to move all the children of this
 
828
                                # entry
 
829
                                update_dirblock(from_path_utf8, to_key,
 
830
                                                to_path_utf8)
 
831
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
832
            except:
 
833
                rollback_rename()
 
834
                raise
 
835
            result.append((from_rel, to_rel))
 
836
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
837
            self._make_dirty(reset_inventory=False)
 
838
 
 
839
        return result
 
840
 
 
841
    def _must_be_locked(self):
 
842
        if not self._control_files._lock_count:
 
843
            raise errors.ObjectNotLocked(self)
 
844
 
 
845
    def _new_tree(self):
 
846
        """Initialize the state in this tree to be a new tree."""
 
847
        self._dirty = True
 
848
 
 
849
    @needs_read_lock
 
850
    def path2id(self, path):
 
851
        """Return the id for path in this tree."""
 
852
        path = path.strip('/')
 
853
        entry = self._get_entry(path=path)
 
854
        if entry == (None, None):
 
855
            return None
 
856
        return entry[0][2]
 
857
 
 
858
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
859
        """See Tree.paths2ids().
 
860
 
 
861
        This specialisation fast-paths the case where all the trees are in the
 
862
        dirstate.
 
863
        """
 
864
        if paths is None:
 
865
            return None
 
866
        parents = self.get_parent_ids()
 
867
        for tree in trees:
 
868
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
869
                parents):
 
870
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
871
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
872
        # -- make all paths utf8 --
 
873
        paths_utf8 = set()
 
874
        for path in paths:
 
875
            paths_utf8.add(path.encode('utf8'))
 
876
        paths = paths_utf8
 
877
        # -- paths is now a utf8 path set --
 
878
        # -- get the state object and prepare it.
 
879
        state = self.current_dirstate()
 
880
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
881
            and '' not in paths):
 
882
            paths2ids = self._paths2ids_using_bisect
 
883
        else:
 
884
            paths2ids = self._paths2ids_in_memory
 
885
        return paths2ids(paths, search_indexes,
 
886
                         require_versioned=require_versioned)
 
887
 
 
888
    def _paths2ids_in_memory(self, paths, search_indexes,
 
889
                             require_versioned=True):
 
890
        state = self.current_dirstate()
 
891
        state._read_dirblocks_if_needed()
 
892
        def _entries_for_path(path):
 
893
            """Return a list with all the entries that match path for all ids.
 
894
            """
 
895
            dirname, basename = os.path.split(path)
 
896
            key = (dirname, basename, '')
 
897
            block_index, present = state._find_block_index_from_key(key)
 
898
            if not present:
 
899
                # the block which should contain path is absent.
 
900
                return []
 
901
            result = []
 
902
            block = state._dirblocks[block_index][1]
 
903
            entry_index, _ = state._find_entry_index(key, block)
 
904
            # we may need to look at multiple entries at this path: walk while the paths match.
 
905
            while (entry_index < len(block) and
 
906
                block[entry_index][0][0:2] == key[0:2]):
 
907
                result.append(block[entry_index])
 
908
                entry_index += 1
 
909
            return result
 
910
        if require_versioned:
 
911
            # -- check all supplied paths are versioned in a search tree. --
 
912
            all_versioned = True
 
913
            for path in paths:
 
914
                path_entries = _entries_for_path(path)
 
915
                if not path_entries:
 
916
                    # this specified path is not present at all: error
 
917
                    all_versioned = False
 
918
                    break
 
919
                found_versioned = False
 
920
                # for each id at this path
 
921
                for entry in path_entries:
 
922
                    # for each tree.
 
923
                    for index in search_indexes:
 
924
                        if entry[1][index][0] != 'a': # absent
 
925
                            found_versioned = True
 
926
                            # all good: found a versioned cell
 
927
                            break
 
928
                if not found_versioned:
 
929
                    # none of the indexes was not 'absent' at all ids for this
 
930
                    # path.
 
931
                    all_versioned = False
 
932
                    break
 
933
            if not all_versioned:
 
934
                raise errors.PathsNotVersionedError(paths)
 
935
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
936
        search_paths = osutils.minimum_path_selection(paths)
 
937
        # sketch: 
 
938
        # for all search_indexs in each path at or under each element of
 
939
        # search_paths, if the detail is relocated: add the id, and add the
 
940
        # relocated path as one to search if its not searched already. If the
 
941
        # detail is not relocated, add the id.
 
942
        searched_paths = set()
 
943
        found_ids = set()
 
944
        def _process_entry(entry):
 
945
            """Look at search_indexes within entry.
 
946
 
 
947
            If a specific tree's details are relocated, add the relocation
 
948
            target to search_paths if not searched already. If it is absent, do
 
949
            nothing. Otherwise add the id to found_ids.
 
950
            """
 
951
            for index in search_indexes:
 
952
                if entry[1][index][0] == 'r': # relocated
 
953
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
954
                        search_paths.add(entry[1][index][1])
 
955
                elif entry[1][index][0] != 'a': # absent
 
956
                    found_ids.add(entry[0][2])
 
957
        while search_paths:
 
958
            current_root = search_paths.pop()
 
959
            searched_paths.add(current_root)
 
960
            # process the entries for this containing directory: the rest will be
 
961
            # found by their parents recursively.
 
962
            root_entries = _entries_for_path(current_root)
 
963
            if not root_entries:
 
964
                # this specified path is not present at all, skip it.
 
965
                continue
 
966
            for entry in root_entries:
 
967
                _process_entry(entry)
 
968
            initial_key = (current_root, '', '')
 
969
            block_index, _ = state._find_block_index_from_key(initial_key)
 
970
            while (block_index < len(state._dirblocks) and
 
971
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
972
                for entry in state._dirblocks[block_index][1]:
 
973
                    _process_entry(entry)
 
974
                block_index += 1
 
975
        return found_ids
 
976
 
 
977
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
978
                                require_versioned=True):
 
979
        state = self.current_dirstate()
 
980
        found_ids = set()
 
981
 
 
982
        split_paths = sorted(osutils.split(p) for p in paths)
 
983
        found = state._bisect_recursive(split_paths)
 
984
 
 
985
        if require_versioned:
 
986
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
987
            for dir_name in split_paths:
 
988
                if dir_name not in found_dir_names:
 
989
                    raise errors.PathsNotVersionedError(paths)
 
990
 
 
991
        for dir_name_id, trees_info in found.iteritems():
 
992
            for index in search_indexes:
 
993
                if trees_info[index][0] not in ('r', 'a'):
 
994
                    found_ids.add(dir_name_id[2])
 
995
        return found_ids
 
996
 
 
997
    def read_working_inventory(self):
 
998
        """Read the working inventory.
 
999
        
 
1000
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1001
        """
 
1002
        return self.inventory
 
1003
 
 
1004
    @needs_read_lock
 
1005
    def revision_tree(self, revision_id):
 
1006
        """See Tree.revision_tree.
 
1007
 
 
1008
        WorkingTree4 supplies revision_trees for any basis tree.
 
1009
        """
 
1010
        dirstate = self.current_dirstate()
 
1011
        parent_ids = dirstate.get_parent_ids()
 
1012
        if revision_id not in parent_ids:
 
1013
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1014
        if revision_id in dirstate.get_ghosts():
 
1015
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1016
        return DirStateRevisionTree(dirstate, revision_id,
 
1017
            self.branch.repository)
 
1018
 
 
1019
    @needs_tree_write_lock
 
1020
    def set_last_revision(self, new_revision):
 
1021
        """Change the last revision in the working tree."""
 
1022
        parents = self.get_parent_ids()
 
1023
        if new_revision in (NULL_REVISION, None):
 
1024
            assert len(parents) < 2, (
 
1025
                "setting the last parent to none with a pending merge is "
 
1026
                "unsupported.")
 
1027
            self.set_parent_ids([])
 
1028
        else:
 
1029
            self.set_parent_ids([new_revision] + parents[1:],
 
1030
                allow_leftmost_as_ghost=True)
 
1031
 
 
1032
    @needs_tree_write_lock
 
1033
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1034
        """Set the parent ids to revision_ids.
 
1035
        
 
1036
        See also set_parent_trees. This api will try to retrieve the tree data
 
1037
        for each element of revision_ids from the trees repository. If you have
 
1038
        tree data already available, it is more efficient to use
 
1039
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1040
        an easier API to use.
 
1041
 
 
1042
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1043
            working tree. Any of these may be ghosts.
 
