/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

Merge bzr.dev

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
 
47
    dirstate,
 
48
    errors,
 
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    transform,
 
59
    urlutils,
 
60
    xml5,
 
61
    xml6,
 
62
    )
 
63
import bzrlib.branch
 
64
from bzrlib.transport import get_transport
 
65
import bzrlib.ui
 
66
""")
 
67
 
 
68
from bzrlib import symbol_versioning
 
69
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
70
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
71
from bzrlib.lockable_files import LockableFiles, TransportLock
 
72
from bzrlib.lockdir import LockDir
 
73
import bzrlib.mutabletree
 
74
from bzrlib.mutabletree import needs_tree_write_lock
 
75
from bzrlib.osutils import (
 
76
    file_kind,
 
77
    isdir,
 
78
    normpath,
 
79
    pathjoin,
 
80
    rand_chars,
 
81
    realpath,
 
82
    safe_unicode,
 
83
    splitpath,
 
84
    )
 
85
from bzrlib.trace import mutter, note
 
86
from bzrlib.transport.local import LocalTransport
 
87
from bzrlib.tree import InterTree
 
88
from bzrlib.progress import DummyProgress, ProgressPhase
 
89
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
90
from bzrlib.rio import RioReader, rio_file, Stanza
 
91
from bzrlib.symbol_versioning import (deprecated_passed,
 
92
        deprecated_method,
 
93
        deprecated_function,
 
94
        DEPRECATED_PARAMETER,
 
95
        )
 
96
from bzrlib.tree import Tree
 
97
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
98
 
 
99
 
 
100
# This is the Windows equivalent of ENOTDIR
 
101
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
102
# just an error code.
 
103
ERROR_PATH_NOT_FOUND = 3
 
104
ERROR_DIRECTORY = 267
 
105
 
 
106
 
 
107
class WorkingTree4(WorkingTree3):
 
108
    """This is the Format 4 working tree.
 
109
 
 
110
    This differs from WorkingTree3 by:
 
111
     - Having a consolidated internal dirstate, stored in a
 
112
       randomly-accessible sorted file on disk.
 
113
     - Not having a regular inventory attribute.  One can be synthesized 
 
114
       on demand but this is expensive and should be avoided.
 
115
 
 
116
    This is new in bzr 0.15.
 
117
    """
 
118
 
 
119
    def __init__(self, basedir,
 
120
                 branch,
 
121
                 _control_files=None,
 
122
                 _format=None,
 
123
                 _bzrdir=None):
 
124
        """Construct a WorkingTree for basedir.
 
125
 
 
126
        If the branch is not supplied, it is opened automatically.
 
127
        If the branch is supplied, it must be the branch for this basedir.
 
128
        (branch.base is not cross checked, because for remote branches that
 
129
        would be meaningless).
 
130
        """
 
131
        self._format = _format
 
132
        self.bzrdir = _bzrdir
 
133
        from bzrlib.trace import note, mutter
 
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
            assert '//' not in f
 
162
            assert '..' not in f
 
163
            if self.path2id(f):
 
164
                # special case tree root handling.
 
165
                if f == '' and self.path2id(f) == ROOT_ID:
 
166
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
167
                continue
 
168
            if file_id is None:
 
169
                file_id = generate_ids.gen_file_id(f)
 
170
            # deliberately add the file with no cached stat or sha1
 
171
            # - on the first access it will be gathered, and we can
 
172
            # always change this once tests are all passing.
 
173
            state.add(f, file_id, kind, None, '')
 
174
        self._make_dirty(reset_inventory=True)
 
175
 
 
176
    def _make_dirty(self, reset_inventory):
 
177
        """Make the tree state dirty.
 
178
 
 
179
        :param reset_inventory: True if the cached inventory should be removed
 
180
            (presuming there is one).
 
181
        """
 
182
        self._dirty = True
 
183
        if reset_inventory and self._inventory is not None:
 
184
            self._inventory = None
 
185
 
 
186
    @needs_tree_write_lock
 
187
    def add_reference(self, sub_tree):
 
188
        # use standard implementation, which calls back to self._add
 
189
        # 
 
190
        # So we don't store the reference_revision in the working dirstate,
 
191
        # it's just recorded at the moment of commit. 
 
192
        self._add_reference(sub_tree)
 
193
 
 
194
    def break_lock(self):
 
195
        """Break a lock if one is present from another instance.
 
196
 
 
197
        Uses the ui factory to ask for confirmation if the lock may be from
 
198
        an active process.
 
199
 
 
200
        This will probe the repository for its lock as well.
 
201
        """
 
202
        # if the dirstate is locked by an active process, reject the break lock
 
203
        # call.
 
204
        try:
 
205
            if self._dirstate is None:
 
206
                clear = True
 
207
            else:
 
208
                clear = False
 
209
            state = self._current_dirstate()
 
210
            if state._lock_token is not None:
 
211
                # we already have it locked. sheese, cant break our own lock.
 
212
                raise errors.LockActive(self.basedir)
 
213
            else:
 
214
                try:
 
215
                    # try for a write lock - need permission to get one anyhow
 
216
                    # to break locks.
 
217
                    state.lock_write()
 
218
                except errors.LockContention:
 
219
                    # oslocks fail when a process is still live: fail.
 
220
                    # TODO: get the locked lockdir info and give to the user to
 
221
                    # assist in debugging.
 
222
                    raise errors.LockActive(self.basedir)
 
223
                else:
 
224
                    state.unlock()
 
225
        finally:
 
226
            if clear:
 
227
                self._dirstate = None
 
228
        self._control_files.break_lock()
 
229
        self.branch.break_lock()
 
230
 
 
231
    def _comparison_data(self, entry, path):
 
232
        kind, executable, stat_value = \
 
233
            WorkingTree3._comparison_data(self, entry, path)
 
234
        # it looks like a plain directory, but it's really a reference -- see
 
235
        # also kind()
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
 
239
            kind = 'tree-reference'
 
240
        return kind, executable, stat_value
 
241
 
 
242
    @needs_write_lock
 
243
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
244
        # mark the tree as dirty post commit - commit
 
245
        # can change the current versioned list by doing deletes.
 
246
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
247
        self._make_dirty(reset_inventory=True)
 
248
        return result
 
249
 
 
250
    def current_dirstate(self):
 
251
        """Return the current dirstate object.
 
252
 
 
253
        This is not part of the tree interface and only exposed for ease of
 
254
        testing.
 
255
 
 
256
        :raises errors.NotWriteLocked: when not in a lock.
 
257
        """
 
258
        self._must_be_locked()
 
259
        return self._current_dirstate()
 
260
 
 
261
    def _current_dirstate(self):
 
262
        """Internal function that does not check lock status.
 
263
 
 
264
        This is needed for break_lock which also needs the dirstate.
 
265
        """
 
266
        if self._dirstate is not None:
 
267
            return self._dirstate
 
268
        local_path = self.bzrdir.get_workingtree_transport(None
 
269
            ).local_abspath('dirstate')
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
 
271
        return self._dirstate
 
272
 
 
273
    def _directory_is_tree_reference(self, relpath):
 
274
        # as a special case, if a directory contains control files then 
 
275
        # it's a tree reference, except that the root of the tree is not
 
276
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
277
        # TODO: We could ask all the control formats whether they
 
278
        # recognize this directory, but at the moment there's no cheap api
 
279
        # to do that.  Since we probably can only nest bzr checkouts and
 
280
        # they always use this name it's ok for now.  -- mbp 20060306
 
281
        #
 
282
        # FIXME: There is an unhandled case here of a subdirectory
 
283
        # containing .bzr but not a branch; that will probably blow up
 
284
        # when you try to commit it.  It might happen if there is a
 
285
        # checkout in a subdirectory.  This can be avoided by not adding
 
286
        # it.  mbp 20070306
 
287
 
 
288
    def filter_unversioned_files(self, paths):
 
289
        """Filter out paths that are versioned.
 
290
 
 
291
        :return: set of paths.
 
292
        """
 
293
        # TODO: make a generic multi-bisect routine roughly that should list
 
294
        # the paths, then process one half at a time recursively, and feed the
 
295
        # results of each bisect in further still
 
296
        paths = sorted(paths)
 
297
        result = set()
 
298
        state = self.current_dirstate()
 
299
        # TODO we want a paths_to_dirblocks helper I think
 
300
        for path in paths:
 
301
            dirname, basename = os.path.split(path.encode('utf8'))
 
302
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
303
                dirname, basename, 0)
 
304
            if not path_is_versioned:
 
305
                result.add(path)
 
306
        return result
 
307
 
 
308
    def flush(self):
 
309
        """Write all cached data to disk."""
 
310
        if self._control_files._lock_mode != 'w':
 
311
            raise errors.NotWriteLocked(self)
 
312
        self.current_dirstate().save()
 
313
        self._inventory = None
 
314
        self._dirty = False
 
315
 
 
316
    @needs_tree_write_lock
 
317
    def _gather_kinds(self, files, kinds):
 
318
        """See MutableTree._gather_kinds."""
 
319
        for pos, f in enumerate(files):
 
320
            if kinds[pos] is None:
 
321
                kinds[pos] = self._kind(f)
 
322
 
 
323
    def _generate_inventory(self):
 
324
        """Create and set self.inventory from the dirstate object.
 
325
        
 
326
        This is relatively expensive: we have to walk the entire dirstate.
 
327
        Ideally we would not, and can deprecate this function.
 
328
        """
 
329
        #: uncomment to trap on inventory requests.
 
330
        # import pdb;pdb.set_trace()
 
331
        state = self.current_dirstate()
 
332
        state._read_dirblocks_if_needed()
 
333
        root_key, current_entry = self._get_entry(path='')
 
334
        current_id = root_key[2]
 
335
        assert current_entry[0][0] == 'd' # directory
 
336
        inv = Inventory(root_id=current_id)
 
337
        # Turn some things into local variables
 
338
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
339
        factory = entry_factory
 
340
        utf8_decode = cache_utf8._utf8_decode
 
341
        inv_byid = inv._byid
 
342
        # we could do this straight out of the dirstate; it might be fast
 
343
        # and should be profiled - RBC 20070216
 
344
        parent_ies = {'' : inv.root}
 
345
        for block in state._dirblocks[1:]: # skip the root
 
346
            dirname = block[0]
 
347
            try:
 
348
                parent_ie = parent_ies[dirname]
 
349
            except KeyError:
 
350
                # all the paths in this block are not versioned in this tree
 
351
                continue
 
352
            for key, entry in block[1]:
 
353
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
354
                if minikind in ('a', 'r'): # absent, relocated
 
355
                    # a parent tree only entry
 
356
                    continue
 
357
                name = key[1]
 
358
                name_unicode = utf8_decode(name)[0]
 
359
                file_id = key[2]
 
360
                kind = minikind_to_kind[minikind]
 
361
                inv_entry = factory[kind](file_id, name_unicode,
 
362
                                          parent_ie.file_id)
 
363
                if kind == 'file':
 
364
                    # This is only needed on win32, where this is the only way
 
365
                    # we know the executable bit.
 
366
                    inv_entry.executable = executable
 
367
                    # not strictly needed: working tree
 
368
                    #inv_entry.text_size = size
 
369
                    #inv_entry.text_sha1 = sha1
 
370
                elif kind == 'directory':
 
371
                    # add this entry to the parent map.
 
372
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
373
                elif kind == 'tree-reference':
 
374
                    assert self._repo_supports_tree_reference, \
 
375
                        "repository of %r " \
 
376
                        "doesn't support tree references " \
 
377
                        "required by entry %r" \
 
378
                        % (self, name)
 
379
                    inv_entry.reference_revision = link_or_sha1 or None
 
380
                elif kind != 'symlink':
 
381
                    raise AssertionError("unknown kind %r" % kind)
 
382
                # These checks cost us around 40ms on a 55k entry tree
 
383
                assert file_id not in inv_byid, ('file_id %s already in'
 
384
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
385
                assert name_unicode not in parent_ie.children
 
386
                inv_byid[file_id] = inv_entry
 
387
                parent_ie.children[name_unicode] = inv_entry
 
388
        self._inventory = inv
 
389
 
 
390
    def _get_entry(self, file_id=None, path=None):
 
391
        """Get the dirstate row for file_id or path.
 
