~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2008-06-20 01:09:18 UTC
  • mfrom: (3505.1.1 ianc-integration)
  • Revision ID: pqm@pqm.ubuntu.com-20080620010918-64z4xylh1ap5hgyf
Accept user names with @s in URLs (Neil Martinsen-Burrell)

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    debug,
 
47
    delta,
 
48
    dirstate,
 
49
    errors,
 
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    trace,
 
59
    transform,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
 
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
 
67
""")
 
68
 
 
69
from bzrlib import symbol_versioning
 
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
 
76
from bzrlib.osutils import (
 
77
    file_kind,
 
78
    isdir,
 
79
    normpath,
 
80
    pathjoin,
 
81
    rand_chars,
 
82
    realpath,
 
83
    safe_unicode,
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
 
87
from bzrlib.transport.local import LocalTransport
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
 
120
    def __init__(self, basedir,
 
121
                 branch,
 
122
                 _control_files=None,
 
123
                 _format=None,
 
124
                 _bzrdir=None):
 
125
        """Construct a WorkingTree for basedir.
 
126
 
 
127
        If the branch is not supplied, it is opened automatically.
 
128
        If the branch is supplied, it must be the branch for this basedir.
 
129
        (branch.base is not cross checked, because for remote branches that
 
130
        would be meaningless).
 
131
        """
 
132
        self._format = _format
 
133
        self.bzrdir = _bzrdir
 
134
        basedir = safe_unicode(basedir)
 
135
        mutter("opening working tree %r", basedir)
 
136
        self._branch = branch
 
137
        self.basedir = realpath(basedir)
 
138
        # if branch is at our basedir and is a format 6 or less
 
139
        # assume all other formats have their own control files.
 
140
        self._control_files = _control_files
 
141
        self._transport = self._control_files._transport
 
142
        self._dirty = None
 
143
        #-------------
 
144
        # during a read or write lock these objects are set, and are
 
145
        # None the rest of the time.
 
146
        self._dirstate = None
 
147
        self._inventory = None
 
148
        #-------------
 
149
        self._setup_directory_is_tree_reference()
 
150
        self._detect_case_handling()
 
151
 
 
152
    @needs_tree_write_lock
 
153
    def _add(self, files, ids, kinds):
 
154
        """See MutableTree._add."""
 
155
        state = self.current_dirstate()
 
156
        for f, file_id, kind in zip(files, ids, kinds):
 
157
            f = f.strip('/')
 
158
            if self.path2id(f):
 
159
                # special case tree root handling.
 
160
                if f == '' and self.path2id(f) == ROOT_ID:
 
161
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
162
                continue
 
163
            if file_id is None:
 
164
                file_id = generate_ids.gen_file_id(f)
 
165
            # deliberately add the file with no cached stat or sha1
 
166
            # - on the first access it will be gathered, and we can
 
167
            # always change this once tests are all passing.
 
168
            state.add(f, file_id, kind, None, '')
 
169
        self._make_dirty(reset_inventory=True)
 
170
 
 
171
    def _make_dirty(self, reset_inventory):
 
172
        """Make the tree state dirty.
 
173
 
 
174
        :param reset_inventory: True if the cached inventory should be removed
 
175
            (presuming there is one).
 
176
        """
 
177
        self._dirty = True
 
178
        if reset_inventory and self._inventory is not None:
 
179
            self._inventory = None
 
180
 
 
181
    @needs_tree_write_lock
 
182
    def add_reference(self, sub_tree):
 
183
        # use standard implementation, which calls back to self._add
 
184
        # 
 
185
        # So we don't store the reference_revision in the working dirstate,
 
186
        # it's just recorded at the moment of commit. 
 
187
        self._add_reference(sub_tree)
 
188
 
 
189
    def break_lock(self):
 
190
        """Break a lock if one is present from another instance.
 
191
 
 
192
        Uses the ui factory to ask for confirmation if the lock may be from
 
193
        an active process.
 
194
 
 
195
        This will probe the repository for its lock as well.
 
196
        """
 
197
        # if the dirstate is locked by an active process, reject the break lock
 
198
        # call.
 