1044
        """
 
1045
        trees = []
 
1046
        for revision_id in revision_ids:
 
1047
            try:
 
1048
                revtree = self.branch.repository.revision_tree(revision_id)
 
1049
                # TODO: jam 20070213 KnitVersionedFile raises
 
1050
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1051
                #       given revision_id is not present. Should Repository be
 
1052
                #       catching it and re-raising NoSuchRevision?
 
1053
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1054
                revtree = None
 
1055
            trees.append((revision_id, revtree))
 
1056
        self.set_parent_trees(trees,
 
1057
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1058
 
 
1059
    @needs_tree_write_lock
 
1060
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1061
        """Set the parents of the working tree.
 
1062
 
 
1063
        :param parents_list: A list of (revision_id, tree) tuples.
 
1064
            If tree is None, then that element is treated as an unreachable
 
1065
            parent tree - i.e. a ghost.
 
1066
        """
 
1067
        dirstate = self.current_dirstate()
 
1068
        if len(parents_list) > 0:
 
1069
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1070
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1071
        real_trees = []
 
1072
        ghosts = []
 
1073
        # convert absent trees to the null tree, which we convert back to
 
1074
        # missing on access.
 
1075
        for rev_id, tree in parents_list:
 
1076
            _mod_revision.check_not_reserved_id(rev_id)
 
1077
            if tree is not None:
 
1078
                real_trees.append((rev_id, tree))
 
1079
            else:
 
1080
                real_trees.append((rev_id,
 
1081
                    self.branch.repository.revision_tree(None)))
 
1082
                ghosts.append(rev_id)
 
1083
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1084
        self._make_dirty(reset_inventory=False)
 
1085
 
 
1086
    def _set_root_id(self, file_id):
 
1087
        """See WorkingTree.set_root_id."""
 
1088
        state = self.current_dirstate()
 
1089
        state.set_path_id('', file_id)
 
1090
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1091
            self._make_dirty(reset_inventory=True)
 
1092
 
 
1093
    def _sha_from_stat(self, path, stat_result):
 
1094
        """Get a sha digest from the tree's stat cache.
 
1095
 
 
1096
        The default implementation assumes no stat cache is present.
 
1097
 
 
1098
        :param path: The path.
 
1099
        :param stat_result: The stat result being looked up.
 
1100
        """
 
1101
        state = self.current_dirstate()
 
1102
        # XXX: should we make the path be passed in as utf8 ?
 
1103
        entry = state._get_entry(0, path_utf8=cache_utf8.encode(path))
 
1104
        tree_details = entry[1][0]
 
1105
        packed_stat = dirstate.pack_stat(stat_result)
 
1106
        if tree_details[4] == packed_stat:
 
1107
            return tree_details[1]
 
1108
        else:
 
1109
            return None
 
1110
 
 
1111
    @needs_read_lock
 
1112
    def supports_tree_reference(self):
 
1113
        return self._repo_supports_tree_reference
 
1114
 
 
1115
    def unlock(self):
 
1116
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1117
        # do non-implementation specific cleanup
 
1118
        self._cleanup()
 
1119
 
 
1120
        if self._control_files._lock_count == 1:
 
1121
            # eventually we should do signature checking during read locks for
 
1122
            # dirstate updates.
 
1123
            if self._control_files._lock_mode == 'w':
 
1124
                if self._dirty:
 
1125
                    self.flush()
 
1126
            if self._dirstate is not None:
 
1127
                # This is a no-op if there are no modifications.
 
1128
                self._dirstate.save()
 
1129
                self._dirstate.unlock()
 
1130
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1131
            #       point. Instead, it could check if the header has been
 
1132
            #       modified when it is locked, and if not, it can hang on to
 
1133
            #       the data it has in memory.
 
1134
            self._dirstate = None
 
1135
            self._inventory = None
 
1136
        # reverse order of locking.
 
1137
        try:
 
1138
            return self._control_files.unlock()
 
1139
        finally:
 
1140
            self.branch.unlock()
 
1141
 
 
1142
    @needs_tree_write_lock
 
1143
    def unversion(self, file_ids):
 
1144
        """Remove the file ids in file_ids from the current versioned set.
 
1145
 
 
1146
        When a file_id is unversioned, all of its children are automatically
 
1147
        unversioned.
 
1148
 
 
1149
        :param file_ids: The file ids to stop versioning.
 
1150
        :raises: NoSuchId if any fileid is not currently versioned.
 
1151
        """
 
1152
        if not file_ids:
 
1153
            return
 
1154
        state = self.current_dirstate()
 
1155
        state._read_dirblocks_if_needed()
 
1156
        ids_to_unversion = set(file_ids)
 
1157
        paths_to_unversion = set()
 
1158
        # sketch:
 
1159
        # check if the root is to be unversioned, if so, assert for now.
 
1160
        # walk the state marking unversioned things as absent.
 
1161
        # if there are any un-unversioned ids at the end, raise
 
1162
        for key, details in state._dirblocks[0][1]:
 
1163
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1164
                key[2] in ids_to_unversion):
 
1165
                # I haven't written the code to unversion / yet - it should be
 
1166
                # supported.
 
1167
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1168
        block_index = 0
 
1169
        while block_index < len(state._dirblocks):
 
1170
            # process one directory at a time.
 
1171
            block = state._dirblocks[block_index]
 
1172
            # first check: is the path one to remove - it or its children
 
1173
            delete_block = False
 
1174
            for path in paths_to_unversion:
 
1175
                if (block[0].startswith(path) and
 
1176
                    (len(block[0]) == len(path) or
 
1177
                     block[0][len(path)] == '/')):
 
1178
                    # this entire block should be deleted - its the block for a
 
1179
                    # path to unversion; or the child of one
 
1180
                    delete_block = True
 
1181
                    break
 
1182
            # TODO: trim paths_to_unversion as we pass by paths
 
1183
            if delete_block:
 
1184
                # this block is to be deleted: process it.
 
1185
                # TODO: we can special case the no-parents case and
 
1186
                # just forget the whole block.
 
1187
                entry_index = 0
 
1188
                while entry_index < len(block[1]):
 
1189
                    # Mark this file id as having been removed
 
1190
                    entry = block[1][entry_index]
 
1191
                    ids_to_unversion.discard(entry[0][2])
 
1192
                    if (entry[1][0][0] == 'a'
 
1193
                        or not state._make_absent(entry)):
 
1194
                        entry_index += 1
 
1195
                # go to the next block. (At the moment we dont delete empty
 
1196
                # dirblocks)
 
1197
                block_index += 1
 
1198
                continue
 
1199
            entry_index = 0
 
1200
            while entry_index < len(block[1]):
 
1201
                entry = block[1][entry_index]
 
1202
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1203
                    # ^ some parent row.
 
1204
                    entry[0][2] not in ids_to_unversion):
 
1205
                    # ^ not an id to unversion
 
1206
                    entry_index += 1
 
1207
                    continue
 
1208
                if entry[1][0][0] == 'd':
 
1209
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1210
                if not state._make_absent(entry):
 
1211
                    entry_index += 1
 
1212
                # we have unversioned this id
 
1213
                ids_to_unversion.remove(entry[0][2])
 
1214
            block_index += 1
 
1215
        if ids_to_unversion:
 
1216
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1217
        self._make_dirty(reset_inventory=False)
 
1218
        # have to change the legacy inventory too.
 
1219
        if self._inventory is not None:
 
1220
            for file_id in file_ids:
 
1221
                self._inventory.remove_recursive_id(file_id)
 
1222
 
 
1223
    @needs_read_lock
 
1224
    def _validate(self):
 
1225
        self._dirstate._validate()
 
1226
 
 
1227
    @needs_tree_write_lock
 
1228
    def _write_inventory(self, inv):
 
1229
        """Write inventory as the current inventory."""
 
1230
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1231
        self.current_dirstate().set_state_from_inventory(inv)
 
1232
        self._make_dirty(reset_inventory=False)
 
1233
        if self._inventory is not None:
 
1234
            self._inventory = inv
 
1235
        self.flush()
 
1236
 
 
1237
 
 
1238
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1239
    """The first consolidated dirstate working tree format.
 
1240
 
 
1241
    This format:
 
1242
        - exists within a metadir controlling .bzr
 
1243
        - includes an explicit version marker for the workingtree control
 
1244
          files, separate from the BzrDir format
 
1245
        - modifies the hash cache format
 
1246
        - is new in bzr 0.15
 
1247
        - uses a LockDir to guard access to it.
 