392
 
 
393
        If either file_id or path is supplied, it is used as the key to lookup.
 
394
        If both are supplied, the fastest lookup is used, and an error is
 
395
        raised if they do not both point at the same row.
 
396
        
 
397
        :param file_id: An optional unicode file_id to be looked up.
 
398
        :param path: An optional unicode path to be looked up.
 
399
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
400
        """
 
401
        if file_id is None and path is None:
 
402
            raise errors.BzrError('must supply file_id or path')
 
403
        state = self.current_dirstate()
 
404
        if path is not None:
 
405
            path = path.encode('utf8')
 
406
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
407
 
 
408
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
409
        # check file id is valid unconditionally.
 
410
        entry = self._get_entry(file_id=file_id, path=path)
 
411
        assert entry[0] is not None, 'what error should this raise'
 
412
        if path is None:
 
413
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
414
 
 
415
        file_abspath = self.abspath(path)
 
416
        state = self.current_dirstate()
 
417
        if stat_value is None:
 
418
            try:
 
419
                stat_value = os.lstat(file_abspath)
 
420
            except OSError, e:
 
421
                if e.errno == errno.ENOENT:
 
422
                    return None
 
423
                else:
 
424
                    raise
 
425
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
426
                                          stat_value=stat_value)
 
427
        if entry[1][0][0] == 'f':
 
428
            return link_or_sha1
 
429
        return None
 
430
 
 
431
    def _get_inventory(self):
 
432
        """Get the inventory for the tree. This is only valid within a lock."""
 
433
        if self._inventory is not None:
 
434
            return self._inventory
 
435
        self._must_be_locked()
 
436
        self._generate_inventory()
 
437
        return self._inventory
 
438
 
 
439
    inventory = property(_get_inventory,
 
440
                         doc="Inventory of this Tree")
 
441
 
 
442
    @needs_read_lock
 
443
    def get_parent_ids(self):
 
444
        """See Tree.get_parent_ids.
 
445
        
 
446
        This implementation requests the ids list from the dirstate file.
 
447
        """
 
448
        return self.current_dirstate().get_parent_ids()
 
449
 
 
450
    def get_reference_revision(self, file_id, path=None):
 
451
        # referenced tree's revision is whatever's currently there
 
452
        return self.get_nested_tree(file_id, path).last_revision()
 
453
 
 
454
    def get_nested_tree(self, file_id, path=None):
 
455
        if path is None:
 
456
            path = self.id2path(file_id)
 
457
        # else: check file_id is at path?
 
458
        return WorkingTree.open(self.abspath(path))
 
459
 
 
460
    @needs_read_lock
 
461
    def get_root_id(self):
 
462
        """Return the id of this trees root"""
 
463
        return self._get_entry(path='')[0][2]
 
464
 
 
465
    def has_id(self, file_id):
 
466
        state = self.current_dirstate()
 
467
        file_id = osutils.safe_file_id(file_id)
 
468
        row, parents = self._get_entry(file_id=file_id)
 
469
        if row is None:
 
470
            return False
 
471
        return osutils.lexists(pathjoin(
 
472
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
473
 
 
474
    @needs_read_lock
 
475
    def id2path(self, file_id):
 
476
        "Convert a file-id to a path."
 
477
        file_id = osutils.safe_file_id(file_id)
 
478
        state = self.current_dirstate()
 
479
        entry = self._get_entry(file_id=file_id)
 
480
        if entry == (None, None):
 
481
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
482
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
483
        return path_utf8.decode('utf8')
 
484
 
 
485
    if not osutils.supports_executable():
 
486
        def is_executable(self, file_id, path=None):
 
487
            """Test if a file is executable or not.
 
488
 
 
489
            Note: The caller is expected to take a read-lock before calling this.
 
490
            """
 
491
            file_id = osutils.safe_file_id(file_id)
 
492
            entry = self._get_entry(file_id=file_id, path=path)
 
493
            if entry == (None, None):
 
494
                return False
 
495
            return entry[1][0][3]
 
496
    else:
 
497
        def is_executable(self, file_id, path=None):
 
498
            """Test if a file is executable or not.
 
499
 
 
500
            Note: The caller is expected to take a read-lock before calling this.
 
501
            """
 
502
            if not path:
 
503
                file_id = osutils.safe_file_id(file_id)
 
504
                path = self.id2path(file_id)
 
505
            mode = os.lstat(self.abspath(path)).st_mode
 
506
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
507
 
 
508
    @needs_read_lock
 
509
    def __iter__(self):
 
510
        """Iterate through file_ids for this tree.
 
511
 
 
512
        file_ids are in a WorkingTree if they are in the working inventory
 
513
        and the working file exists.
 
514
        """
 
515
        result = []
 
516
        for key, tree_details in self.current_dirstate()._iter_entries():
 
517
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
518
                # not relevant to the working tree
 
519
                continue
 
520
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
521
            if osutils.lexists(path):
 
522
                result.append(key[2])
 
523
        return iter(result)
 
524
 
 
525
    def iter_references(self):
 
526
        for key, tree_details in self.current_dirstate()._iter_entries():
 
527
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
528
                # not relevant to the working tree
 
529
                continue
 
530
            if not key[1]:
 
531
                # the root is not a reference.
 
532
                continue
 
533
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
534
            try:
 
535
                if self._kind(path) == 'tree-reference':
 
536
                    yield path, key[2]
 
537
            except errors.NoSuchFile:
 
538
                # path is missing on disk.
 
539
                continue
 
540
 
 
541
    def kind(self, file_id):
 
542
        """Return the kind of a file.
 
543
 
 
544
        This is always the actual kind that's on disk, regardless of what it
 
545
        was added as.
 
546
 
 
547
        Note: The caller is expected to take a read-lock before calling this.
 
548
        """
 
549
        relpath = self.id2path(file_id)
 
550
        assert relpath != None, \
 
551
            "path for id {%s} is None!" % file_id
 
552
        return self._kind(relpath)
 
553
 
 
554
    def _kind(self, relpath):
 
555
        abspath = self.abspath(relpath)
 
556
        kind = file_kind(abspath)
 
557
        if (self._repo_supports_tree_reference and
 
558
            kind == 'directory' and
 
559
            self._directory_is_tree_reference(relpath)):
 
560
            kind = 'tree-reference'
 
561
        return kind
 
562
 
 
563
    @needs_read_lock
 
564
    def _last_revision(self):
 
565
        """See Mutable.last_revision."""
 
566
        parent_ids = self.current_dirstate().get_parent_ids()
 
567
        if parent_ids:
 
568
            return parent_ids[0]
 
569
        else:
 
570
            return None
 
571
 
 
572
    def lock_read(self):
 
573
        """See Branch.lock_read, and WorkingTree.unlock."""
 
574
        self.branch.lock_read()
 
575
        try:
 
576
            self._control_files.lock_read()
 
577
            try:
 
578
                state = self.current_dirstate()
 
579
                if not state._lock_token:
 
580
                    state.lock_read()
 
581
                # set our support for tree references from the repository in
 
582
                # use.
 
583
                self._repo_supports_tree_reference = getattr(
 
584
                    self.branch.repository._format, "supports_tree_reference",
 
585
                    False)
 
586
            except:
 
587
                self._control_files.unlock()
 
588
                raise
 
589
        except:
 
590
            self.branch.unlock()
 
591
            raise
 
592
 
 
593
    def _lock_self_write(self):
 
594
        """This should be called after the branch is locked."""
 
595
        try:
 
596
            self._control_files.lock_write()
 
597
            try:
 
598
                state = self.current_dirstate()
 
599
                if not state._lock_token:
 
600
                    state.lock_write()
 
601
                # set our support for tree references from the repository in
 
602
                # use.
 
603
                self._repo_supports_tree_reference = getattr(
 
604
                    self.branch.repository._format, "supports_tree_reference",
 
605
                    False)
 
606
            except:
 
607
                self._control_files.unlock()
 
608
                raise
 
609
        except:
 
610
            self.branch.unlock()
 
611
            raise
 
612
 
 
613
    def lock_tree_write(self):
 
614
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
615
        self.branch.lock_read()
 
616
        self._lock_self_write()
 
617
 
 
618
    def lock_write(self):
 
619
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
620
        self.branch.lock_write()
 
621
        self._lock_self_write()
 
622
 
 
623
    @needs_tree_write_lock
 
624
    def move(self, from_paths, to_dir, after=False):
 
625
        """See WorkingTree.move()."""
 
626
        result = []
 
627
        if not from_paths:
 
628
            return result
 
629
 
 
630
        state = self.current_dirstate()
 
631
 
 
632
        assert not isinstance(from_paths, basestring)
 
633
        to_dir_utf8 = to_dir.encode('utf8')
 
634
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
635
        id_index = state._get_id_index()
 
636
        # check destination directory
 
637
        # get the details for it
 
638
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
639
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
640
        if not entry_present:
 
641
            raise errors.BzrMoveFailedError('', to_dir,
 
642
                errors.NotVersionedError(to_dir))
 
643
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
644
        # get a handle on the block itself.
 
645
        to_block_index = state._ensure_block(
 
646
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
647
        to_block = state._dirblocks[to_block_index]
 
648
        to_abs = self.abspath(to_dir)
 
649
        if not isdir(to_abs):
 
650
            raise errors.BzrMoveFailedError('',to_dir,
 
651
                errors.NotADirectory(to_abs))
 
652
 
 
653
        if to_entry[1][0][0] != 'd':
 
654
            raise errors.BzrMoveFailedError('',to_dir,
 
655
                errors.NotADirectory(to_abs))
 
656
 
 
657
        if self._inventory is not None:
 
658
            update_inventory = True
 
659
            inv = self.inventory
 
660
            to_dir_ie = inv[to_dir_id]
 
661
            to_dir_id = to_entry[0][2]
 
662
        else:
 
663
            update_inventory = False
 
664
 
 
665
        rollbacks = []
 
666
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
667
                     fingerprint, packed_stat, size,
 
668
                     to_block, to_key, to_path_utf8):
 
669
            state._make_absent(old_entry)
 
670
            from_key = old_entry[0]
 
671
            rollbacks.append(
 
672
                lambda:state.update_minimal(from_key,
 
673
                    minikind,
 
674
                    executable=executable,
 
675
                    fingerprint=fingerprint,
 
676
                    packed_stat=packed_stat,
 
677
                    size=size,
 
678
                    path_utf8=from_path_utf8))
 
679
            state.update_minimal(to_key,
 
680
                    minikind,
 
681
                    executable=executable,
 
682
                    fingerprint=fingerprint,
 
683
                    packed_stat=packed_stat,
 
684
                    size=size,
 
685
                    path_utf8=to_path_utf8)
 
686
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
687
            new_entry = to_block[1][added_entry_index]
 
688
            rollbacks.append(lambda:state._make_absent(new_entry))
 
689
 
 
690
        for from_rel in from_paths:
 
691
            # from_rel is 'pathinroot/foo/bar'
 
692
            from_rel_utf8 = from_rel.encode('utf8')
 
693
            from_dirname, from_tail = osutils.split(from_rel)
 
694
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
695
            from_entry = self._get_entry(path=from_rel)
 
696
            if from_entry == (None, None):
 
697
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
698
                    errors.NotVersionedError(path=str(from_rel)))
 
699
 
 
700
            from_id = from_entry[0][2]
 
701
            to_rel = pathjoin(to_dir, from_tail)
 
702
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
703
            item_to_entry = self._get_entry(path=to_rel)
 
704
            if item_to_entry != (None, None):
 
705
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
706
                    "Target is already versioned.")
 