199
        try:
 
200
            if self._dirstate is None:
 
201
                clear = True
 
202
            else:
 
203
                clear = False
 
204
            state = self._current_dirstate()
 
205
            if state._lock_token is not None:
 
206
                # we already have it locked. sheese, cant break our own lock.
 
207
                raise errors.LockActive(self.basedir)
 
208
            else:
 
209
                try:
 
210
                    # try for a write lock - need permission to get one anyhow
 
211
                    # to break locks.
 
212
                    state.lock_write()
 
213
                except errors.LockContention:
 
214
                    # oslocks fail when a process is still live: fail.
 
215
                    # TODO: get the locked lockdir info and give to the user to
 
216
                    # assist in debugging.
 
217
                    raise errors.LockActive(self.basedir)
 
218
                else:
 
219
                    state.unlock()
 
220
        finally:
 
221
            if clear:
 
222
                self._dirstate = None
 
223
        self._control_files.break_lock()
 
224
        self.branch.break_lock()
 
225
 
 
226
    def _comparison_data(self, entry, path):
 
227
        kind, executable, stat_value = \
 
228
            WorkingTree3._comparison_data(self, entry, path)
 
229
        # it looks like a plain directory, but it's really a reference -- see
 
230
        # also kind()
 
231
        if (self._repo_supports_tree_reference and
 
232
            kind == 'directory' and
 
233
            self._directory_is_tree_reference(path)):
 
234
            kind = 'tree-reference'
 
235
        return kind, executable, stat_value
 
236
 
 
237
    @needs_write_lock
 
238
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
239
        # mark the tree as dirty post commit - commit
 
240
        # can change the current versioned list by doing deletes.
 
241
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
242
        self._make_dirty(reset_inventory=True)
 
243
        return result
 
244
 
 
245
    def current_dirstate(self):
 
246
        """Return the current dirstate object.
 
247
 
 
248
        This is not part of the tree interface and only exposed for ease of
 
249
        testing.
 
250
 
 
251
        :raises errors.NotWriteLocked: when not in a lock.
 
252
        """
 
253
        self._must_be_locked()
 
254
        return self._current_dirstate()
 
255
 
 
256
    def _current_dirstate(self):
 
257
        """Internal function that does not check lock status.
 
258
 
 
259
        This is needed for break_lock which also needs the dirstate.
 
260
        """
 
261
        if self._dirstate is not None:
 
262
            return self._dirstate
 
263
        local_path = self.bzrdir.get_workingtree_transport(None
 
264
            ).local_abspath('dirstate')
 
265
        self._dirstate = dirstate.DirState.on_file(local_path)
 
266
        return self._dirstate
 
267
 
 
268
    def filter_unversioned_files(self, paths):
 
269
        """Filter out paths that are versioned.
 
270
 
 
271
        :return: set of paths.
 
272
        """
 
273
        # TODO: make a generic multi-bisect routine roughly that should list
 
274
        # the paths, then process one half at a time recursively, and feed the
 
275
        # results of each bisect in further still
 
276
        paths = sorted(paths)
 
277
        result = set()
 
278
        state = self.current_dirstate()
 
279
        # TODO we want a paths_to_dirblocks helper I think
 
280
        for path in paths:
 
281
            dirname, basename = os.path.split(path.encode('utf8'))
 
282
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
283
                dirname, basename, 0)
 
284
            if not path_is_versioned:
 
285
                result.add(path)
 
286
        return result
 
287
 
 
288
    def flush(self):
 
289
        """Write all cached data to disk."""
 
290
        if self._control_files._lock_mode != 'w':
 
291
            raise errors.NotWriteLocked(self)
 
292
        self.current_dirstate().save()
 
293
        self._inventory = None
 
294
        self._dirty = False
 
295
 
 
296
    @needs_tree_write_lock
 
297
    def _gather_kinds(self, files, kinds):
 
298
        """See MutableTree._gather_kinds."""
 
299
        for pos, f in enumerate(files):
 
300
            if kinds[pos] is None:
 
301
                kinds[pos] = self._kind(f)
 
302
 
 
303
    def _generate_inventory(self):
 
304
        """Create and set self.inventory from the dirstate object.
 
305
        
 
306
        This is relatively expensive: we have to walk the entire dirstate.
 
307
        Ideally we would not, and can deprecate this function.
 
308
        """
 
309
        #: uncomment to trap on inventory requests.
 