1248
    """
 
1249
 
 
1250
    upgrade_recommended = False
 
1251
 
 
1252
    def get_format_string(self):
 
1253
        """See WorkingTreeFormat.get_format_string()."""
 
1254
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1255
 
 
1256
    def get_format_description(self):
 
1257
        """See WorkingTreeFormat.get_format_description()."""
 
1258
        return "Working tree format 4"
 
1259
 
 
1260
    def initialize(self, a_bzrdir, revision_id=None):
 
1261
        """See WorkingTreeFormat.initialize().
 
1262
 
 
1263
        :param revision_id: allows creating a working tree at a different
 
1264
        revision than the branch is at.
 
1265
 
 
1266
        These trees get an initial random root id, if their repository supports
 
1267
        rich root data, TREE_ROOT otherwise.
 
1268
        """
 
1269
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1270
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1271
        transport = a_bzrdir.get_workingtree_transport(self)
 
1272
        control_files = self._open_control_files(a_bzrdir)
 
1273
        control_files.create_lock()
 
1274
        control_files.lock_write()
 
1275
        control_files.put_utf8('format', self.get_format_string())
 
1276
        branch = a_bzrdir.open_branch()
 
1277
        if revision_id is None:
 
1278
            revision_id = branch.last_revision()
 
1279
        local_path = transport.local_abspath('dirstate')
 
1280
        # write out new dirstate (must exist when we create the tree)
 
1281
        state = dirstate.DirState.initialize(local_path)
 
1282
        state.unlock()
 
1283
        del state
 
1284
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1285
                         branch,
 
1286
                         _format=self,
 
1287
                         _bzrdir=a_bzrdir,
 
1288
                         _control_files=control_files)
 
1289
        wt._new_tree()
 
1290
        wt.lock_tree_write()
 
1291
        try:
 
1292
            if revision_id in (None, NULL_REVISION):
 
1293
                if branch.repository.supports_rich_root():
 
1294
                    wt._set_root_id(generate_ids.gen_root_id())
 
1295
                else:
 
1296
                    wt._set_root_id(ROOT_ID)
 
1297
                wt.flush()
 
1298
            wt.set_last_revision(revision_id)
 
1299
            wt.flush()
 
1300
            basis = wt.basis_tree()
 
1301
            basis.lock_read()
 
1302
            # if the basis has a root id we have to use that; otherwise we use
 
1303
            # a new random one
 
1304
            basis_root_id = basis.get_root_id()
 
1305
            if basis_root_id is not None:
 
1306
                wt._set_root_id(basis_root_id)
 
1307
                wt.flush()
 
1308
            transform.build_tree(basis, wt)
 
1309
            basis.unlock()
 
1310
        finally:
 
1311
            control_files.unlock()
 
1312
            wt.unlock()
 
1313
        return wt
 
1314
 
 
1315
    def _open(self, a_bzrdir, control_files):
 
1316
        """Open the tree itself.
 
1317
 
 
1318
        :param a_bzrdir: the dir for the tree.
 
1319
        :param control_files: the control files for the tree.
 
1320
        """
 
1321
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1322
                           branch=a_bzrdir.open_branch(),
 
1323
                           _format=self,
 
1324
                           _bzrdir=a_bzrdir,
 
1325
                           _control_files=control_files)
 
1326
 
 
1327
    def __get_matchingbzrdir(self):
 
1328
        # please test against something that will let us do tree references
 
1329
        return bzrdir.format_registry.make_bzrdir(
 
1330
            'dirstate-with-subtree')
 
1331
 
 
1332
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1333
 
 
1334
 
 
1335
class DirStateRevisionTree(Tree):
 
1336
    """A revision tree pulling the inventory from a dirstate."""
 
1337
 
 
1338
    def __init__(self, dirstate, revision_id, repository):
 
1339
        self._dirstate = dirstate
 
1340
        self._revision_id = revision_id
 
1341
        self._repository = repository
 
1342
        self._inventory = None
 
1343
        self._locked = 0
 
1344
        self._dirstate_locked = False
 
1345
 
 
1346
    def __repr__(self):
 
1347
        return "<%s of %s in %s>" % \
 
1348
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1349
 
 
1350
    def annotate_iter(self, file_id,
 
1351
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1352
        """See Tree.annotate_iter"""
 
1353
        w = self._get_weave(file_id)
 
1354
        return w.annotate_iter(self.inventory[file_id].revision)
 
1355
 
 
1356
    def _get_ancestors(self, default_revision):
 
1357
        return set(self._repository.get_ancestry(self._revision_id,
 
1358
                                                 topo_sorted=False))
 
1359
    def _comparison_data(self, entry, path):
 
1360
        """See Tree._comparison_data."""
 
1361
        if entry is None:
 
1362
            return None, False, None
 
1363
        # trust the entry as RevisionTree does, but this may not be
 
1364
        # sensible: the entry might not have come from us?
 
1365
        return entry.kind, entry.executable, None
 
1366
 
 
1367
    def _file_size(self, entry, stat_value):
 
1368
        return entry.text_size
 
1369
 
 
1370
    def filter_unversioned_files(self, paths):
 
1371
        """Filter out paths that are not versioned.
 
1372
 
 
1373
        :return: set of paths.
 
1374
        """
 
1375
        pred = self.has_filename
 
1376
        return set((p for p in paths if not pred(p)))
 
1377
 
 
1378
    def get_root_id(self):
 
1379
        return self.path2id('')
 
1380
 
 
1381
    def _get_parent_index(self):
 
1382
        """Return the index in the dirstate referenced by this tree."""
 
1383
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1384
 
 
1385
    def _get_entry(self, file_id=None, path=None):
 
1386
        """Get the dirstate row for file_id or path.
 
1387
 
 
1388
        If either file_id or path is supplied, it is used as the key to lookup.
 
1389
        If both are supplied, the fastest lookup is used, and an error is
 
1390
        raised if they do not both point at the same row.
 
1391
        
 
1392
        :param file_id: An optional unicode file_id to be looked up.
 
1393
        :param path: An optional unicode path to be looked up.
 
1394
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1395
        """
 
1396
        if file_id is None and path is None:
 
1397
            raise errors.BzrError('must supply file_id or path')
 
1398
        if path is not None:
 
1399
            path = path.encode('utf8')
 
1400
        parent_index = self._get_parent_index()
 
1401
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1402
 
 
1403
    def _generate_inventory(self):
 
1404
        """Create and set self.inventory from the dirstate object.
 
1405
 
 
1406
        (So this is only called the first time the inventory is requested for
 
1407
        this tree; it then remains in memory until it's out of date.)
 
1408
 
 
1409
        This is relatively expensive: we have to walk the entire dirstate.
 
1410
        """
 
1411
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1412
            'dirstate revision tree'
 
1413
        # separate call for profiling - makes it clear where the costs are.
 
1414
        self._dirstate._read_dirblocks_if_needed()
 
1415
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1416
            'parent %s has disappeared from %s' % (
 
1417
            self._revision_id, self._dirstate.get_parent_ids())
 
1418
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1419
        # This is identical now to the WorkingTree _generate_inventory except
 
1420
        # for the tree index use.
 
1421
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1422
        current_id = root_key[2]
 
1423
        assert current_entry[parent_index][0] == 'd'
 
1424
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1425
        inv.root.revision = current_entry[parent_index][4]
 
1426
        # Turn some things into local variables
 
1427
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1428
        factory = entry_factory
 
1429
        utf8_decode = cache_utf8._utf8_decode
 
1430
        inv_byid = inv._byid
 
1431
        # we could do this straight out of the dirstate; it might be fast
 
1432
        # and should be profiled - RBC 20070216
 
1433
        parent_ies = {'' : inv.root}
 
1434
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1435
            dirname = block[0]
 
1436
            try:
 
1437
                parent_ie = parent_ies[dirname]
 
1438
            except KeyError:
 
1439
                # all the paths in this block are not versioned in this tree
 
1440
                continue
 
1441
            for key, entry in block[1]:
 
1442
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1443
                if minikind in ('a', 'r'): # absent, relocated
 
1444
                    # not this tree
 
1445
                    continue
 
1446
                name = key[1]
 
1447
                name_unicode = utf8_decode(name)[0]
 
1448
                file_id = key[2]
 
1449
                kind = minikind_to_kind[minikind]
 
1450
                inv_entry = factory[kind](file_id, name_unicode,
 
1451
                                          parent_ie.file_id)
 
1452
                inv_entry.revision = revid
 
1453
                if kind == 'file':
 
1454
                    inv_entry.executable = executable
 
1455
                    inv_entry.text_size = size
 
1456
                    inv_entry.text_sha1 = fingerprint
 
1457
                elif kind == 'directory':
 
1458
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1459
                elif kind == 'symlink':
 
1460
                    inv_entry.executable = False
 
1461
                    inv_entry.text_size = None
 
1462
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1463
                elif kind == 'tree-reference':
 
1464
                    inv_entry.reference_revision = fingerprint or None
 
1465
                else:
 
1466
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1467
                            % entry)
 
1468
                # These checks cost us around 40ms on a 55k entry tree
 
1469
                assert file_id not in inv_byid
 
1470
                assert name_unicode not in parent_ie.children
 
1471
                inv_byid[file_id] = inv_entry
 
1472
                parent_ie.children[name_unicode] = inv_entry
 
1473
        self._inventory = inv
 
1474
 
 
1475
    def get_file_mtime(self, file_id, path=None):
 
1476
        """Return the modification time for this record.
 