707
 
 
708
            if from_rel == to_rel:
 
709
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
710
                    "Source and target are identical.")
 
711
 
 
712
            from_missing = not self.has_filename(from_rel)
 
713
            to_missing = not self.has_filename(to_rel)
 
714
            if after:
 
715
                move_file = False
 
716
            else:
 
717
                move_file = True
 
718
            if to_missing:
 
719
                if not move_file:
 
720
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
721
                        errors.NoSuchFile(path=to_rel,
 
722
                        extra="New file has not been created yet"))
 
723
                elif from_missing:
 
724
                    # neither path exists
 
725
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
726
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
727
            else:
 
728
                if from_missing: # implicitly just update our path mapping
 
729
                    move_file = False
 
730
                elif not after:
 
731
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
732
                        extra="(Use --after to update the Bazaar id)")
 
733
 
 
734
            rollbacks = []
 
735
            def rollback_rename():
 
736
                """A single rename has failed, roll it back."""
 
737
                # roll back everything, even if we encounter trouble doing one
 
738
                # of them.
 
739
                #
 
740
                # TODO: at least log the other exceptions rather than just
 
741
                # losing them mbp 20070307
 
742
                exc_info = None
 
743
                for rollback in reversed(rollbacks):
 
744
                    try:
 
745
                        rollback()
 
746
                    except Exception, e:
 
747
                        exc_info = sys.exc_info()
 
748
                if exc_info:
 
749
                    raise exc_info[0], exc_info[1], exc_info[2]
 
750
 
 
751
            # perform the disk move first - its the most likely failure point.
 
752
            if move_file:
 
753
                from_rel_abs = self.abspath(from_rel)
 
754
                to_rel_abs = self.abspath(to_rel)
 
755
                try:
 
756
                    osutils.rename(from_rel_abs, to_rel_abs)
 
757
                except OSError, e:
 
758
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
759
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
760
            try:
 
761
                # perform the rename in the inventory next if needed: its easy
 
762
                # to rollback
 
763
                if update_inventory:
 
764
                    # rename the entry
 
765
                    from_entry = inv[from_id]
 
766
                    current_parent = from_entry.parent_id
 
767
                    inv.rename(from_id, to_dir_id, from_tail)
 
768
                    rollbacks.append(
 
769
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
770
                # finally do the rename in the dirstate, which is a little
 
771
                # tricky to rollback, but least likely to need it.
 
772
                old_block_index, old_entry_index, dir_present, file_present = \
 
773
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
774
                old_block = state._dirblocks[old_block_index][1]
 
775
                old_entry = old_block[old_entry_index]
 
776
                from_key, old_entry_details = old_entry
 
777
                cur_details = old_entry_details[0]
 
778
                # remove the old row
 
779
                to_key = ((to_block[0],) + from_key[1:3])
 
780
                minikind = cur_details[0]
 
781
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
782
                         minikind=minikind,
 
783
                         executable=cur_details[3],
 
784
                         fingerprint=cur_details[1],
 
785
                         packed_stat=cur_details[4],
 
786
                         size=cur_details[2],
 
787
                         to_block=to_block,
 
788
                         to_key=to_key,
 
789
                         to_path_utf8=to_rel_utf8)
 
790
 
 
791
                if minikind == 'd':
 
792
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
793
                        """Recursively update all entries in this dirblock."""
 
794
                        assert from_dir != '', "renaming root not supported"
 
795
                        from_key = (from_dir, '')
 
796
                        from_block_idx, present = \
 
797
                            state._find_block_index_from_key(from_key)
 
798
                        if not present:
 
799
                            # This is the old record, if it isn't present, then
 
800
                            # there is theoretically nothing to update.
 
801
                            # (Unless it isn't present because of lazy loading,
 
802
                            # but we don't do that yet)
 
803
                            return
 
804
                        from_block = state._dirblocks[from_block_idx]
 
805
                        to_block_index, to_entry_index, _, _ = \
 
806
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
807
                        to_block_index = state._ensure_block(
 
808
                            to_block_index, to_entry_index, to_dir_utf8)
 
809
                        to_block = state._dirblocks[to_block_index]
 
810
 
 
811
                        # Grab a copy since move_one may update the list.
 
812
                        for entry in from_block[1][:]:
 
813
                            assert entry[0][0] == from_dir
 
814
                            cur_details = entry[1][0]
 
815
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
816
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
817
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
818
                            minikind = cur_details[0]
 
819
                            if minikind in 'ar':
 
820
                                # Deleted children of a renamed directory
 
821
                                # Do not need to be updated.
 
822
                                # Children that have been renamed out of this
 
823
                                # directory should also not be updated
 
824
                                continue
 
825
                            move_one(entry, from_path_utf8=from_path_utf8,
 
826
                                     minikind=minikind,
 
827
                                     executable=cur_details[3],
 
828
                                     fingerprint=cur_details[1],
 
829
                                     packed_stat=cur_details[4],
 
830
                                     size=cur_details[2],
 
831
                                     to_block=to_block,
 
832
                                     to_key=to_key,
 
833
                                     to_path_utf8=to_path_utf8)
 
834
                            if minikind == 'd':
 
835
                                # We need to move all the children of this
 
836
                                # entry
 
837
                                update_dirblock(from_path_utf8, to_key,
 
838
                                                to_path_utf8)
 
839
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
840
            except:
 
841
                rollback_rename()
 
842
                raise
 
843
            result.append((from_rel, to_rel))
 
844
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
845
            self._make_dirty(reset_inventory=False)
 
846
 
 
847
        return result
 
848
 
 
849
    def _must_be_locked(self):
 
850
        if not self._control_files._lock_count:
 
851
            raise errors.ObjectNotLocked(self)
 
852
 
 
853
    def _new_tree(self):
 
854
        """Initialize the state in this tree to be a new tree."""
 
855
        self._dirty = True
 
856
 
 
857
    @needs_read_lock
 
858
    def path2id(self, path):
 
859
        """Return the id for path in this tree."""
 
860
        path = path.strip('/')
 
861
        entry = self._get_entry(path=path)
 
862
        if entry == (None, None):
 
863
            return None
 
864
        return entry[0][2]
 
865
 
 
866
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
867
        """See Tree.paths2ids().
 
868
 
 
869
        This specialisation fast-paths the case where all the trees are in the
 
870
        dirstate.
 
871
        """
 
872
        if paths is None:
 
873
            return None
 
874
        parents = self.get_parent_ids()
 
875
        for tree in trees:
 
876
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
877
                parents):
 
878
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
879
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
880
        # -- make all paths utf8 --
 
881
        paths_utf8 = set()
 
882
        for path in paths:
 
883
            paths_utf8.add(path.encode('utf8'))
 
884
        paths = paths_utf8
 
885
        # -- paths is now a utf8 path set --
 
886
        # -- get the state object and prepare it.
 
887
        state = self.current_dirstate()
 
888
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
889
            and '' not in paths):
 
890
            paths2ids = self._paths2ids_using_bisect
 
891
        else:
 
892
            paths2ids = self._paths2ids_in_memory
 
893
        return paths2ids(paths, search_indexes,
 
894
                         require_versioned=require_versioned)
 
895
 
 
896
    def _paths2ids_in_memory(self, paths, search_indexes,
 
897
                             require_versioned=True):
 
898
        state = self.current_dirstate()
 
899
        state._read_dirblocks_if_needed()
 
900
        def _entries_for_path(path):
 
901
            """Return a list with all the entries that match path for all ids.
 
902
            """
 
903
            dirname, basename = os.path.split(path)
 
904
            key = (dirname, basename, '')
 
905
            block_index, present = state._find_block_index_from_key(key)
 
906
            if not present:
 
907
                # the block which should contain path is absent.
 
908
                return []
 
909
            result = []
 
910
            block = state._dirblocks[block_index][1]
 
911
            entry_index, _ = state._find_entry_index(key, block)
 
912
            # we may need to look at multiple entries at this path: walk while the paths match.
 
913
            while (entry_index < len(block) and
 
914
                block[entry_index][0][0:2] == key[0:2]):
 
915
                result.append(block[entry_index])
 
916
                entry_index += 1
 
917
            return result
 
918
        if require_versioned:
 
919
            # -- check all supplied paths are versioned in a search tree. --
 
920
            all_versioned = True
 
921
            for path in paths:
 
922
                path_entries = _entries_for_path(path)
 
923
                if not path_entries:
 
924
                    # this specified path is not present at all: error
 
925
                    all_versioned = False
 
926
                    break
 
927
                found_versioned = False
 
928
                # for each id at this path
 
929
                for entry in path_entries:
 
930
                    # for each tree.
 
931
                    for index in search_indexes:
 
932
                        if entry[1][index][0] != 'a': # absent
 
933
                            found_versioned = True
 
934
                            # all good: found a versioned cell
 
935
                            break
 
936
                if not found_versioned:
 
937
                    # none of the indexes was not 'absent' at all ids for this
 
938
                    # path.
 
939
                    all_versioned = False
 
940
                    break
 
941
            if not all_versioned:
 
942
                raise errors.PathsNotVersionedError(paths)
 
943
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
944
        search_paths = set()
 
945
        for path in paths:
 
946
            other_paths = paths.difference(set([path]))
 
947
            if not osutils.is_inside_any(other_paths, path):
 
948
                # this is a top level path, we must check it.
 
949
                search_paths.add(path)
 
950
        # sketch: 
 
951
        # for all search_indexs in each path at or under each element of
 
952
        # search_paths, if the detail is relocated: add the id, and add the
 
953
        # relocated path as one to search if its not searched already. If the
 
954
        # detail is not relocated, add the id.
 
955
        searched_paths = set()
 
956
        found_ids = set()
 
957
        def _process_entry(entry):
 
958
            """Look at search_indexes within entry.
 
959
 
 
960
            If a specific tree's details are relocated, add the relocation
 
961
            target to search_paths if not searched already. If it is absent, do
 
962
            nothing. Otherwise add the id to found_ids.
 
963
            """
 
964
            for index in search_indexes:
 
965
                if entry[1][index][0] == 'r': # relocated
 
966
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
967
                        search_paths.add(entry[1][index][1])
 
968
                elif entry[1][index][0] != 'a': # absent
 
969
                    found_ids.add(entry[0][2])
 
970
        while search_paths:
 
971
            current_root = search_paths.pop()
 
972
            searched_paths.add(current_root)
 
973
            # process the entries for this containing directory: the rest will be
 
974
            # found by their parents recursively.
 
975
            root_entries = _entries_for_path(current_root)
 
976
            if not root_entries:
 
977
                # this specified path is not present at all, skip it.
 
978
                continue
 
979
            for entry in root_entries:
 
980
                _process_entry(entry)
 
981
            initial_key = (current_root, '', '')
 
982
            block_index, _ = state._find_block_index_from_key(initial_key)
 
983
            while (block_index < len(state._dirblocks) and
 
984
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
985
                for entry in state._dirblocks[block_index][1]:
 
986
                    _process_entry(entry)
 
987
                block_index += 1
 
988
        return found_ids
 
989
 
 
990
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
991
                                require_versioned=True):
 
992
        state = self.current_dirstate()
 
993
        found_ids = set()
 
994
 
 
995
        split_paths = sorted(osutils.split(p) for p in paths)
 
996
        found = state._bisect_recursive(split_paths)
 
997
 
 
998
        if require_versioned:
 
999
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
1000
            for dir_name in split_paths:
 
1001
                if dir_name not in found_dir_names:
 
1002
                    raise errors.PathsNotVersionedError(paths)
 
1003
 
 
1004
        for dir_name_id, trees_info in found.iteritems():
 
1005
            for index in search_indexes:
 
1006
                if trees_info[index][0] not in ('r', 'a'):
 
1007
                    found_ids.add(dir_name_id[2])
 
1008
        return found_ids
 
1009
 
 
1010
    def read_working_inventory(self):
 
1011
        """Read the working inventory.
 