310
        # import pdb;pdb.set_trace()
 
311
        state = self.current_dirstate()
 
312
        state._read_dirblocks_if_needed()
 
313
        root_key, current_entry = self._get_entry(path='')
 
314
        current_id = root_key[2]
 
315
        if not (current_entry[0][0] == 'd'): # directory
 
316
            raise AssertionError(current_entry)
 
317
        inv = Inventory(root_id=current_id)
 
318
        # Turn some things into local variables
 
319
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
320
        factory = entry_factory
 
321
        utf8_decode = cache_utf8._utf8_decode
 
322
        inv_byid = inv._byid
 
323
        # we could do this straight out of the dirstate; it might be fast
 
324
        # and should be profiled - RBC 20070216
 
325
        parent_ies = {'' : inv.root}
 
326
        for block in state._dirblocks[1:]: # skip the root
 
327
            dirname = block[0]
 
328
            try:
 
329
                parent_ie = parent_ies[dirname]
 
330
            except KeyError:
 
331
                # all the paths in this block are not versioned in this tree
 
332
                continue
 
333
            for key, entry in block[1]:
 
334
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
335
                if minikind in ('a', 'r'): # absent, relocated
 
336
                    # a parent tree only entry
 
337
                    continue
 
338
                name = key[1]
 
339
                name_unicode = utf8_decode(name)[0]
 
340
                file_id = key[2]
 
341
                kind = minikind_to_kind[minikind]
 
342
                inv_entry = factory[kind](file_id, name_unicode,
 
343
                                          parent_ie.file_id)
 
344
                if kind == 'file':
 
345
                    # This is only needed on win32, where this is the only way
 
346
                    # we know the executable bit.
 
347
                    inv_entry.executable = executable
 
348
                    # not strictly needed: working tree
 
349
                    #inv_entry.text_size = size
 
350
                    #inv_entry.text_sha1 = sha1
 
351
                elif kind == 'directory':
 
352
                    # add this entry to the parent map.
 
353
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
354
                elif kind == 'tree-reference':
 
355
                    if not self._repo_supports_tree_reference:
 
356
                        raise AssertionError(
 
357
                            "repository of %r "
 
358
                            "doesn't support tree references "
 
359
                            "required by entry %r"
 
360
                            % (self, name))
 
361
                    inv_entry.reference_revision = link_or_sha1 or None
 
362
                elif kind != 'symlink':
 
363
                    raise AssertionError("unknown kind %r" % kind)
 
364
                # These checks cost us around 40ms on a 55k entry tree
 
365
                if file_id in inv_byid:
 
366
                    raise AssertionError('file_id %s already in'
 
367
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                if name_unicode in parent_ie.children:
 
369
                    raise AssertionError('name %r already in parent'
 
370
                        % (name_unicode,))
 
371
                inv_byid[file_id] = inv_entry
 
372
                parent_ie.children[name_unicode] = inv_entry
 
373
        self._inventory = inv
 
374
 
 
375
    def _get_entry(self, file_id=None, path=None):
 
376
        """Get the dirstate row for file_id or path.
 
377
 
 
378
        If either file_id or path is supplied, it is used as the key to lookup.
 
379
        If both are supplied, the fastest lookup is used, and an error is
 
380
        raised if they do not both point at the same row.
 
381
        
 
382
        :param file_id: An optional unicode file_id to be looked up.
 
383
        :param path: An optional unicode path to be looked up.
 
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
385
        """
 
386
        if file_id is None and path is None:
 
387
            raise errors.BzrError('must supply file_id or path')
 
388
        state = self.current_dirstate()
 
389
        if path is not None:
 
390
            path = path.encode('utf8')
 
391
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
392
 
 
393
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
394
        # check file id is valid unconditionally.
 