1477
 
 
1478
        We return the timestamp of the last-changed revision.
 
1479
        """
 
1480
        # Make sure the file exists
 
1481
        entry = self._get_entry(file_id, path=path)
 
1482
        if entry == (None, None): # do we raise?
 
1483
            return None
 
1484
        parent_index = self._get_parent_index()
 
1485
        last_changed_revision = entry[1][parent_index][4]
 
1486
        return self._repository.get_revision(last_changed_revision).timestamp
 
1487
 
 
1488
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1489
        entry = self._get_entry(file_id=file_id, path=path)
 
1490
        parent_index = self._get_parent_index()
 
1491
        parent_details = entry[1][parent_index]
 
1492
        if parent_details[0] == 'f':
 
1493
            return parent_details[1]
 
1494
        return None
 
1495
 
 
1496
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1497
    def get_weave(self, file_id):
 
1498
        return self._get_weave(file_id)
 
1499
 
 
1500
    def _get_weave(self, file_id):
 
1501
        return self._repository.weave_store.get_weave(file_id,
 
1502
                self._repository.get_transaction())
 
1503
 
 
1504
    def get_file(self, file_id, path=None):
 
1505
        return StringIO(self.get_file_text(file_id))
 
1506
 
 
1507
    def get_file_lines(self, file_id):
 
1508
        ie = self.inventory[file_id]
 
1509
        return self._get_weave(file_id).get_lines(ie.revision)
 
1510
 
 
1511
    def get_file_size(self, file_id):
 
1512
        return self.inventory[file_id].text_size
 
1513
 
 
1514
    def get_file_text(self, file_id):
 
1515
        return ''.join(self.get_file_lines(file_id))
 
1516
 
 
1517
    def get_reference_revision(self, file_id, path=None):
 
1518
        return self.inventory[file_id].reference_revision
 
1519
 
 
1520
    def iter_files_bytes(self, desired_files):
 
1521
        """See Tree.iter_files_bytes.
 
1522
 
 
1523
        This version is implemented on top of Repository.iter_files_bytes"""
 
1524
        parent_index = self._get_parent_index()
 
1525
        repo_desired_files = []
 
1526
        for file_id, identifier in desired_files:
 
1527
            entry = self._get_entry(file_id)
 
1528
            if entry == (None, None):
 
1529
                raise errors.NoSuchId(self, file_id)
 
1530
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1531
                                       identifier))
 
1532
        return self._repository.iter_files_bytes(repo_desired_files)
 
1533
 
 
1534
    def get_symlink_target(self, file_id):
 
1535
        entry = self._get_entry(file_id=file_id)
 
1536
        parent_index = self._get_parent_index()
 
1537
        if entry[1][parent_index][0] != 'l':
 
1538
            return None
 
1539
        else:
 
1540
            # At present, none of the tree implementations supports non-ascii
 
1541
            # symlink targets. So we will just assume that the dirstate path is
 
1542
            # correct.
 
1543
            return entry[1][parent_index][1]
 
1544
 
 
1545
    def get_revision_id(self):
 
1546
        """Return the revision id for this tree."""
 
1547
        return self._revision_id
 
1548
 
 
1549
    def _get_inventory(self):
 
1550
        if self._inventory is not None:
 
1551
            return self._inventory
 
1552
        self._must_be_locked()
 
1553
        self._generate_inventory()
 
1554
        return self._inventory
 
1555
 
 
1556
    inventory = property(_get_inventory,
 
1557
                         doc="Inventory of this Tree")
 
1558
 
 
1559
    def get_parent_ids(self):
 
1560
        """The parents of a tree in the dirstate are not cached."""
 
1561
        return self._repository.get_revision(self._revision_id).parent_ids
 
1562
 
 
1563
    def has_filename(self, filename):
 
1564
        return bool(self.path2id(filename))
 
1565
 
 
1566
    def kind(self, file_id):
 
1567
        return self.inventory[file_id].kind
 
1568
 
 
1569
    def path_content_summary(self, path):
 
1570
        """See Tree.path_content_summary."""
 
1571
        id = self.inventory.path2id(path)
 
1572
        if id is None:
 
1573
            return ('missing', None, None, None)
 
1574
        entry = self._inventory[id]
 
1575
        kind = entry.kind
 
1576
        if kind == 'file':
 
1577
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1578
        elif kind == 'symlink':
 
1579
            return (kind, None, None, entry.symlink_target)
 
1580
        else:
 
1581
            return (kind, None, None, None)
 
1582
 
 
1583
    def is_executable(self, file_id, path=None):
 
1584
        ie = self.inventory[file_id]
 
1585
        if ie.kind != "file":
 
1586
            return None
 
1587
        return ie.executable
 
1588
 
 
1589
    def list_files(self, include_root=False):
 
1590
        # We use a standard implementation, because DirStateRevisionTree is
 
1591
        # dealing with one of the parents of the current state
 
1592
        inv = self._get_inventory()
 
1593
        entries = inv.iter_entries()
 
1594
        if self.inventory.root is not None and not include_root:
 
1595
            entries.next()
 
1596
        for path, entry in entries:
 
1597
            yield path, 'V', entry.kind, entry.file_id, entry
 
1598
 
 
1599
    def lock_read(self):
 
1600
        """Lock the tree for a set of operations."""
 
1601
        if not self._locked:
 
1602
            self._repository.lock_read()
 
1603
            if self._dirstate._lock_token is None:
 
1604
                self._dirstate.lock_read()
 
1605
                self._dirstate_locked = True
 
1606
        self._locked += 1
 
1607
 
 
1608
    def _must_be_locked(self):
 
1609
        if not self._locked:
 
1610
            raise errors.ObjectNotLocked(self)
 
1611
 
 
1612
    @needs_read_lock
 
1613
    def path2id(self, path):
 
1614
        """Return the id for path in this tree."""
 
1615
        # lookup by path: faster than splitting and walking the ivnentory.
 
1616
        entry = self._get_entry(path=path)
 
1617
        if entry == (None, None):
 
1618
            return None
 
1619
        return entry[0][2]
 
1620
 
 
1621
    def unlock(self):
 
1622
        """Unlock, freeing any cache memory used during the lock."""
 
1623
        # outside of a lock, the inventory is suspect: release it.
 
1624
        self._locked -=1
 
1625
        if not self._locked:
 
1626
            self._inventory = None
 
1627
            self._locked = 0
 
1628
            if self._dirstate_locked:
 
1629
                self._dirstate.unlock()
 
1630
                self._dirstate_locked = False
 
1631
            self._repository.unlock()
 
1632
 
 
1633
    def walkdirs(self, prefix=""):
 
1634
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1635
        # implementation based on an inventory.  
 
1636
        # This should be cleaned up to use the much faster Dirstate code
 
1637
        # So for now, we just build up the parent inventory, and extract
 
1638
        # it the same way RevisionTree does.
 
1639
        _directory = 'directory'
 
1640
        inv = self._get_inventory()
 
1641
        top_id = inv.path2id(prefix)
 
1642
        if top_id is None:
 
1643
            pending = []
 
1644
        else:
 
1645
            pending = [(prefix, top_id)]
 
1646
        while pending:
 
1647
            dirblock = []
 
1648
            relpath, file_id = pending.pop()
 
1649
            # 0 - relpath, 1- file-id
 
1650
            if relpath:
 
1651
                relroot = relpath + '/'
 
1652
            else:
 
1653
                relroot = ""
 
1654
            # FIXME: stash the node in pending
 
1655
            entry = inv[file_id]
 
1656
            for name, child in entry.sorted_children():
 
1657
                toppath = relroot + name
 
1658
                dirblock.append((toppath, name, child.kind, None,
 
1659
                    child.file_id, child.kind
 
1660
                    ))
 
1661
            yield (relpath, entry.file_id), dirblock
 
1662
            # push the user specified dirs from dirblock
 
1663
            for dir in reversed(dirblock):
 
1664
                if dir[2] == _directory:
 
1665
                    pending.append((dir[0], dir[4]))
 
1666
 
 
1667
 
 
1668
class InterDirStateTree(InterTree):
 
1669
    """Fast path optimiser for changes_from with dirstate trees.
 