1012
        
 
1013
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1014
        """
 
1015
        return self.inventory
 
1016
 
 
1017
    @needs_read_lock
 
1018
    def revision_tree(self, revision_id):
 
1019
        """See Tree.revision_tree.
 
1020
 
 
1021
        WorkingTree4 supplies revision_trees for any basis tree.
 
1022
        """
 
1023
        revision_id = osutils.safe_revision_id(revision_id)
 
1024
        dirstate = self.current_dirstate()
 
1025
        parent_ids = dirstate.get_parent_ids()
 
1026
        if revision_id not in parent_ids:
 
1027
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1028
        if revision_id in dirstate.get_ghosts():
 
1029
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1030
        return DirStateRevisionTree(dirstate, revision_id,
 
1031
            self.branch.repository)
 
1032
 
 
1033
    @needs_tree_write_lock
 
1034
    def set_last_revision(self, new_revision):
 
1035
        """Change the last revision in the working tree."""
 
1036
        new_revision = osutils.safe_revision_id(new_revision)
 
1037
        parents = self.get_parent_ids()
 
1038
        if new_revision in (NULL_REVISION, None):
 
1039
            assert len(parents) < 2, (
 
1040
                "setting the last parent to none with a pending merge is "
 
1041
                "unsupported.")
 
1042
            self.set_parent_ids([])
 
1043
        else:
 
1044
            self.set_parent_ids([new_revision] + parents[1:],
 
1045
                allow_leftmost_as_ghost=True)
 
1046
 
 
1047
    @needs_tree_write_lock
 
1048
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1049
        """Set the parent ids to revision_ids.
 
1050
        
 
1051
        See also set_parent_trees. This api will try to retrieve the tree data
 
1052
        for each element of revision_ids from the trees repository. If you have
 
1053
        tree data already available, it is more efficient to use
 
1054
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1055
        an easier API to use.
 
1056
 
 
1057
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1058
            working tree. Any of these may be ghosts.
 
1059
        """
 
1060
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1061
        trees = []
 
1062
        for revision_id in revision_ids:
 
1063
            try:
 
1064
                revtree = self.branch.repository.revision_tree(revision_id)
 
1065
                # TODO: jam 20070213 KnitVersionedFile raises
 
1066
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1067
                #       given revision_id is not present. Should Repository be
 
1068
                #       catching it and re-raising NoSuchRevision?
 
1069
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1070
                revtree = None
 
1071
            trees.append((revision_id, revtree))
 
1072
        self.set_parent_trees(trees,
 
1073
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1074
 
 
1075
    @needs_tree_write_lock
 
1076
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1077
        """Set the parents of the working tree.
 
1078
 
 
1079
        :param parents_list: A list of (revision_id, tree) tuples.
 
1080
            If tree is None, then that element is treated as an unreachable
 
1081
            parent tree - i.e. a ghost.
 
1082
        """
 
1083
        dirstate = self.current_dirstate()
 
1084
        if len(parents_list) > 0:
 
1085
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1086
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1087
        real_trees = []
 
1088
        ghosts = []
 
1089
        # convert absent trees to the null tree, which we convert back to
 
1090
        # missing on access.
 
1091
        for rev_id, tree in parents_list:
 
1092
            rev_id = osutils.safe_revision_id(rev_id)
 
1093
            _mod_revision.check_not_reserved_id(rev_id)
 
1094
            if tree is not None:
 
1095
                real_trees.append((rev_id, tree))
 
1096
            else:
 
1097
                real_trees.append((rev_id,
 
1098
                    self.branch.repository.revision_tree(None)))
 
1099
                ghosts.append(rev_id)
 
1100
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1101
        self._make_dirty(reset_inventory=False)
 
1102
 
 
1103
    def _set_root_id(self, file_id):
 
1104
        """See WorkingTree.set_root_id."""
 
1105
        state = self.current_dirstate()
 
1106
        state.set_path_id('', file_id)
 
1107
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1108
            self._make_dirty(reset_inventory=True)
 
1109
 
 
1110
    @needs_read_lock
 
1111
    def supports_tree_reference(self):
 
1112
        return self._repo_supports_tree_reference
 
1113
 
 
1114
    def unlock(self):
 
1115
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1116
        if self._control_files._lock_count == 1:
 
1117
            # eventually we should do signature checking during read locks for
 
1118
            # dirstate updates.
 
1119
            if self._control_files._lock_mode == 'w':
 
1120
                if self._dirty:
 
1121
                    self.flush()
 
1122
            if self._dirstate is not None:
 
1123
                # This is a no-op if there are no modifications.
 
1124
                self._dirstate.save()
 
1125
                self._dirstate.unlock()
 
1126
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1127
            #       point. Instead, it could check if the header has been
 
1128
            #       modified when it is locked, and if not, it can hang on to
 
1129
            #       the data it has in memory.
 
1130
            self._dirstate = None
 
1131
            self._inventory = None
 
1132
        # reverse order of locking.
 
1133
        try:
 
1134
            return self._control_files.unlock()
 
1135
        finally:
 
1136
            self.branch.unlock()
 
1137
 
 
1138
    @needs_tree_write_lock
 
1139
    def unversion(self, file_ids):
 
1140
        """Remove the file ids in file_ids from the current versioned set.
 
1141
 
 
1142
        When a file_id is unversioned, all of its children are automatically
 
1143
        unversioned.
 
1144
 
 
1145
        :param file_ids: The file ids to stop versioning.
 
1146
        :raises: NoSuchId if any fileid is not currently versioned.
 
1147
        """
 
1148
        if not file_ids:
 
1149
            return
 
1150
        state = self.current_dirstate()
 
1151
        state._read_dirblocks_if_needed()
 
1152
        ids_to_unversion = set()
 
1153
        for file_id in file_ids:
 
1154
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1155
        paths_to_unversion = set()
 
1156
        # sketch:
 
1157
        # check if the root is to be unversioned, if so, assert for now.
 
1158
        # walk the state marking unversioned things as absent.
 
1159
        # if there are any un-unversioned ids at the end, raise
 
1160
        for key, details in state._dirblocks[0][1]:
 
1161
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1162
                key[2] in ids_to_unversion):
 
1163
                # I haven't written the code to unversion / yet - it should be
 
1164
                # supported.
 
1165
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1166
        block_index = 0
 
1167
        while block_index < len(state._dirblocks):
 
1168
            # process one directory at a time.
 
1169
            block = state._dirblocks[block_index]
 
1170
            # first check: is the path one to remove - it or its children
 
1171
            delete_block = False
 
1172
            for path in paths_to_unversion:
 
1173
                if (block[0].startswith(path) and
 
1174
                    (len(block[0]) == len(path) or
 
1175
                     block[0][len(path)] == '/')):
 
1176
                    # this entire block should be deleted - its the block for a
 
1177
                    # path to unversion; or the child of one
 
1178
                    delete_block = True
 
1179
                    break
 
1180
            # TODO: trim paths_to_unversion as we pass by paths
 
1181
            if delete_block:
 
1182
                # this block is to be deleted: process it.
 
1183
                # TODO: we can special case the no-parents case and
 
1184
                # just forget the whole block.
 
1185
                entry_index = 0
 
1186
                while entry_index < len(block[1]):
 
1187
                    # Mark this file id as having been removed
 
1188
                    entry = block[1][entry_index]
 
1189
                    ids_to_unversion.discard(entry[0][2])
 
1190
                    if (entry[1][0][0] == 'a'
 
1191
                        or not state._make_absent(entry)):
 
1192
                        entry_index += 1
 
1193
                # go to the next block. (At the moment we dont delete empty
 
1194
                # dirblocks)
 
1195
                block_index += 1
 
1196
                continue
 
1197
            entry_index = 0
 
1198
            while entry_index < len(block[1]):
 
1199
                entry = block[1][entry_index]
 
1200
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1201
                    # ^ some parent row.
 
1202
                    entry[0][2] not in ids_to_unversion):
 
1203
                    # ^ not an id to unversion
 
1204
                    entry_index += 1
 
1205
                    continue
 
1206
                if entry[1][0][0] == 'd':
 
1207
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1208
                if not state._make_absent(entry):
 
1209
                    entry_index += 1
 
1210
                # we have unversioned this id
 
1211
                ids_to_unversion.remove(entry[0][2])
 
1212
            block_index += 1
 
1213
        if ids_to_unversion:
 
1214
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1215
        self._make_dirty(reset_inventory=False)
 
1216
        # have to change the legacy inventory too.
 
1217
        if self._inventory is not None:
 
1218
            for file_id in file_ids:
 
1219
                self._inventory.remove_recursive_id(file_id)
 
1220
 
 
1221
    @needs_read_lock
 
1222
    def _validate(self):
 
1223
        self._dirstate._validate()
 
1224
 
 
1225
    @needs_tree_write_lock
 
1226
    def _write_inventory(self, inv):
 
1227
        """Write inventory as the current inventory."""
 
1228
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1229
        self.current_dirstate().set_state_from_inventory(inv)
 
1230
        self._make_dirty(reset_inventory=False)
 
1231
        if self._inventory is not None:
 
1232
            self._inventory = inv
 
1233
        self.flush()
 
1234
 
 
1235
 
 
1236
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1237
    """The first consolidated dirstate working tree format.
 
1238
 
 
1239
    This format:
 
1240
        - exists within a metadir controlling .bzr
 
1241
        - includes an explicit version marker for the workingtree control
 
1242
          files, separate from the BzrDir format
 
1243
        - modifies the hash cache format
 
1244
        - is new in bzr 0.15
 
1245
        - uses a LockDir to guard access to it.
 
1246
    """
 
1247
 
 
1248
    upgrade_recommended = False
 
1249
 
 
1250
    def get_format_string(self):
 
1251
        """See WorkingTreeFormat.get_format_string()."""
 
1252
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1253
 
 
1254
    def get_format_description(self):
 
1255
        """See WorkingTreeFormat.get_format_description()."""
 
1256
        return "Working tree format 4"
 
1257
 
 
1258
    def initialize(self, a_bzrdir, revision_id=None):
 
1259
        """See WorkingTreeFormat.initialize().
 
1260
 
 
1261
        :param revision_id: allows creating a working tree at a different
 
1262
        revision than the branch is at.
 
1263
 
 
1264
        These trees get an initial random root id, if their repository supports
 
1265
        rich root data, TREE_ROOT otherwise.
 
1266
        """
 
1267
        revision_id = osutils.safe_revision_id(revision_id)
 
1268
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1269
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1270
        transport = a_bzrdir.get_workingtree_transport(self)
 
1271
        control_files = self._open_control_files(a_bzrdir)
 
1272
        control_files.create_lock()
 
1273
        control_files.lock_write()
 
1274
        control_files.put_utf8('format', self.get_format_string())
 
1275
        branch = a_bzrdir.open_branch()
 
1276
        if revision_id is None:
 
1277
            revision_id = branch.last_revision()
 
1278
        local_path = transport.local_abspath('dirstate')
 
1279
        # write out new dirstate (must exist when we create the tree)
 
1280
        state = dirstate.DirState.initialize(local_path)
 
1281
        state.unlock()
 
1282
        del state
 
1283
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1284
                         branch,
 
1285
                         _format=self,
 
1286
                         _bzrdir=a_bzrdir,
 
1287
                         _control_files=control_files)
 
1288
        wt._new_tree()
 
1289
        wt.lock_tree_write()
 
1290
        try:
 
1291
            if revision_id in (None, NULL_REVISION):
 
1292
                if branch.repository.supports_rich_root():
 
1293
                    wt._set_root_id(generate_ids.gen_root_id())
 
1294
                else:
 
1295
                    wt._set_root_id(ROOT_ID)
 
1296
                wt.flush()
 
1297
            wt.set_last_revision(revision_id)
 
1298
            wt.flush()
 
1299
            basis = wt.basis_tree()
 
1300
            basis.lock_read()
 
1301
            # if the basis has a root id we have to use that; otherwise we use
 
1302
            # a new random one
 
1303
            basis_root_id = basis.get_root_id()
 
1304
            if basis_root_id is not None:
 
1305
                wt._set_root_id(basis_root_id)
 
1306
                wt.flush()
 
1307
            transform.build_tree(basis, wt)
 
1308
            basis.unlock()
 
1309
        finally:
 
1310
            control_files.unlock()
 
1311
            wt.unlock()
 
1312
        return wt
 
1313
 
 
1314
    def _open(self, a_bzrdir, control_files):
 
1315
        """Open the tree itself.
 