395
        entry = self._get_entry(file_id=file_id, path=path)
 
396
        if entry[0] is None:
 
397
            raise errors.NoSuchId(self, file_id)
 
398
        if path is None:
 
399
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
400
 
 
401
        file_abspath = self.abspath(path)
 
402
        state = self.current_dirstate()
 
403
        if stat_value is None:
 
404
            try:
 
405
                stat_value = os.lstat(file_abspath)
 
406
            except OSError, e:
 
407
                if e.errno == errno.ENOENT:
 
408
                    return None
 
409
                else:
 
410
                    raise
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
 
413
        if entry[1][0][0] == 'f':
 
414
            return link_or_sha1
 
415
        return None
 
416
 
 
417
    def _get_inventory(self):
 
418
        """Get the inventory for the tree. This is only valid within a lock."""
 
419
        if 'evil' in debug.debug_flags:
 
420
            trace.mutter_callsite(2,
 
421
                "accessing .inventory forces a size of tree translation.")
 
422
        if self._inventory is not None:
 
423
            return self._inventory
 
424
        self._must_be_locked()
 
425
        self._generate_inventory()
 
426
        return self._inventory
 
427
 
 
428
    inventory = property(_get_inventory,
 
429
                         doc="Inventory of this Tree")
 
430
 
 
431
    @needs_read_lock
 
432
    def get_parent_ids(self):
 
433
        """See Tree.get_parent_ids.
 
434
        
 
435
        This implementation requests the ids list from the dirstate file.
 
436
        """
 
437
        return self.current_dirstate().get_parent_ids()
 
438
 
 
439
    def get_reference_revision(self, file_id, path=None):
 
440
        # referenced tree's revision is whatever's currently there
 
441
        return self.get_nested_tree(file_id, path).last_revision()
 
442
 
 
443
    def get_nested_tree(self, file_id, path=None):
 
444
        if path is None:
 
445
            path = self.id2path(file_id)
 
446
        # else: check file_id is at path?
 
447
        return WorkingTree.open(self.abspath(path))
 
448
 
 
449
    @needs_read_lock
 
450
    def get_root_id(self):
 
451
        """Return the id of this trees root"""
 
452
        return self._get_entry(path='')[0][2]
 
453
 
 
454
    def has_id(self, file_id):
 
455
        state = self.current_dirstate()
 
456
        row, parents = self._get_entry(file_id=file_id)
 
457
        if row is None:
 
458
            return False
 
459
        return osutils.lexists(pathjoin(
 
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
461
 
 
462
    @needs_read_lock
 
463
    def id2path(self, file_id):
 
464
        "Convert a file-id to a path."
 
465
        state = self.current_dirstate()
 
466
        entry = self._get_entry(file_id=file_id)
 
467
        if entry == (None, None):
 
468
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
469
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
470
        return path_utf8.decode('utf8')
 
471
 
 
472
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
473
        entry = self._get_entry(path=path)
 
474
        if entry == (None, None):
 
475
            return False # Missing entries are not executable
 
476
        return entry[1][0][3] # Executable?
 
477
 
 
478
    if not osutils.supports_executable():
 
479
        def is_executable(self, file_id, path=None):
 
480
            """Test if a file is executable or not.
 
481
 
 
482
            Note: The caller is expected to take a read-lock before calling this.
 
483
            """
 
484
            entry = self._get_entry(file_id=file_id, path=path)
 
485
            if entry == (None, None):
 
486
                return False
 
487
            return entry[1][0][3]
 
488
 
 
489
        _is_executable_from_path_and_stat = \
 
490
            _is_executable_from_path_and_stat_from_basis
 
491
    else:
 
492
        def is_executable(self, file_id, path=None):
 
493
            """Test if a file is executable or not.
 
494
 
 
495
            Note: The caller is expected to take a read-lock before calling this.
 
496
            """
 
497
            self._must_be_locked()
 
498
            if not path:
 
499
                path = self.id2path(file_id)
 
500
            mode = os.lstat(self.abspath(path)).st_mode
 
501
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
502
 
 
503
    def all_file_ids(self):
 
504
        """See Tree.iter_all_file_ids"""
 
505
        self._must_be_locked()
 
506
        result = set()
 
507
        for key, tree_details in self.current_dirstate()._iter_entries():
 
508
            if tree_details[0][0] in ('a', 'r'): # relocated
 
509
                continue
 
510
            result.add(key[2])
 
511
        return result
 
512
 
 
513
    @needs_read_lock
 
514
    def __iter__(self):
 
515
        """Iterate through file_ids for this tree.
 