1670
    
 
1671
    This is used only when both trees are in the dirstate working file, and 
 
1672
    the source is any parent within the dirstate, and the destination is 
 
1673
    the current working tree of the same dirstate.
 
1674
    """
 
1675
    # this could be generalized to allow comparisons between any trees in the
 
1676
    # dirstate, and possibly between trees stored in different dirstates.
 
1677
 
 
1678
    def __init__(self, source, target):
 
1679
        super(InterDirStateTree, self).__init__(source, target)
 
1680
        if not InterDirStateTree.is_compatible(source, target):
 
1681
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1682
 
 
1683
    @staticmethod
 
1684
    def make_source_parent_tree(source, target):
 
1685
        """Change the source tree into a parent of the target."""
 
1686
        revid = source.commit('record tree')
 
1687
        target.branch.repository.fetch(source.branch.repository, revid)
 
1688
        target.set_parent_ids([revid])
 
1689
        return target.basis_tree(), target
 
1690
 
 
1691
    _matching_from_tree_format = WorkingTreeFormat4()
 
1692
    _matching_to_tree_format = WorkingTreeFormat4()
 
1693
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1694
 
 
1695
    def _iter_changes(self, include_unchanged=False,
 
1696
                      specific_files=None, pb=None, extra_trees=[],
 
1697
                      require_versioned=True, want_unversioned=False):
 
1698
        """Return the changes from source to target.
 
1699
 
 
1700
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1701
            for details.
 
1702
        :param specific_files: An optional list of file paths to restrict the
 
1703
            comparison to. When mapping filenames to ids, all matches in all
 
1704
            trees (including optional extra_trees) are used, and all children of
 
1705
            matched directories are included.
 
1706
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1707
            unchanged entries in the result.
 
1708
        :param extra_trees: An optional list of additional trees to use when
 
1709
            mapping the contents of specific_files (paths) to file_ids.
 
1710
        :param require_versioned: If True, all files in specific_files must be
 
1711
            versioned in one of source, target, extra_trees or
 
1712
            PathsNotVersionedError is raised.
 
1713
        :param want_unversioned: Should unversioned files be returned in the
 
1714
            output. An unversioned file is defined as one with (False, False)
 
1715
            for the versioned pair.
 
1716
        """
 
1717
        utf8_decode = cache_utf8._utf8_decode
 
1718
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1719
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1720
        # NB: show_status depends on being able to pass in non-versioned files
 
1721
        # and report them as unknown
 
1722
        # TODO: handle extra trees in the dirstate.
 
1723
        # TODO: handle comparisons as an empty tree as a different special
 
1724
        # case? mbp 20070226
 
1725
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1726
            or specific_files == []):
 
1727
            # we can't fast-path these cases (yet)
 
1728
            for f in super(InterDirStateTree, self)._iter_changes(
 
1729
                include_unchanged, specific_files, pb, extra_trees,
 
1730
                require_versioned, want_unversioned=want_unversioned):
 
1731
                yield f
 
1732
            return
 
1733
        parent_ids = self.target.get_parent_ids()
 
1734
        assert (self.source._revision_id in parent_ids), \
 
1735
                "revision {%s} is not stored in {%s}, but %s " \
 
1736
                "can only be used for trees stored in the dirstate" \
 
1737
                % (self.source._revision_id, self.target, self._iter_changes)
 
1738
        target_index = 0
 
1739
        if self.source._revision_id == NULL_REVISION:
 
1740
            source_index = None
 
1741
            indices = (target_index,)
 
1742
        else:
 
1743
            assert (self.source._revision_id in parent_ids), \
 
1744
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1745
                self.source._revision_id, parent_ids)
 
1746
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1747
            indices = (source_index,target_index)
 
1748
        # -- make all specific_files utf8 --
 
1749
        if specific_files:
 
1750
            specific_files_utf8 = set()
 
1751
            for path in specific_files:
 
1752
                specific_files_utf8.add(path.encode('utf8'))
 
1753
            specific_files = specific_files_utf8
 
1754
        else:
 
1755
            specific_files = set([''])
 
1756
        # -- specific_files is now a utf8 path set --
 
1757
        # -- get the state object and prepare it.
 
1758
        state = self.target.current_dirstate()
 
1759
        state._read_dirblocks_if_needed()
 
1760
        def _entries_for_path(path):
 
1761
            """Return a list with all the entries that match path for all ids.
 
1762
            """
 
1763
            dirname, basename = os.path.split(path)
 
1764
            key = (dirname, basename, '')
 
1765
            block_index, present = state._find_block_index_from_key(key)
 
1766
            if not present:
 
1767
                # the block which should contain path is absent.
 
1768
                return []
 
1769
            result = []
 
1770
            block = state._dirblocks[block_index][1]
 
1771
            entry_index, _ = state._find_entry_index(key, block)
 
1772
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1773
            while (entry_index < len(block) and
 
1774
                block[entry_index][0][0:2] == key[0:2]):
 
1775
                result.append(block[entry_index])
 
1776
                entry_index += 1
 
1777
            return result
 
1778
        if require_versioned:
 
1779
            # -- check all supplied paths are versioned in a search tree. --
 
1780
            all_versioned = True
 
1781
            for path in specific_files:
 
1782
                path_entries = _entries_for_path(path)
 
1783
                if not path_entries:
 
1784
                    # this specified path is not present at all: error
 
1785
                    all_versioned = False
 
1786
                    break
 
1787
                found_versioned = False
 
1788
                # for each id at this path
 
1789
                for entry in path_entries:
 
1790
                    # for each tree.
 
1791
                    for index in indices:
 
1792
                        if entry[1][index][0] != 'a': # absent
 
1793
                            found_versioned = True
 
1794
                            # all good: found a versioned cell
 
1795
                            break
 
1796
                if not found_versioned:
 
1797
                    # none of the indexes was not 'absent' at all ids for this
 
1798
                    # path.
 
1799
                    all_versioned = False
 
1800
                    break
 
1801
            if not all_versioned:
 
1802
                raise errors.PathsNotVersionedError(specific_files)
 
1803
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1804
        search_specific_files = set()
 
1805
        for path in specific_files:
 
1806
            other_specific_files = specific_files.difference(set([path]))
 
1807
            if not osutils.is_inside_any(other_specific_files, path):
 
1808
                # this is a top level path, we must check it.
 
1809
                search_specific_files.add(path)
 
1810
        # sketch: 
 
1811
        # compare source_index and target_index at or under each element of search_specific_files.
 
1812
        # follow the following comparison table. Note that we only want to do diff operations when
 
1813
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1814
        # for the target.
 
1815
        # cases:
 
1816
        # 
 
1817
        # Source | Target | disk | action
 
1818
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1819
        #        |        |      | diff check on source-target
 
1820
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1821
        #        |        |      | ???
 
1822
        #   r    |  a     |      | add source to search
 
1823
        #   r    |  a     |  a   | 
 
1824
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1825
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1826
        #   a    | fdlt   |      | add new id
 
1827
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1828
        #   a    |  a     |      | not present in either tree, skip
 
1829
        #   a    |  a     |  a   | not present in any tree, skip
 
1830
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1831
        #        |        |      | may not be selected by the users list of paths.
 
1832
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1833
        #        |        |      | may not be selected by the users list of paths.
 
1834
        #  fdlt  | fdlt   |      | content in both: diff them
 
1835
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1836
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1837
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1838
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1839
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1840
        #        |        |      | this id at the other path.
 
1841
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1842
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1843
        #        |        |      | this id at the other path.
 
1844
 
 
1845
        # for all search_indexs in each path at or under each element of
 
1846
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1847
        # relocated path as one to search if its not searched already. If the
 
1848
        # detail is not relocated, add the id.
 