1316
 
 
1317
        :param a_bzrdir: the dir for the tree.
 
1318
        :param control_files: the control files for the tree.
 
1319
        """
 
1320
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1321
                           branch=a_bzrdir.open_branch(),
 
1322
                           _format=self,
 
1323
                           _bzrdir=a_bzrdir,
 
1324
                           _control_files=control_files)
 
1325
 
 
1326
    def __get_matchingbzrdir(self):
 
1327
        # please test against something that will let us do tree references
 
1328
        return bzrdir.format_registry.make_bzrdir(
 
1329
            'dirstate-with-subtree')
 
1330
 
 
1331
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1332
 
 
1333
 
 
1334
class DirStateRevisionTree(Tree):
 
1335
    """A revision tree pulling the inventory from a dirstate."""
 
1336
 
 
1337
    def __init__(self, dirstate, revision_id, repository):
 
1338
        self._dirstate = dirstate
 
1339
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1340
        self._repository = repository
 
1341
        self._inventory = None
 
1342
        self._locked = 0
 
1343
        self._dirstate_locked = False
 
1344
 
 
1345
    def __repr__(self):
 
1346
        return "<%s of %s in %s>" % \
 
1347
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1348
 
 
1349
    def annotate_iter(self, file_id):
 
1350
        """See Tree.annotate_iter"""
 
1351
        w = self._repository.weave_store.get_weave(file_id,
 
1352
                           self._repository.get_transaction())
 
1353
        return w.annotate_iter(self.inventory[file_id].revision)
 
1354
 
 
1355
    def _comparison_data(self, entry, path):
 
1356
        """See Tree._comparison_data."""
 
1357
        if entry is None:
 
1358
            return None, False, None
 
1359
        # trust the entry as RevisionTree does, but this may not be
 
1360
        # sensible: the entry might not have come from us?
 
1361
        return entry.kind, entry.executable, None
 
1362
 
 
1363
    def _file_size(self, entry, stat_value):
 
1364
        return entry.text_size
 
1365
 
 
1366
    def filter_unversioned_files(self, paths):
 
1367
        """Filter out paths that are not versioned.
 
1368
 
 
1369
        :return: set of paths.
 
1370
        """
 
1371
        pred = self.has_filename
 
1372
        return set((p for p in paths if not pred(p)))
 
1373
 
 
1374
    def get_root_id(self):
 
1375
        return self.path2id('')
 
1376
 
 
1377
    def _get_parent_index(self):
 
1378
        """Return the index in the dirstate referenced by this tree."""
 
1379
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1380
 
 
1381
    def _get_entry(self, file_id=None, path=None):
 
1382
        """Get the dirstate row for file_id or path.
 
1383
 
 
1384
        If either file_id or path is supplied, it is used as the key to lookup.
 
1385
        If both are supplied, the fastest lookup is used, and an error is
 
1386
        raised if they do not both point at the same row.
 
1387
        
 
1388
        :param file_id: An optional unicode file_id to be looked up.
 
1389
        :param path: An optional unicode path to be looked up.
 
1390
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1391
        """
 
1392
        if file_id is None and path is None:
 
1393
            raise errors.BzrError('must supply file_id or path')
 
1394
        file_id = osutils.safe_file_id(file_id)
 
1395
        if path is not None:
 
1396
            path = path.encode('utf8')
 
1397
        parent_index = self._get_parent_index()
 
1398
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1399
 
 
1400
    def _generate_inventory(self):
 
1401
        """Create and set self.inventory from the dirstate object.
 
1402
 
 
1403
        (So this is only called the first time the inventory is requested for
 
1404
        this tree; it then remains in memory until it's out of date.)
 
1405
 
 
1406
        This is relatively expensive: we have to walk the entire dirstate.
 
1407
        """
 
1408
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1409
            'dirstate revision tree'
 
1410
        # separate call for profiling - makes it clear where the costs are.
 
1411
        self._dirstate._read_dirblocks_if_needed()
 
1412
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1413
            'parent %s has disappeared from %s' % (
 
1414
            self._revision_id, self._dirstate.get_parent_ids())
 
1415
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1416
        # This is identical now to the WorkingTree _generate_inventory except
 
1417
        # for the tree index use.
 
1418
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1419
        current_id = root_key[2]
 
1420
        assert current_entry[parent_index][0] == 'd'
 
1421
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1422
        inv.root.revision = current_entry[parent_index][4]
 
1423
        # Turn some things into local variables
 
1424
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1425
        factory = entry_factory
 
1426
        utf8_decode = cache_utf8._utf8_decode
 
1427
        inv_byid = inv._byid
 
1428
        # we could do this straight out of the dirstate; it might be fast
 
1429
        # and should be profiled - RBC 20070216
 
1430
        parent_ies = {'' : inv.root}
 
1431
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1432
            dirname = block[0]
 
1433
            try:
 
1434
                parent_ie = parent_ies[dirname]
 
1435
            except KeyError:
 
1436
                # all the paths in this block are not versioned in this tree
 
1437
                continue
 
1438
            for key, entry in block[1]:
 
1439
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1440
                if minikind in ('a', 'r'): # absent, relocated
 
1441
                    # not this tree
 
1442
                    continue
 
1443
                name = key[1]
 
1444
                name_unicode = utf8_decode(name)[0]
 
1445
                file_id = key[2]
 
1446
                kind = minikind_to_kind[minikind]
 
1447
                inv_entry = factory[kind](file_id, name_unicode,
 
1448
                                          parent_ie.file_id)
 
1449
                inv_entry.revision = revid
 
1450
                if kind == 'file':
 
1451
                    inv_entry.executable = executable
 
1452
                    inv_entry.text_size = size
 
1453
                    inv_entry.text_sha1 = fingerprint
 
1454
                elif kind == 'directory':
 
1455
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1456
                elif kind == 'symlink':
 
1457
                    inv_entry.executable = False
 
1458
                    inv_entry.text_size = None
 
1459
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1460
                elif kind == 'tree-reference':
 
1461
                    inv_entry.reference_revision = fingerprint or None
 
1462
                else:
 
1463
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1464
                            % entry)
 
1465
                # These checks cost us around 40ms on a 55k entry tree
 
1466
                assert file_id not in inv_byid
 
1467
                assert name_unicode not in parent_ie.children
 
1468
                inv_byid[file_id] = inv_entry
 
1469
                parent_ie.children[name_unicode] = inv_entry
 
1470
        self._inventory = inv
 
1471
 
 
1472
    def get_file_mtime(self, file_id, path=None):
 
1473
        """Return the modification time for this record.
 
1474
 
 
1475
        We return the timestamp of the last-changed revision.
 
1476
        """
 
1477
        # Make sure the file exists
 
1478
        entry = self._get_entry(file_id, path=path)
 
1479
        if entry == (None, None): # do we raise?
 
1480
            return None
 
1481
        parent_index = self._get_parent_index()
 
1482
        last_changed_revision = entry[1][parent_index][4]
 
1483
        return self._repository.get_revision(last_changed_revision).timestamp
 
1484
 
 
1485
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1486
        entry = self._get_entry(file_id=file_id, path=path)
 
1487
        parent_index = self._get_parent_index()
 
1488
        parent_details = entry[1][parent_index]
 
1489
        if parent_details[0] == 'f':
 
1490
            return parent_details[1]
 
1491
        return None
 
1492
 
 
1493
    def get_file(self, file_id):
 
1494
        return StringIO(self.get_file_text(file_id))
 
1495
 
 
1496
    def get_file_lines(self, file_id):
 
1497
        ie = self.inventory[file_id]
 
1498
        return self._repository.weave_store.get_weave(file_id,
 
1499
                self._repository.get_transaction()).get_lines(ie.revision)
 
1500
 
 
1501
    def get_file_size(self, file_id):
 
1502
        return self.inventory[file_id].text_size
 
1503
 
 
1504
    def get_file_text(self, file_id):
 
1505
        return ''.join(self.get_file_lines(file_id))
 
1506
 
 
1507
    def get_reference_revision(self, file_id, path=None):
 
1508
        return self.inventory[file_id].reference_revision
 
1509
 
 
1510
    def get_symlink_target(self, file_id):
 
1511
        entry = self._get_entry(file_id=file_id)
 
1512
        parent_index = self._get_parent_index()
 
1513
        if entry[1][parent_index][0] != 'l':
 
1514
            return None
 
1515
        else:
 
1516
            # At present, none of the tree implementations supports non-ascii
 
1517
            # symlink targets. So we will just assume that the dirstate path is
 
1518
            # correct.
 
1519
            return entry[1][parent_index][1]
 
1520
 
 
1521
    def get_revision_id(self):
 
1522
        """Return the revision id for this tree."""
 
1523
        return self._revision_id
 
1524
 
 
1525
    def _get_inventory(self):
 
1526
        if self._inventory is not None:
 
1527
            return self._inventory
 
1528
        self._must_be_locked()
 
1529
        self._generate_inventory()
 
1530
        return self._inventory
 
1531
 
 
1532
    inventory = property(_get_inventory,
 
1533
                         doc="Inventory of this Tree")
 
1534
 
 
1535
    def get_parent_ids(self):
 
1536
        """The parents of a tree in the dirstate are not cached."""
 
1537
        return self._repository.get_revision(self._revision_id).parent_ids
 
1538
 
 
1539
    def has_filename(self, filename):
 
1540
        return bool(self.path2id(filename))
 
1541
 
 
1542
    def kind(self, file_id):
 
1543
        return self.inventory[file_id].kind
 
1544
 
 
1545
    def is_executable(self, file_id, path=None):
 
1546
        ie = self.inventory[file_id]
 
1547
        if ie.kind != "file":
 
1548
            return None
 
1549
        return ie.executable
 
1550
 
 
1551
    def list_files(self, include_root=False):
 
1552
        # We use a standard implementation, because DirStateRevisionTree is
 
1553
        # dealing with one of the parents of the current state
 
1554
        inv = self._get_inventory()
 
1555
        entries = inv.iter_entries()
 
1556
        if self.inventory.root is not None and not include_root:
 
1557
            entries.next()
 
1558
        for path, entry in entries:
 
1559
            yield path, 'V', entry.kind, entry.file_id, entry
 
1560
 
 
1561
    def lock_read(self):
 
1562
        """Lock the tree for a set of operations."""
 
1563
        if not self._locked:
 
1564
            self._repository.lock_read()
 
1565
            if self._dirstate._lock_token is None:
 
1566
                self._dirstate.lock_read()
 
1567
                self._dirstate_locked = True
 
1568
        self._locked += 1
 
1569
 
 
1570
    def _must_be_locked(self):
 
1571
        if not self._locked:
 
1572
            raise errors.ObjectNotLocked(self)
 
1573
 
 
1574
    @needs_read_lock
 
1575
    def path2id(self, path):
 
1576
        """Return the id for path in this tree."""
 