516
 
 
517
        file_ids are in a WorkingTree if they are in the working inventory
 
518
        and the working file exists.
 
519
        """
 
520
        result = []
 
521
        for key, tree_details in self.current_dirstate()._iter_entries():
 
522
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
523
                # not relevant to the working tree
 
524
                continue
 
525
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
526
            if osutils.lexists(path):
 
527
                result.append(key[2])
 
528
        return iter(result)
 
529
 
 
530
    def iter_references(self):
 
531
        for key, tree_details in self.current_dirstate()._iter_entries():
 
532
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
533
                # not relevant to the working tree
 
534
                continue
 
535
            if not key[1]:
 
536
                # the root is not a reference.
 
537
                continue
 
538
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
539
            try:
 
540
                if self._kind(path) == 'tree-reference':
 
541
                    yield path, key[2]
 
542
            except errors.NoSuchFile:
 
543
                # path is missing on disk.
 
544
                continue
 
545
 
 
546
    def kind(self, file_id):
 
547
        """Return the kind of a file.
 
548
 
 
549
        This is always the actual kind that's on disk, regardless of what it
 
550
        was added as.
 
551
 
 
552
        Note: The caller is expected to take a read-lock before calling this.
 
553
        """
 
554
        relpath = self.id2path(file_id)
 
555
        if relpath is None:
 
556
            raise AssertionError(
 
557
                "path for id {%s} is None!" % file_id)
 
558
        return self._kind(relpath)
 
559
 
 
560
    def _kind(self, relpath):
 
561
        abspath = self.abspath(relpath)
 
562
        kind = file_kind(abspath)
 
563
        if (self._repo_supports_tree_reference and
 
564
            kind == 'directory' and
 
565
            self._directory_is_tree_reference(relpath)):
 
566
            kind = 'tree-reference'
 
567
        return kind
 
568
 
 
569
    @needs_read_lock
 
570
    def _last_revision(self):
 
571
        """See Mutable.last_revision."""
 
572
        parent_ids = self.current_dirstate().get_parent_ids()
 
573
        if parent_ids:
 
574
            return parent_ids[0]
 
575
        else:
 
576
            return _mod_revision.NULL_REVISION
 
577
 
 
578
    def lock_read(self):
 
579
        """See Branch.lock_read, and WorkingTree.unlock."""
 
580
        self.branch.lock_read()
 
581
        try:
 
582
            self._control_files.lock_read()
 
583
            try:
 
584
                state = self.current_dirstate()
 
585
                if not state._lock_token:
 
586
                    state.lock_read()
 
587
                # set our support for tree references from the repository in
 
588
                # use.
 
589
                self._repo_supports_tree_reference = getattr(
 
590
                    self.branch.repository._format, "supports_tree_reference",
 
591
                    False)
 
592
            except:
 
593
                self._control_files.unlock()
 
594
                raise
 
595
        except:
 
596
            self.branch.unlock()
 
597
            raise
 
598
 
 
599
    def _lock_self_write(self):
 
600
        """This should be called after the branch is locked."""
 
601
        try:
 
602
            self._control_files.lock_write()
 
603
            try:
 
604
                state = self.current_dirstate()
 
605
                if not state._lock_token:
 
606
                    state.lock_write()
 
607
                # set our support for tree references from the repository in
 
608
                # use.
 
609
                self._repo_supports_tree_reference = getattr(
 
610
                    self.branch.repository._format, "supports_tree_reference",
 
611
                    False)
 
612
            except:
 
613
                self._control_files.unlock()
 
614
                raise
 
615
        except:
 
616
            self.branch.unlock()
 
617
            raise
 
618
 
 
619
    def lock_tree_write(self):
 
620
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
621
        self.branch.lock_read()
 
622
        self._lock_self_write()
 
623
 
 
624
    def lock_write(self):
 
625
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
626
        self.branch.lock_write()
 
627
        self._lock_self_write()
 
628
 
 
629
    @needs_tree_write_lock
 
630
    def move(self, from_paths, to_dir, after=False):
 
631
        """See WorkingTree.move()."""
 