1849
        searched_specific_files = set()
 
1850
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1851
        # Using a list so that we can access the values and change them in
 
1852
        # nested scope. Each one is [path, file_id, entry]
 
1853
        last_source_parent = [None, None]
 
1854
        last_target_parent = [None, None]
 
1855
 
 
1856
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1857
 
 
1858
        # Just a sentry, so that _process_entry can say that this
 
1859
        # record is handled, but isn't interesting to process (unchanged)
 
1860
        uninteresting = object()
 
1861
 
 
1862
 
 
1863
        old_dirname_to_file_id = {}
 
1864
        new_dirname_to_file_id = {}
 
1865
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1866
        #       keeping a cache of directories that we have seen.
 
1867
 
 
1868
        def _process_entry(entry, path_info):
 
1869
            """Compare an entry and real disk to generate delta information.
 
1870
 
 
1871
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1872
                the path of entry. If None, then the path is considered absent.
 
1873
                (Perhaps we should pass in a concrete entry for this ?)
 
1874
                Basename is returned as a utf8 string because we expect this
 
1875
                tuple will be ignored, and don't want to take the time to
 
1876
                decode.
 
1877
            :return: None if these don't match
 
1878
                     A tuple of information about the change, or
 
1879
                     the object 'uninteresting' if these match, but are
 
1880
                     basically identical.
 
1881
            """
 
1882
            if source_index is None:
 
1883
                source_details = NULL_PARENT_DETAILS
 
1884
            else:
 
1885
                source_details = entry[1][source_index]
 
1886
            target_details = entry[1][target_index]
 
1887
            target_minikind = target_details[0]
 
1888
            if path_info is not None and target_minikind in 'fdlt':
 
1889
                assert target_index == 0
 
1890
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1891
                                                  stat_value=path_info[3])
 
1892
                # The entry may have been modified by update_entry
 
1893
                target_details = entry[1][target_index]
 
1894
                target_minikind = target_details[0]
 
1895
            else:
 
1896
                link_or_sha1 = None
 
1897
            file_id = entry[0][2]
 
1898
            source_minikind = source_details[0]
 
1899
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1900
                # claimed content in both: diff
 
1901
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1902
                #        |        |      | diff check on source-target
 
1903
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1904
                #        |        |      | ???
 
1905
                if source_minikind in 'r':
 
1906
                    # add the source to the search path to find any children it
 
1907
                    # has.  TODO ? : only add if it is a container ?
 
1908
                    if not osutils.is_inside_any(searched_specific_files,
 
1909
                                                 source_details[1]):
 
1910
                        search_specific_files.add(source_details[1])
 
1911
                    # generate the old path; this is needed for stating later
 
1912
                    # as well.
 
1913
                    old_path = source_details[1]
 
1914
                    old_dirname, old_basename = os.path.split(old_path)
 
1915
                    path = pathjoin(entry[0][0], entry[0][1])
 
1916
                    old_entry = state._get_entry(source_index,
 
1917
                                                 path_utf8=old_path)
 
1918
                    # update the source details variable to be the real
 
1919
                    # location.
 
1920
                    source_details = old_entry[1][source_index]
 
1921
                    source_minikind = source_details[0]
 
1922
                else:
 
1923
                    old_dirname = entry[0][0]
 
1924
                    old_basename = entry[0][1]
 
1925
                    old_path = path = None
 
1926
                if path_info is None:
 
1927
                    # the file is missing on disk, show as removed.
 
1928
                    content_change = True
 
1929
                    target_kind = None
 
1930
                    target_exec = False
 
1931
                else:
 
1932
                    # source and target are both versioned and disk file is present.
 
1933
                    target_kind = path_info[2]
 
1934
                    if target_kind == 'directory':
 
1935
                        if path is None:
 
1936
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1937
                        new_dirname_to_file_id[path] = file_id
 
1938
                        if source_minikind != 'd':
 
1939
                            content_change = True
 
1940
                        else:
 
1941
                            # directories have no fingerprint
 
1942
                            content_change = False
 
1943
                        target_exec = False
 
1944
                    elif target_kind == 'file':
 
1945
                        if source_minikind != 'f':
 
1946
                            content_change = True
 
1947
                        else:
 
1948
                            # We could check the size, but we already have the
 
1949
                            # sha1 hash.
 
1950
                            content_change = (link_or_sha1 != source_details[1])
 
1951
                        # Target details is updated at update_entry time
 
1952
                        if use_filesystem_for_exec:
 
1953
                            # We don't need S_ISREG here, because we are sure
 
1954
                            # we are dealing with a file.
 
1955
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1956
                        else:
 
1957
                            target_exec = target_details[3]
 
1958
                    elif target_kind == 'symlink':
 
1959
                        if source_minikind != 'l':
 
1960
                            content_change = True
 
1961
                        else:
 
1962
                            content_change = (link_or_sha1 != source_details[1])
 
1963
                        target_exec = False
 
1964
                    elif target_kind == 'tree-reference':
 
1965
                        if source_minikind != 't':
 
1966
                            content_change = True
 
1967
                        else:
 
1968
                            content_change = False
 
1969
                        target_exec = False
 
1970
                    else:
 
1971
                        raise Exception, "unknown kind %s" % path_info[2]
 
1972
                if source_minikind == 'd':
 
1973
                    if path is None:
 
1974
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1975
                    old_dirname_to_file_id[old_path] = file_id
 
1976
                # parent id is the entry for the path in the target tree
 
1977
                if old_dirname == last_source_parent[0]:
 
1978
                    source_parent_id = last_source_parent[1]
 
1979
                else:
 
1980
                    try:
 
1981
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1982
                    except KeyError:
 
1983
                        source_parent_entry = state._get_entry(source_index,
 
1984
                                                               path_utf8=old_dirname)
 
1985
                        source_parent_id = source_parent_entry[0][2]
 
1986
                    if source_parent_id == entry[0][2]:
 
1987
                        # This is the root, so the parent is None
 
1988
                        source_parent_id = None
 
1989
                    else:
 
1990
                        last_source_parent[0] = old_dirname
 
1991
                        last_source_parent[1] = source_parent_id
 
1992
                new_dirname = entry[0][0]
 
1993
                if new_dirname == last_target_parent[0]:
 
1994
                    target_parent_id = last_target_parent[1]
 
1995
                else:
 
1996
                    try:
 
1997
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1998
                    except KeyError:
 
1999
                        # TODO: We don't always need to do the lookup, because the
 
2000
                        #       parent entry will be the same as the source entry.
 
2001
                        target_parent_entry = state._get_entry(target_index,
 
2002
                                                               path_utf8=new_dirname)
 
2003
                        assert target_parent_entry != (None, None), (
 
2004
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2005
                            % (new_dirname, entry))
 
2006
                        target_parent_id = target_parent_entry[0][2]
 
2007
                    if target_parent_id == entry[0][2]:
 
2008
                        # This is the root, so the parent is None
 
2009
                        target_parent_id = None
 
2010
                    else:
 
2011
                        last_target_parent[0] = new_dirname
 
2012
                        last_target_parent[1] = target_parent_id
 
2013
 
 
2014
                source_exec = source_details[3]
 
2015
                if (include_unchanged
 
2016
                    or content_change
 
2017
                    or source_parent_id != target_parent_id
 
2018
                    or old_basename != entry[0][1]
 
2019
                    or source_exec != target_exec
 
2020
                    ):
 
2021
                    if old_path is None:
 
2022
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2023
                        old_path_u = utf8_decode(old_path)[0]
 
2024
                        path_u = old_path_u
 
2025
                    else:
 
2026
                        old_path_u = utf8_decode(old_path)[0]
 
2027
                        if old_path == path:
 
2028
                            path_u = old_path_u
 
2029
                        else:
 
2030
                            path_u = utf8_decode(path)[0]
 
2031
                    source_kind = _minikind_to_kind[source_minikind]
 
2032
                    return (entry[0][2],
 
2033
                           (old_path_u, path_u),
 
2034
                           content_change,
 
2035
                           (True, True),
 
2036
                           (source_parent_id, target_parent_id),
 
2037
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2038
                           (source_kind, target_kind),
 
2039
                           (source_exec, target_exec))
 
2040
                else:
 
2041
                    return uninteresting
 
2042
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2043
                # looks like a new file
 
2044
                if path_info is not None:
 
2045
                    path = pathjoin(entry[0][0], entry[0][1])
 
2046
                    # parent id is the entry for the path in the target tree
 
2047
                    # TODO: these are the same for an entire directory: cache em.
 