1577
        # lookup by path: faster than splitting and walking the ivnentory.
 
1578
        entry = self._get_entry(path=path)
 
1579
        if entry == (None, None):
 
1580
            return None
 
1581
        return entry[0][2]
 
1582
 
 
1583
    def unlock(self):
 
1584
        """Unlock, freeing any cache memory used during the lock."""
 
1585
        # outside of a lock, the inventory is suspect: release it.
 
1586
        self._locked -=1
 
1587
        if not self._locked:
 
1588
            self._inventory = None
 
1589
            self._locked = 0
 
1590
            if self._dirstate_locked:
 
1591
                self._dirstate.unlock()
 
1592
                self._dirstate_locked = False
 
1593
            self._repository.unlock()
 
1594
 
 
1595
    def walkdirs(self, prefix=""):
 
1596
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1597
        # implementation based on an inventory.  
 
1598
        # This should be cleaned up to use the much faster Dirstate code
 
1599
        # So for now, we just build up the parent inventory, and extract
 
1600
        # it the same way RevisionTree does.
 
1601
        _directory = 'directory'
 
1602
        inv = self._get_inventory()
 
1603
        top_id = inv.path2id(prefix)
 
1604
        if top_id is None:
 
1605
            pending = []
 
1606
        else:
 
1607
            pending = [(prefix, top_id)]
 
1608
        while pending:
 
1609
            dirblock = []
 
1610
            relpath, file_id = pending.pop()
 
1611
            # 0 - relpath, 1- file-id
 
1612
            if relpath:
 
1613
                relroot = relpath + '/'
 
1614
            else:
 
1615
                relroot = ""
 
1616
            # FIXME: stash the node in pending
 
1617
            entry = inv[file_id]
 
1618
            for name, child in entry.sorted_children():
 
1619
                toppath = relroot + name
 
1620
                dirblock.append((toppath, name, child.kind, None,
 
1621
                    child.file_id, child.kind
 
1622
                    ))
 
1623
            yield (relpath, entry.file_id), dirblock
 
1624
            # push the user specified dirs from dirblock
 
1625
            for dir in reversed(dirblock):
 
1626
                if dir[2] == _directory:
 
1627
                    pending.append((dir[0], dir[4]))
 
1628
 
 
1629
 
 
1630
class InterDirStateTree(InterTree):
 
1631
    """Fast path optimiser for changes_from with dirstate trees.
 
1632
    
 
1633
    This is used only when both trees are in the dirstate working file, and 
 
1634
    the source is any parent within the dirstate, and the destination is 
 
1635
    the current working tree of the same dirstate.
 
1636
    """
 
1637
    # this could be generalized to allow comparisons between any trees in the
 
1638
    # dirstate, and possibly between trees stored in different dirstates.
 
1639
 
 
1640
    def __init__(self, source, target):
 
1641
        super(InterDirStateTree, self).__init__(source, target)
 
1642
        if not InterDirStateTree.is_compatible(source, target):
 
1643
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1644
 
 
1645
    @staticmethod
 
1646
    def make_source_parent_tree(source, target):
 
1647
        """Change the source tree into a parent of the target."""
 
1648
        revid = source.commit('record tree')
 
1649
        target.branch.repository.fetch(source.branch.repository, revid)
 
1650
        target.set_parent_ids([revid])
 
1651
        return target.basis_tree(), target
 
1652
 
 
1653
    _matching_from_tree_format = WorkingTreeFormat4()
 
1654
    _matching_to_tree_format = WorkingTreeFormat4()
 
1655
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1656
 
 
1657
    def _iter_changes(self, include_unchanged=False,
 
1658
                      specific_files=None, pb=None, extra_trees=[],
 
1659
                      require_versioned=True, want_unversioned=False):
 
1660
        """Return the changes from source to target.
 
1661
 
 
1662
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1663
            for details.
 
1664
        :param specific_files: An optional list of file paths to restrict the
 
1665
            comparison to. When mapping filenames to ids, all matches in all
 
1666
            trees (including optional extra_trees) are used, and all children of
 
1667
            matched directories are included.
 
1668
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1669
            unchanged entries in the result.
 
1670
        :param extra_trees: An optional list of additional trees to use when
 
1671
            mapping the contents of specific_files (paths) to file_ids.
 
1672
        :param require_versioned: If True, all files in specific_files must be
 
1673
            versioned in one of source, target, extra_trees or
 
1674
            PathsNotVersionedError is raised.
 
1675
        :param want_unversioned: Should unversioned files be returned in the
 
1676
            output. An unversioned file is defined as one with (False, False)
 
1677
            for the versioned pair.
 
1678
        """
 
1679
        utf8_decode = cache_utf8._utf8_decode
 
1680
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1681
        # NB: show_status depends on being able to pass in non-versioned files
 
1682
        # and report them as unknown
 
1683
        # TODO: handle extra trees in the dirstate.
 
1684
        # TODO: handle comparisons as an empty tree as a different special
 
1685
        # case? mbp 20070226
 
1686
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1687
            # we can't fast-path these cases (yet)
 
1688
            for f in super(InterDirStateTree, self)._iter_changes(
 
1689
                include_unchanged, specific_files, pb, extra_trees,
 
1690
                require_versioned, want_unversioned=want_unversioned):
 
1691
                yield f
 
1692
            return
 
1693
        parent_ids = self.target.get_parent_ids()
 
1694
        assert (self.source._revision_id in parent_ids), \
 
1695
                "revision {%s} is not stored in {%s}, but %s " \
 
1696
                "can only be used for trees stored in the dirstate" \
 
1697
                % (self.source._revision_id, self.target, self._iter_changes)
 
1698
        target_index = 0
 
1699
        if self.source._revision_id == NULL_REVISION:
 
1700
            source_index = None
 
1701
            indices = (target_index,)
 
1702
        else:
 
1703
            assert (self.source._revision_id in parent_ids), \
 
1704
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1705
                self.source._revision_id, parent_ids)
 
1706
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1707
            indices = (source_index,target_index)
 
1708
        # -- make all specific_files utf8 --
 
1709
        if specific_files:
 
1710
            specific_files_utf8 = set()
 
1711
            for path in specific_files:
 
1712
                specific_files_utf8.add(path.encode('utf8'))
 
1713
            specific_files = specific_files_utf8
 
1714
        else:
 
1715
            specific_files = set([''])
 
1716
        # -- specific_files is now a utf8 path set --
 
1717
        # -- get the state object and prepare it.
 
1718
        state = self.target.current_dirstate()
 
1719
        state._read_dirblocks_if_needed()
 
1720
        def _entries_for_path(path):
 
1721
            """Return a list with all the entries that match path for all ids.
 
1722
            """
 
1723
            dirname, basename = os.path.split(path)
 
1724
            key = (dirname, basename, '')
 
1725
            block_index, present = state._find_block_index_from_key(key)
 
1726
            if not present:
 
1727
                # the block which should contain path is absent.
 
1728
                return []
 
1729
            result = []
 
1730
            block = state._dirblocks[block_index][1]
 
1731
            entry_index, _ = state._find_entry_index(key, block)
 
1732
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1733
            while (entry_index < len(block) and
 
1734
                block[entry_index][0][0:2] == key[0:2]):
 
1735
                result.append(block[entry_index])
 
1736
                entry_index += 1
 
1737
            return result
 
1738
        if require_versioned:
 
1739
            # -- check all supplied paths are versioned in a search tree. --
 
1740
            all_versioned = True
 
1741
            for path in specific_files:
 
1742
                path_entries = _entries_for_path(path)
 
1743
                if not path_entries:
 
1744
                    # this specified path is not present at all: error
 
1745
                    all_versioned = False
 
1746
                    break
 
1747
                found_versioned = False
 
1748
                # for each id at this path
 
1749
                for entry in path_entries:
 
1750
                    # for each tree.
 
1751
                    for index in indices:
 
1752
                        if entry[1][index][0] != 'a': # absent
 
1753
                            found_versioned = True
 
1754
                            # all good: found a versioned cell
 
1755
                            break
 
1756
                if not found_versioned:
 
1757
                    # none of the indexes was not 'absent' at all ids for this
 
1758
                    # path.
 
1759
                    all_versioned = False
 
1760
                    break
 
1761
            if not all_versioned:
 
1762
                raise errors.PathsNotVersionedError(specific_files)
 
1763
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1764
        search_specific_files = set()
 
1765
        for path in specific_files:
 
1766
            other_specific_files = specific_files.difference(set([path]))
 
1767
            if not osutils.is_inside_any(other_specific_files, path):
 
1768
                # this is a top level path, we must check it.
 
1769
                search_specific_files.add(path)
 
1770
        # sketch: 
 
1771
        # compare source_index and target_index at or under each element of search_specific_files.
 
1772
        # follow the following comparison table. Note that we only want to do diff operations when
 
1773
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1774
        # for the target.
 
1775
        # cases:
 
1776
        # 
 
1777
        # Source | Target | disk | action
 
1778
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1779
        #        |        |      | diff check on source-target
 
1780
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1781
        #        |        |      | ???
 
1782
        #   r    |  a     |      | add source to search
 
1783
        #   r    |  a     |  a   | 
 
1784
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1785
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1786
        #   a    | fdlt   |      | add new id
 
1787
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1788
        #   a    |  a     |      | not present in either tree, skip
 
1789
        #   a    |  a     |  a   | not present in any tree, skip
 
1790
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1791
        #        |        |      | may not be selected by the users list of paths.
 
1792
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1793
        #        |        |      | may not be selected by the users list of paths.
 
1794
        #  fdlt  | fdlt   |      | content in both: diff them
 
1795
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1796
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1797
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1798
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1799
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1800
        #        |        |      | this id at the other path.
 
1801
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1802
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1803
        #        |        |      | this id at the other path.
 
1804
 
 
1805
        # for all search_indexs in each path at or under each element of
 
1806
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1807
        # relocated path as one to search if its not searched already. If the
 
1808
        # detail is not relocated, add the id.
 
1809
        searched_specific_files = set()
 
1810
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1811
        # Using a list so that we can access the values and change them in
 
1812
        # nested scope. Each one is [path, file_id, entry]
 
1813
        last_source_parent = [None, None]
 
1814
        last_target_parent = [None, None]
 
1815
 
 
1816
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1817
 
 
1818
        # Just a sentry, so that _process_entry can say that this
 
1819
        # record is handled, but isn't interesting to process (unchanged)
 
1820
        uninteresting = object()
 
1821
 
 
1822
 
 
1823
        old_dirname_to_file_id = {}
 
1824
        new_dirname_to_file_id = {}
 
1825
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1826
        #       keeping a cache of directories that we have seen.
 
1827
 
 
1828
        def _process_entry(entry, path_info):
 
1829
            """Compare an entry and real disk to generate delta information.
 
1830
 
 
1831
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1832
                the path of entry. If None, then the path is considered absent.
 
1833
                (Perhaps we should pass in a concrete entry for this ?)
 
1834
                Basename is returned as a utf8 string because we expect this
 
1835
                tuple will be ignored, and don't want to take the time to
 
1836
                decode.
 
1837
            :return: None if these don't match
 
1838
                     A tuple of information about the change, or
 
1839
                     the object 'uninteresting' if these match, but are
 
1840
                     basically identical.
 
1841
            """
 
1842
            if source_index is None:
 
1843
                source_details = NULL_PARENT_DETAILS
 
1844
            else:
 
1845
                source_details = entry[1][source_index]
 
1846
            target_details = entry[1][target_index]
 
1847
            target_minikind = target_details[0]
 
1848
            if path_info is not None and target_minikind in 'fdlt':
 
1849
                assert target_index == 0
 
1850
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1851
                                                  stat_value=path_info[3])
 
1852
                # The entry may have been modified by update_entry
 
1853
                target_details = entry[1][target_index]
 
1854
                target_minikind = target_details[0]
 
1855
            else:
 
1856
                link_or_sha1 = None
 
1857
            file_id = entry[0][2]
 
1858
            source_minikind = source_details[0]
 
1859
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1860
                # claimed content in both: diff
 
1861
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1862
                #        |        |      | diff check on source-target
 
1863
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1864
                #        |        |      | ???
 
1865
                if source_minikind in 'r':
 
1866
                    # add the source to the search path to find any children it
 
1867
                    # has.  TODO ? : only add if it is a container ?
 
1868
                    if not osutils.is_inside_any(searched_specific_files,
 
1869
                                                 source_details[1]):
 
1870
                        search_specific_files.add(source_details[1])
 
1871
                    # generate the old path; this is needed for stating later
 
1872
                    # as well.
 
1873
                    old_path = source_details[1]
 
1874
                    old_dirname, old_basename = os.path.split(old_path)
 
1875
                    path = pathjoin(entry[0][0], entry[0][1])
 
1876
                    old_entry = state._get_entry(source_index,
 
1877
                                                 path_utf8=old_path)
 
1878
                    # update the source details variable to be the real
 
1879
                    # location.
 
1880
                    source_details = old_entry[1][source_index]
 
1881
                    source_minikind = source_details[0]
 
1882
                else:
 
1883
                    old_dirname = entry[0][0]
 
1884
                    old_basename = entry[0][1]
 
1885
                    old_path = path = None
 
1886
                if path_info is None:
 
1887
                    # the file is missing on disk, show as removed.
 
1888
                    content_change = True
 
1889
                    target_kind = None
 
1890
                    target_exec = False
 
1891
                else:
 
1892
                    # source and target are both versioned and disk file is present.
 
1893
                    target_kind = path_info[2]
 
1894
                    if target_kind == 'directory':
 
1895
                        if path is None:
 
1896
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1897
                        new_dirname_to_file_id[path] = file_id
 
1898
                        if source_minikind != 'd':
 
1899
                            content_change = True
 
1900
                        else:
 
1901
                            # directories have no fingerprint
 
1902
                            content_change = False
 
1903
                        target_exec = False
 
1904
                    elif target_kind == 'file':
 
1905
                        if source_minikind != 'f':
 
1906
                            content_change = True
 
1907
                        else:
 
1908
                            # We could check the size, but we already have the
 
1909
                            # sha1 hash.
 
1910
                            content_change = (link_or_sha1 != source_details[1])
 
1911
                        # Target details is updated at update_entry time
 
1912
                        if use_filesystem_for_exec:
 
1913
                            # We don't need S_ISREG here, because we are sure
 
1914
                            # we are dealing with a file.
 
1915
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1916
                        else:
 
1917
                            target_exec = target_details[3]
 
1918
                    elif target_kind == 'symlink':
 
1919
                        if source_minikind != 'l':
 
1920
                            content_change = True
 
1921
                        else:
 
1922
                            content_change = (link_or_sha1 != source_details[1])
 
1923
                        target_exec = False
 
1924
                    elif target_kind == 'tree-reference':
 
1925
                        if source_minikind != 't':
 
1926
                            content_change = True
 
1927
                        else:
 
1928
                            content_change = False
 
1929
                        target_exec = False
 
1930
                    else:
 
1931
                        raise Exception, "unknown kind %s" % path_info[2]
 
1932
                if source_minikind == 'd':
 
1933
                    if path is None:
 
1934
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1935
                    old_dirname_to_file_id[old_path] = file_id
 
1936
                # parent id is the entry for the path in the target tree
 
1937
                if old_dirname == last_source_parent[0]:
 
1938
                    source_parent_id = last_source_parent[1]
 
1939
                else:
 
1940
                    try:
 
1941
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1942
                    except KeyError:
 
1943
                        source_parent_entry = state._get_entry(source_index,
 
1944
                                                               path_utf8=old_dirname)
 
1945
                        source_parent_id = source_parent_entry[0][2]
 
1946
                    if source_parent_id == entry[0][2]:
 
1947
                        # This is the root, so the parent is None
 
1948
                        source_parent_id = None
 
1949
                    else:
 
1950
                        last_source_parent[0] = old_dirname
 
1951
                        last_source_parent[1] = source_parent_id
 
1952
                new_dirname = entry[0][0]
 
1953
                if new_dirname == last_target_parent[0]:
 
1954
                    target_parent_id = last_target_parent[1]
 
1955
                else:
 
1956
                    try:
 
1957
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1958
                    except KeyError:
 
1959
                        # TODO: We don't always need to do the lookup, because the
 
1960
                        #       parent entry will be the same as the source entry.
 
1961
                        target_parent_entry = state._get_entry(target_index,
 
1962
                                                               path_utf8=new_dirname)
 
1963
                        assert target_parent_entry != (None, None), (
 
1964
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1965
                            % (new_dirname, entry))
 
1966
                        target_parent_id = target_parent_entry[0][2]
 
1967
                    if target_parent_id == entry[0][2]:
 
1968
                        # This is the root, so the parent is None
 
1969
                        target_parent_id = None
 
1970
                    else:
 
1971
                        last_target_parent[0] = new_dirname
 
1972
                        last_target_parent[1] = target_parent_id
 
1973
 
 
1974
                source_exec = source_details[3]
 
1975
                if (include_unchanged
 
1976
                    or content_change
 
1977
                    or source_parent_id != target_parent_id
 
1978
                    or old_basename != entry[0][1]
 
1979
                    or source_exec != target_exec
 
1980
                    ):
 
1981
                    if old_path is None:
 
1982
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1983
                        old_path_u = utf8_decode(old_path)[0]
 
1984
                        path_u = old_path_u
 
1985
                    else:
 
1986
                        old_path_u = utf8_decode(old_path)[0]
 
1987
                        if old_path == path:
 
1988
                            path_u = old_path_u
 
1989
                        else:
 
1990
                            path_u = utf8_decode(path)[0]
 
1991
                    source_kind = _minikind_to_kind[source_minikind]
 
1992
                    return (entry[0][2],
 
1993
                           (old_path_u, path_u),
 
1994
                           content_change,
 
1995
                           (True, True),
 
1996
                           (source_parent_id, target_parent_id),
 
1997
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
1998
                           (source_kind, target_kind),
 
1999
                           (source_exec, target_exec))
 
2000
                else:
 
2001
                    return uninteresting
 
2002
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2003
                # looks like a new file
 
2004
                if path_info is not None:
 
2005
                    path = pathjoin(entry[0][0], entry[0][1])
 
2006
                    # parent id is the entry for the path in the target tree
 
2007
                    # TODO: these are the same for an entire directory: cache em.
 
2008
                    parent_id = state._get_entry(target_index,
 
2009
                                                 path_utf8=entry[0][0])[0][2]
 
2010
                    if parent_id == entry[0][2]:
 
2011
                        parent_id = None
 
2012
                    if use_filesystem_for_exec:
 
2013
                        # We need S_ISREG here, because we aren't sure if this
 
2014
                        # is a file or not.
 
2015
                        target_exec = bool(
 
2016
                            stat.S_ISREG(path_info[3].st_mode)
 
2017
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2018
                    else:
 
2019
                        target_exec = target_details[3]
 
2020
                    return (entry[0][2],
 
2021
                           (None, utf8_decode(path)[0]),
 
2022
                           True,
 
2023
                           (False, True),
 
2024
                           (None, parent_id),
 
2025
                           (None, utf8_decode(entry[0][1])[0]),
 
2026
                           (None, path_info[2]),
 
2027
                           (None, target_exec))
 
2028
                else:
 
2029
                    # but its not on disk: we deliberately treat this as just
 
2030
                    # never-present. (Why ?! - RBC 20070224)
 
2031
                    pass
 
2032
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2033
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2034
                # if its still on disk, *and* theres no other entry at this
 
2035
                # path [we dont know this in this routine at the moment -
 
2036
                # perhaps we should change this - then it would be an unknown.
 
2037
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2038
                # parent id is the entry for the path in the target tree
 
2039
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2040
                if parent_id == entry[0][2]:
 
2041
                    parent_id = None
 
2042
                return (entry[0][2],
 
2043
                       (utf8_decode(old_path)[0], None),
 
2044
                       True,
 
2045
                       (True, False),
 
2046
                       (parent_id, None),
 
2047
                       (utf8_decode(entry[0][1])[0], None),
 
2048
                       (_minikind_to_kind[source_minikind], None),
 
2049
                       (source_details[3], None))
 
2050
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2051
                # a rename; could be a true rename, or a rename inherited from
 
2052
                # a renamed parent. TODO: handle this efficiently. Its not
 
2053
                # common case to rename dirs though, so a correct but slow
 
2054
                # implementation will do.
 
2055
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2056
                    search_specific_files.add(target_details[1])
 
2057
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2058
                # neither of the selected trees contain this file,
 
2059
                # so skip over it. This is not currently directly tested, but
 
2060
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2061
                pass
 
2062
            else:
 
2063
                raise AssertionError("don't know how to compare "
 
2064
                    "source_minikind=%r, target_minikind=%r"
 
2065
                    % (source_minikind, target_minikind))
 
2066
                ## import pdb;pdb.set_trace()
 
2067
            return None
 
2068
 
 
2069
        while search_specific_files:
 
2070
            # TODO: the pending list should be lexically sorted?  the
 
2071
            # interface doesn't require it.
 
2072
            current_root = search_specific_files.pop()
 
2073
            current_root_unicode = current_root.decode('utf8')
 
2074
            searched_specific_files.add(current_root)
 
2075
            # process the entries for this containing directory: the rest will be
 
2076
            # found by their parents recursively.
 
2077
            root_entries = _entries_for_path(current_root)
 
2078
            root_abspath = self.target.abspath(current_root_unicode)
 
2079
            try:
 
2080
                root_stat = os.lstat(root_abspath)
 
2081
            except OSError, e:
 
2082
                if e.errno == errno.ENOENT:
 
2083
                    # the path does not exist: let _process_entry know that.
 
2084
                    root_dir_info = None
 
2085
                else:
 
2086
                    # some other random error: hand it up.
 
2087
                    raise
 
2088
            else:
 
2089
                root_dir_info = ('', current_root,
 
2090
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2091
                    root_abspath)
 
2092
                if root_dir_info[2] == 'directory':
 
2093
                    if self.target._directory_is_tree_reference(
 
2094
                        current_root.decode('utf8')):
 
2095
                        root_dir_info = root_dir_info[:2] + \
 
2096
                            ('tree-reference',) + root_dir_info[3:]
 
2097
 
 
2098
            if not root_entries and not root_dir_info:
 
2099
                # this specified path is not present at all, skip it.
 
2100
                continue
 
2101
            path_handled = False
 
2102
            for entry in root_entries:
 
2103
                result = _process_entry(entry, root_dir_info)
 
2104
                if result is not None:
 
2105
                    path_handled = True
 
2106
                    if result is not uninteresting:
 
2107
                        yield result
 
2108
            if want_unversioned and not path_handled and root_dir_info:
 
2109
                new_executable = bool(
 
2110
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2111
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2112
                yield (None,
 
2113
                       (None, current_root_unicode),
 
2114
                       True,
 
2115
                       (False, False),
 
2116
                       (None, None),
 
2117
                       (None, splitpath(current_root_unicode)[-1]),
 
2118
                       (None, root_dir_info[2]),
 
2119
                       (None, new_executable)
 
2120
                      )
 
2121
            initial_key = (current_root, '', '')
 
2122
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2123
            if block_index == 0:
 
2124
                # we have processed the total root already, but because the
 
2125
                # initial key matched it we should skip it here.
 