2048
                    parent_id = state._get_entry(target_index,
 
2049
                                                 path_utf8=entry[0][0])[0][2]
 
2050
                    if parent_id == entry[0][2]:
 
2051
                        parent_id = None
 
2052
                    if use_filesystem_for_exec:
 
2053
                        # We need S_ISREG here, because we aren't sure if this
 
2054
                        # is a file or not.
 
2055
                        target_exec = bool(
 
2056
                            stat.S_ISREG(path_info[3].st_mode)
 
2057
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2058
                    else:
 
2059
                        target_exec = target_details[3]
 
2060
                    return (entry[0][2],
 
2061
                           (None, utf8_decode(path)[0]),
 
2062
                           True,
 
2063
                           (False, True),
 
2064
                           (None, parent_id),
 
2065
                           (None, utf8_decode(entry[0][1])[0]),
 
2066
                           (None, path_info[2]),
 
2067
                           (None, target_exec))
 
2068
                else:
 
2069
                    # but its not on disk: we deliberately treat this as just
 
2070
                    # never-present. (Why ?! - RBC 20070224)
 
2071
                    pass
 
2072
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2073
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2074
                # if its still on disk, *and* theres no other entry at this
 
2075
                # path [we dont know this in this routine at the moment -
 
2076
                # perhaps we should change this - then it would be an unknown.
 
2077
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2078
                # parent id is the entry for the path in the target tree
 
2079
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2080
                if parent_id == entry[0][2]:
 
2081
                    parent_id = None
 
2082
                return (entry[0][2],
 
2083
                       (utf8_decode(old_path)[0], None),
 
2084
                       True,
 
2085
                       (True, False),
 
2086
                       (parent_id, None),
 
2087
                       (utf8_decode(entry[0][1])[0], None),
 
2088
                       (_minikind_to_kind[source_minikind], None),
 
2089
                       (source_details[3], None))
 
2090
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2091
                # a rename; could be a true rename, or a rename inherited from
 
2092
                # a renamed parent. TODO: handle this efficiently. Its not
 
2093
                # common case to rename dirs though, so a correct but slow
 
2094
                # implementation will do.
 
2095
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2096
                    search_specific_files.add(target_details[1])
 
2097
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2098
                # neither of the selected trees contain this file,
 
2099
                # so skip over it. This is not currently directly tested, but
 
2100
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2101
                pass
 
2102
            else:
 
2103
                raise AssertionError("don't know how to compare "
 
2104
                    "source_minikind=%r, target_minikind=%r"
 
2105
                    % (source_minikind, target_minikind))
 
2106
                ## import pdb;pdb.set_trace()
 
2107
            return None
 
2108
 
 
2109
        while search_specific_files:
 
2110
            # TODO: the pending list should be lexically sorted?  the
 
2111
            # interface doesn't require it.
 
2112
            current_root = search_specific_files.pop()
 
2113
            current_root_unicode = current_root.decode('utf8')
 
2114
            searched_specific_files.add(current_root)
 
2115
            # process the entries for this containing directory: the rest will be
 
2116
            # found by their parents recursively.
 
2117
            root_entries = _entries_for_path(current_root)
 
2118
            root_abspath = self.target.abspath(current_root_unicode)
 
2119
            try:
 
2120
                root_stat = os.lstat(root_abspath)
 
2121
            except OSError, e:
 
2122
                if e.errno == errno.ENOENT:
 
2123
                    # the path does not exist: let _process_entry know that.
 
2124
                    root_dir_info = None
 
2125
                else:
 
2126
                    # some other random error: hand it up.
 
2127
                    raise
 
2128
            else:
 
2129
                root_dir_info = ('', current_root,
 
2130
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2131
                    root_abspath)
 
2132
                if root_dir_info[2] == 'directory':
 
2133
                    if self.target._directory_is_tree_reference(
 
2134
                        current_root.decode('utf8')):
 
2135
                        root_dir_info = root_dir_info[:2] + \
 
2136
                            ('tree-reference',) + root_dir_info[3:]
 
2137
 
 
2138
            if not root_entries and not root_dir_info:
 
2139
                # this specified path is not present at all, skip it.
 
2140
                continue
 
2141
            path_handled = False
 
2142
            for entry in root_entries:
 
2143
                result = _process_entry(entry, root_dir_info)
 
2144
                if result is not None:
 
2145
                    path_handled = True
 
2146
                    if result is not uninteresting:
 
2147
                        yield result
 
2148
            if want_unversioned and not path_handled and root_dir_info:
 
2149
                new_executable = bool(
 
2150
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2151
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2152
                yield (None,
 
2153
                       (None, current_root_unicode),
 
2154
                       True,
 
2155
                       (False, False),
 
2156
                       (None, None),
 
2157
                       (None, splitpath(current_root_unicode)[-1]),
 
2158
                       (None, root_dir_info[2]),
 
2159
                       (None, new_executable)
 
2160
                      )
 
2161
            initial_key = (current_root, '', '')
 
2162
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2163
            if block_index == 0:
 
2164
                # we have processed the total root already, but because the
 
2165
                # initial key matched it we should skip it here.
 
2166
                block_index +=1
 
2167
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2168
                current_dir_info = None
 
2169
            else:
 
2170
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2171
                try:
 
2172
                    current_dir_info = dir_iterator.next()
 
2173
                except OSError, e:
 
2174
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2175
                    # python 2.5 has e.errno == EINVAL,
 
2176
                    #            and e.winerror == ERROR_DIRECTORY
 
2177
                    e_winerror = getattr(e, 'winerror', None)
 
2178
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2179
                    # there may be directories in the inventory even though
 
2180
                    # this path is not a file on disk: so mark it as end of
 
2181
                    # iterator
 
2182
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2183
                        current_dir_info = None
 
2184
                    elif (sys.platform == 'win32'
 
2185
                          and (e.errno in win_errors
 
2186
                               or e_winerror in win_errors)):
 
2187
                        current_dir_info = None
 
2188
                    else:
 
2189
                        raise
 
2190
                else:
 
2191
                    if current_dir_info[0][0] == '':
 
2192
                        # remove .bzr from iteration
 
2193
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2194
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2195
                        del current_dir_info[1][bzr_index]
 
2196
            # walk until both the directory listing and the versioned metadata
 
2197
            # are exhausted. 
 
2198
            if (block_index < len(state._dirblocks) and
 
2199
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2200
                current_block = state._dirblocks[block_index]
 
2201
            else:
 
2202
                current_block = None
 
2203
            while (current_dir_info is not None or
 
2204
                   current_block is not None):
 
2205
                if (current_dir_info and current_block
 
2206
                    and current_dir_info[0][0] != current_block[0]):
 
2207
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2208
                        # filesystem data refers to paths not covered by the dirblock.
 
2209
                        # this has two possibilities:
 
2210
                        # A) it is versioned but empty, so there is no block for it
 
2211
                        # B) it is not versioned.
 
2212
 
 
2213
                        # if (A) then we need to recurse into it to check for
 
2214
                        # new unknown files or directories.
 
2215
                        # if (B) then we should ignore it, because we don't
 
2216
                        # recurse into unknown directories.
 
2217
                        path_index = 0
 
2218
                        while path_index < len(current_dir_info[1]):
 
2219
                                current_path_info = current_dir_info[1][path_index]
 
2220
                                if want_unversioned:
 
2221
                                    if current_path_info[2] == 'directory':
 
2222
                                        if self.target._directory_is_tree_reference(
 
2223
                                            current_path_info[0].decode('utf8')):
 
2224
                                            current_path_info = current_path_info[:2] + \
 
2225
                                                ('tree-reference',) + current_path_info[3:]
 
2226
                                    new_executable = bool(
 
2227
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2228
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2229
                                    yield (None,
 
2230
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2231
                                        True,
 
2232
                                        (False, False),
 
2233
                                        (None, None),
 
2234
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2235
                                        (None, current_path_info[2]),
 
2236
                                        (None, new_executable))
 
2237
                                # dont descend into this unversioned path if it is
 
2238
                                # a dir
 
2239
                                if current_path_info[2] in ('directory',
 
2240
                                                            'tree-reference'):
 
2241
                                    del current_dir_info[1][path_index]
 
2242
                                    path_index -= 1
 
2243
                                path_index += 1
 
2244
 
 
2245
                        # This dir info has been handled, go to the next
 
2246
                        try:
 
2247
                            current_dir_info = dir_iterator.next()
 
2248
                        except StopIteration:
 
2249
                            current_dir_info = None
 
2250
                    else:
 
2251
                        # We have a dirblock entry for this location, but there
 
2252
                        # is no filesystem path for this. This is most likely
 
2253
                        # because a directory was removed from the disk.
 