2126
                block_index +=1
 
2127
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2128
                current_dir_info = None
 
2129
            else:
 
2130
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2131
                try:
 
2132
                    current_dir_info = dir_iterator.next()
 
2133
                except OSError, e:
 
2134
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2135
                    # python 2.5 has e.errno == EINVAL,
 
2136
                    #            and e.winerror == ERROR_DIRECTORY
 
2137
                    e_winerror = getattr(e, 'winerror', None)
 
2138
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2139
                    # there may be directories in the inventory even though
 
2140
                    # this path is not a file on disk: so mark it as end of
 
2141
                    # iterator
 
2142
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2143
                        current_dir_info = None
 
2144
                    elif (sys.platform == 'win32'
 
2145
                          and (e.errno in win_errors
 
2146
                               or e_winerror in win_errors)):
 
2147
                        current_dir_info = None
 
2148
                    else:
 
2149
                        raise
 
2150
                else:
 
2151
                    if current_dir_info[0][0] == '':
 
2152
                        # remove .bzr from iteration
 
2153
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2154
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2155
                        del current_dir_info[1][bzr_index]
 
2156
            # walk until both the directory listing and the versioned metadata
 
2157
            # are exhausted. 
 
2158
            if (block_index < len(state._dirblocks) and
 
2159
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2160
                current_block = state._dirblocks[block_index]
 
2161
            else:
 
2162
                current_block = None
 
2163
            while (current_dir_info is not None or
 
2164
                   current_block is not None):
 
2165
                if (current_dir_info and current_block
 
2166
                    and current_dir_info[0][0] != current_block[0]):
 
2167
                    if current_dir_info[0][0].split('/') < current_block[0].split('/'):
 
2168
                        # filesystem data refers to paths not covered by the dirblock.
 
2169
                        # this has two possibilities:
 
2170
                        # A) it is versioned but empty, so there is no block for it
 
2171
                        # B) it is not versioned.
 
2172
 
 
2173
                        # if (A) then we need to recurse into it to check for
 
2174
                        # new unknown files or directories.
 
2175
                        # if (B) then we should ignore it, because we don't
 
2176
                        # recurse into unknown directories.
 
2177
                        path_index = 0
 
2178
                        while path_index < len(current_dir_info[1]):
 
2179
                                current_path_info = current_dir_info[1][path_index]
 
2180
                                if want_unversioned:
 
2181
                                    if current_path_info[2] == 'directory':
 
2182
                                        if self.target._directory_is_tree_reference(
 
2183
                                            current_path_info[0].decode('utf8')):
 
2184
                                            current_path_info = current_path_info[:2] + \
 
2185
                                                ('tree-reference',) + current_path_info[3:]
 
2186
                                    new_executable = bool(
 
2187
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2188
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2189
                                    yield (None,
 
2190
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2191
                                        True,
 
2192
                                        (False, False),
 
2193
                                        (None, None),
 
2194
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2195
                                        (None, current_path_info[2]),
 
2196
                                        (None, new_executable))
 
2197
                                # dont descend into this unversioned path if it is
 
2198
                                # a dir
 
2199
                                if current_path_info[2] in ('directory',
 
2200
                                                            'tree-reference'):
 
2201
                                    del current_dir_info[1][path_index]
 
2202
                                    path_index -= 1
 
2203
                                path_index += 1
 
2204
 
 
2205
                        # This dir info has been handled, go to the next
 
2206
                        try:
 
2207
                            current_dir_info = dir_iterator.next()
 
2208
                        except StopIteration:
 
2209
                            current_dir_info = None
 
2210
                    else:
 
2211
                        # We have a dirblock entry for this location, but there
 
2212
                        # is no filesystem path for this. This is most likely
 
2213
                        # because a directory was removed from the disk.
 
2214
                        # We don't have to report the missing directory,
 
2215
                        # because that should have already been handled, but we
 
2216
                        # need to handle all of the files that are contained
 
2217
                        # within.
 
2218
                        for current_entry in current_block[1]:
 
2219
                            # entry referring to file not present on disk.
 
2220
                            # advance the entry only, after processing.
 
2221
                            result = _process_entry(current_entry, None)
 
2222
                            if result is not None:
 
2223
                                if result is not uninteresting:
 
2224
                                    yield result
 
2225
                        block_index +=1
 
2226
                        if (block_index < len(state._dirblocks) and
 
2227
                            osutils.is_inside(current_root,
 
2228
                                              state._dirblocks[block_index][0])):
 
2229
                            current_block = state._dirblocks[block_index]
 
2230
                        else:
 
2231
                            current_block = None
 
2232
                    continue
 
2233
                entry_index = 0
 
2234
                if current_block and entry_index < len(current_block[1]):
 
2235
                    current_entry = current_block[1][entry_index]
 
2236
                else:
 
2237
                    current_entry = None
 
2238
                advance_entry = True
 
2239
                path_index = 0
 
2240
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2241
                    current_path_info = current_dir_info[1][path_index]
 
2242
                    if current_path_info[2] == 'directory':
 
2243
                        if self.target._directory_is_tree_reference(
 
2244
                            current_path_info[0].decode('utf8')):
 
2245
                            current_path_info = current_path_info[:2] + \
 
2246
                                ('tree-reference',) + current_path_info[3:]
 
2247
                else:
 
2248
                    current_path_info = None
 
2249
                advance_path = True
 
2250
                path_handled = False
 
2251
                while (current_entry is not None or
 
2252
                    current_path_info is not None):
 
2253
                    if current_entry is None:
 
2254
                        # the check for path_handled when the path is adnvaced
 
2255
                        # will yield this path if needed.
 
2256
                        pass
 
2257
                    elif current_path_info is None:
 
2258
                        # no path is fine: the per entry code will handle it.
 
2259
                        result = _process_entry(current_entry, current_path_info)
 
2260
                        if result is not None:
 
2261
                            if result is not uninteresting:
 
2262
                                yield result
 
2263
                    elif (current_entry[0][1] != current_path_info[1]
 
2264
                          or current_entry[1][target_index][0] in 'ar'):
 
2265
                        # The current path on disk doesn't match the dirblock
 
2266
                        # record. Either the dirblock is marked as absent, or
 
2267
                        # the file on disk is not present at all in the
 
2268
                        # dirblock. Either way, report about the dirblock
 
2269
                        # entry, and let other code handle the filesystem one.
 
2270
 
 
2271
                        # Compare the basename for these files to determine
 
2272
                        # which comes first
 
2273
                        if current_path_info[1] < current_entry[0][1]:
 
2274
                            # extra file on disk: pass for now, but only
 
2275
                            # increment the path, not the entry
 
2276
                            advance_entry = False
 
2277
                        else:
 
2278
                            # entry referring to file not present on disk.
 
2279
                            # advance the entry only, after processing.
 
2280
                            result = _process_entry(current_entry, None)
 
2281
                            if result is not None:
 
2282
                                if result is not uninteresting:
 
2283
                                    yield result
 
2284
                            advance_path = False
 
2285
                    else:
 
2286
                        result = _process_entry(current_entry, current_path_info)
 
2287
                        if result is not None:
 
2288
                            path_handled = True
 
2289
                            if result is not uninteresting:
 
2290
                                yield result
 
2291
                    if advance_entry and current_entry is not None:
 
2292
                        entry_index += 1
 
2293
                        if entry_index < len(current_block[1]):
 
2294
                            current_entry = current_block[1][entry_index]
 
2295
                        else:
 
2296
                            current_entry = None
 
2297
                    else:
 
2298
                        advance_entry = True # reset the advance flaga
 
2299
                    if advance_path and current_path_info is not None:
 
2300
                        if not path_handled:
 
2301
                            # unversioned in all regards
 
2302
                            if want_unversioned:
 
2303
                                new_executable = bool(
 
2304
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2305
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2306
                                yield (None,
 
2307
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2308
                                    True,
 
2309
                                    (False, False),
 
2310
                                    (None, None),
 
2311
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2312
                                    (None, current_path_info[2]),
 
2313
                                    (None, new_executable))
 
2314
                            # dont descend into this unversioned path if it is
 
2315
                            # a dir
 
2316
                            if current_path_info[2] in ('directory'):
 
2317
                                del current_dir_info[1][path_index]
 
2318
                                path_index -= 1
 
2319
                        # dont descend the disk iterator into any tree 
 
2320
                        # paths.
 
2321
                        if current_path_info[2] == 'tree-reference':
 
2322
                            del current_dir_info[1][path_index]
 
2323
                            path_index -= 1
 
2324
                        path_index += 1
 
2325
                        if path_index < len(current_dir_info[1]):
 
2326
                            current_path_info = current_dir_info[1][path_index]
 
2327
                            if current_path_info[2] == 'directory':
 
2328
                                if self.target._directory_is_tree_reference(
 
2329
                                    current_path_info[0].decode('utf8')):
 
2330
                                    current_path_info = current_path_info[:2] + \
 
2331
                                        ('tree-reference',) + current_path_info[3:]
 
2332
                        else:
 
2333
                            current_path_info = None
 
2334
                        path_handled = False
 
2335
                    else:
 
2336
                        advance_path = True # reset the advance flagg.
 
2337
                if current_block is not None:
 
2338
                    block_index += 1
 
2339
                    if (block_index < len(state._dirblocks) and
 
2340
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2341
                        current_block = state._dirblocks[block_index]
 
2342
                    else:
 
2343
                        current_block = None
 
2344
                if current_dir_info is not None:
 
2345
                    try:
 
2346
                        current_dir_info = dir_iterator.next()
 
2347
                    except StopIteration:
 
2348
                        current_dir_info = None
 
2349
 
 
2350
 
 
2351
    @staticmethod
 
2352
    def is_compatible(source, target):
 
2353
        # the target must be a dirstate working tree
 
2354
        if not isinstance(target, WorkingTree4):
 
2355
            return False
 
2356
        # the source must be a revtreee or dirstate rev tree.
 
2357
        if not isinstance(source,
 
2358
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2359
            return False
 
2360
        # the source revid must be in the target dirstate
 
2361
        if not (source._revision_id == NULL_REVISION or
 
2362
            source._revision_id in target.get_parent_ids()):
 
2363
            # TODO: what about ghosts? it may well need to 
 
2364
            # check for them explicitly.
 
2365
            return False
 
2366
        return True
 
2367
 
 
2368
InterTree.register_optimiser(InterDirStateTree)
 
2369
 
 
2370
 
 
2371
class Converter3to4(object):
 
2372
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2373
 
 
2374
    def __init__(self):
 
2375
        self.target_format = WorkingTreeFormat4()
 
2376
 
 
2377
    def convert(self, tree):
 
2378
        # lock the control files not the tree, so that we dont get tree
 
2379
        # on-unlock behaviours, and so that noone else diddles with the 
 
2380
        # tree during upgrade.
 
2381
        tree._control_files.lock_write()
 
2382
        try:
 
2383
            tree.read_working_inventory()
 
2384
            self.create_dirstate_data(tree)
 
2385
            self.update_format(tree)
 
2386
            self.remove_xml_files(tree)
 
2387
        finally:
 
2388
            tree._control_files.unlock()
 
2389
 
 
2390
    def create_dirstate_data(self, tree):
 
2391
        """Create the dirstate based data for tree."""
 
2392
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2393
            ).local_abspath('dirstate')
 
2394
        state = dirstate.DirState.from_tree(tree, local_path)
 
2395
        state.save()
 
2396
        state.unlock()
 
2397
 
 
2398
    def remove_xml_files(self, tree):
 
2399
        """Remove the oldformat 3 data."""
 
2400
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2401
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2402
            'pending-merges', 'stat-cache']:
 
2403
            try:
 
2404
                transport.delete(path)
 
2405
            except errors.NoSuchFile:
 
2406
                # some files are optional - just deal.
 
2407
                pass
 
2408
 
 
2409
    def update_format(self, tree):
 
2410
        """Change the format marker."""
 
2411
        tree._control_files.put_utf8('format',
 
2412
            self.target_format.get_format_string())