2254
                        # We don't have to report the missing directory,
 
2255
                        # because that should have already been handled, but we
 
2256
                        # need to handle all of the files that are contained
 
2257
                        # within.
 
2258
                        for current_entry in current_block[1]:
 
2259
                            # entry referring to file not present on disk.
 
2260
                            # advance the entry only, after processing.
 
2261
                            result = _process_entry(current_entry, None)
 
2262
                            if result is not None:
 
2263
                                if result is not uninteresting:
 
2264
                                    yield result
 
2265
                        block_index +=1
 
2266
                        if (block_index < len(state._dirblocks) and
 
2267
                            osutils.is_inside(current_root,
 
2268
                                              state._dirblocks[block_index][0])):
 
2269
                            current_block = state._dirblocks[block_index]
 
2270
                        else:
 
2271
                            current_block = None
 
2272
                    continue
 
2273
                entry_index = 0
 
2274
                if current_block and entry_index < len(current_block[1]):
 
2275
                    current_entry = current_block[1][entry_index]
 
2276
                else:
 
2277
                    current_entry = None
 
2278
                advance_entry = True
 
2279
                path_index = 0
 
2280
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2281
                    current_path_info = current_dir_info[1][path_index]
 
2282
                    if current_path_info[2] == 'directory':
 
2283
                        if self.target._directory_is_tree_reference(
 
2284
                            current_path_info[0].decode('utf8')):
 
2285
                            current_path_info = current_path_info[:2] + \
 
2286
                                ('tree-reference',) + current_path_info[3:]
 
2287
                else:
 
2288
                    current_path_info = None
 
2289
                advance_path = True
 
2290
                path_handled = False
 
2291
                while (current_entry is not None or
 
2292
                    current_path_info is not None):
 
2293
                    if current_entry is None:
 
2294
                        # the check for path_handled when the path is adnvaced
 
2295
                        # will yield this path if needed.
 
2296
                        pass
 
2297
                    elif current_path_info is None:
 
2298
                        # no path is fine: the per entry code will handle it.
 
2299
                        result = _process_entry(current_entry, current_path_info)
 
2300
                        if result is not None:
 
2301
                            if result is not uninteresting:
 
2302
                                yield result
 
2303
                    elif (current_entry[0][1] != current_path_info[1]
 
2304
                          or current_entry[1][target_index][0] in 'ar'):
 
2305
                        # The current path on disk doesn't match the dirblock
 
2306
                        # record. Either the dirblock is marked as absent, or
 
2307
                        # the file on disk is not present at all in the
 
2308
                        # dirblock. Either way, report about the dirblock
 
2309
                        # entry, and let other code handle the filesystem one.
 
2310
 
 
2311
                        # Compare the basename for these files to determine
 
2312
                        # which comes first
 
2313
                        if current_path_info[1] < current_entry[0][1]:
 
2314
                            # extra file on disk: pass for now, but only
 
2315
                            # increment the path, not the entry
 
2316
                            advance_entry = False
 
2317
                        else:
 
2318
                            # entry referring to file not present on disk.
 
2319
                            # advance the entry only, after processing.
 
2320
                            result = _process_entry(current_entry, None)
 
2321
                            if result is not None:
 
2322
                                if result is not uninteresting:
 
2323
                                    yield result
 
2324
                            advance_path = False
 
2325
                    else:
 
2326
                        result = _process_entry(current_entry, current_path_info)
 
2327
                        if result is not None:
 
2328
                            path_handled = True
 
2329
                            if result is not uninteresting:
 
2330
                                yield result
 
2331
                    if advance_entry and current_entry is not None:
 
2332
                        entry_index += 1
 
2333
                        if entry_index < len(current_block[1]):
 
2334
                            current_entry = current_block[1][entry_index]
 
2335
                        else:
 
2336
                            current_entry = None
 
2337
                    else:
 
2338
                        advance_entry = True # reset the advance flaga
 
2339
                    if advance_path and current_path_info is not None:
 
2340
                        if not path_handled:
 
2341
                            # unversioned in all regards
 
2342
                            if want_unversioned:
 
2343
                                new_executable = bool(
 
2344
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2345
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2346
                                yield (None,
 
2347
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2348
                                    True,
 
2349
                                    (False, False),
 
2350
                                    (None, None),
 
2351
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2352
                                    (None, current_path_info[2]),
 
2353
                                    (None, new_executable))
 
2354
                            # dont descend into this unversioned path if it is
 
2355
                            # a dir
 
2356
                            if current_path_info[2] in ('directory'):
 
2357
                                del current_dir_info[1][path_index]
 
2358
                                path_index -= 1
 
2359
                        # dont descend the disk iterator into any tree 
 
2360
                        # paths.
 
2361
                        if current_path_info[2] == 'tree-reference':
 
2362
                            del current_dir_info[1][path_index]
 
2363
                            path_index -= 1
 
2364
                        path_index += 1
 
2365
                        if path_index < len(current_dir_info[1]):
 
2366
                            current_path_info = current_dir_info[1][path_index]
 
2367
                            if current_path_info[2] == 'directory':
 
2368
                                if self.target._directory_is_tree_reference(
 
2369
                                    current_path_info[0].decode('utf8')):
 
2370
                                    current_path_info = current_path_info[:2] + \
 
2371
                                        ('tree-reference',) + current_path_info[3:]
 
2372
                        else:
 
2373
                            current_path_info = None
 
2374
                        path_handled = False
 
2375
                    else:
 
2376
                        advance_path = True # reset the advance flagg.
 
2377
                if current_block is not None:
 
2378
                    block_index += 1
 
2379
                    if (block_index < len(state._dirblocks) and
 
2380
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2381
                        current_block = state._dirblocks[block_index]
 
2382
                    else:
 
2383
                        current_block = None
 
2384
                if current_dir_info is not None:
 
2385
                    try:
 
2386
                        current_dir_info = dir_iterator.next()
 
2387
                    except StopIteration:
 
2388
                        current_dir_info = None
 
2389
 
 
2390
 
 
2391
    @staticmethod
 
2392
    def is_compatible(source, target):
 
2393
        # the target must be a dirstate working tree
 
2394
        if not isinstance(target, WorkingTree4):
 
2395
            return False
 
2396
        # the source must be a revtreee or dirstate rev tree.
 
2397
        if not isinstance(source,
 
2398
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2399
            return False
 
2400
        # the source revid must be in the target dirstate
 
2401
        if not (source._revision_id == NULL_REVISION or
 
2402
            source._revision_id in target.get_parent_ids()):
 
2403
            # TODO: what about ghosts? it may well need to 
 
2404
            # check for them explicitly.
 
2405
            return False
 
2406
        return True
 
2407
 
 
2408
InterTree.register_optimiser(InterDirStateTree)
 
2409
 
 
2410
 
 
2411
class Converter3to4(object):
 
2412
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2413
 
 
2414
    def __init__(self):
 
2415
        self.target_format = WorkingTreeFormat4()
 
2416
 
 
2417
    def convert(self, tree):
 
2418
        # lock the control files not the tree, so that we dont get tree
 
2419
        # on-unlock behaviours, and so that noone else diddles with the 
 
2420
        # tree during upgrade.
 
2421
        tree._control_files.lock_write()
 
2422
        try:
 
2423
            tree.read_working_inventory()
 
2424
            self.create_dirstate_data(tree)
 
2425
            self.update_format(tree)
 
2426
            self.remove_xml_files(tree)
 
2427
        finally:
 
2428
            tree._control_files.unlock()
 
2429
 
 
2430
    def create_dirstate_data(self, tree):
 
2431
        """Create the dirstate based data for tree."""
 
2432
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2433
            ).local_abspath('dirstate')
 
2434
        state = dirstate.DirState.from_tree(tree, local_path)
 
2435
        state.save()
 
2436
        state.unlock()
 
2437
 
 
2438
    def remove_xml_files(self, tree):
 
2439
        """Remove the oldformat 3 data."""
 
2440
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2441
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2442
            'pending-merges', 'stat-cache']:
 
2443
            try:
 
2444
                transport.delete(path)
 
2445
            except errors.NoSuchFile:
 
2446
                # some files are optional - just deal.
 
2447
                pass
 
2448
 
 
2449
    def update_format(self, tree):
 
2450
        """Change the format marker."""
 
2451
        tree._control_files.put_utf8('format',
 
2452
            self.target_format.get_format_string())