632
        result = []
 
633
        if not from_paths:
 
634
            return result
 
635
        state = self.current_dirstate()
 
636
        if isinstance(from_paths, basestring):
 
637
            raise ValueError()
 
638
        to_dir_utf8 = to_dir.encode('utf8')
 
639
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
640
        id_index = state._get_id_index()
 
641
        # check destination directory
 
642
        # get the details for it
 
643
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
644
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
645
        if not entry_present:
 
646
            raise errors.BzrMoveFailedError('', to_dir,
 
647
                errors.NotVersionedError(to_dir))
 
648
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
649
        # get a handle on the block itself.
 
650
        to_block_index = state._ensure_block(
 
651
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
652
        to_block = state._dirblocks[to_block_index]
 
653
        to_abs = self.abspath(to_dir)
 
654
        if not isdir(to_abs):
 
655
            raise errors.BzrMoveFailedError('',to_dir,
 
656
                errors.NotADirectory(to_abs))
 
657
 
 
658
        if to_entry[1][0][0] != 'd':
 
659
            raise errors.BzrMoveFailedError('',to_dir,
 
660
                errors.NotADirectory(to_abs))
 
661
 
 
662
        if self._inventory is not None:
 
663
            update_inventory = True
 
664
            inv = self.inventory
 
665
            to_dir_id = to_entry[0][2]
 
666
            to_dir_ie = inv[to_dir_id]
 
667
        else:
 
668
            update_inventory = False
 
669
 
 
670
        rollbacks = []
 
671
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
672
                     fingerprint, packed_stat, size,
 
673
                     to_block, to_key, to_path_utf8):
 
674
            state._make_absent(old_entry)
 
675
            from_key = old_entry[0]
 
676
            rollbacks.append(
 
677
                lambda:state.update_minimal(from_key,
 
678
                    minikind,
 
679
                    executable=executable,
 
680
                    fingerprint=fingerprint,
 
681
                    packed_stat=packed_stat,
 
682
                    size=size,
 
683
                    path_utf8=from_path_utf8))
 
684
            state.update_minimal(to_key,
 
685
                    minikind,
 
686
                    executable=executable,
 
687
                    fingerprint=fingerprint,
 
688
                    packed_stat=packed_stat,
 
689
                    size=size,
 
690
                    path_utf8=to_path_utf8)
 
691
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
692
            new_entry = to_block[1][added_entry_index]
 
693
            rollbacks.append(lambda:state._make_absent(new_entry))
 
694
 
 
695
        for from_rel in from_paths:
 
696
            # from_rel is 'pathinroot/foo/bar'
 
697
            from_rel_utf8 = from_rel.encode('utf8')
 
698
            from_dirname, from_tail = osutils.split(from_rel)
 
699
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
700
            from_entry = self._get_entry(path=from_rel)
 
701
            if from_entry == (None, None):
 
702
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
703
                    errors.NotVersionedError(path=str(from_rel)))
 
704
 
 
705
            from_id = from_entry[0][2]
 
706
            to_rel = pathjoin(to_dir, from_tail)
 
707
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
708
            item_to_entry = self._get_entry(path=to_rel)
 
709
            if item_to_entry != (None, None):
 
710
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
711
                    "Target is already versioned.")
 
712
 
 
713
            if from_rel == to_rel:
 
714
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
715
                    "Source and target are identical.")
 
716
 
 
717
            from_missing = not self.has_filename(from_rel)
 
718
            to_missing = not self.has_filename(to_rel)
 
719
            if after:
 
720
                move_file = False
 
721
            else:
 
722
                move_file = True
 
723
            if to_missing:
 
724
                if not move_file:
 
725
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
726
                        errors.NoSuchFile(path=to_rel,
 
727
                        extra="New file has not been created yet"))
 
728
                elif from_missing:
 
729
                    # neither path exists
 
730
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
731
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
732
            else:
 
733
                if from_missing: # implicitly just update our path mapping
 
734
                    move_file = False
 
735
                elif not after:
 
736
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
737
 
 
738
            rollbacks = []
 
739
            def rollback_rename():
 
740
                """A single rename has failed, roll it back."""
 
741
                # roll back everything, even if we encounter trouble doing one
 
742
                # of them.
 
743
                #
 
744
                # TODO: at least log the other exceptions rather than just
 
745
                # losing them mbp 20070307
 
746
                exc_info = None
 
747
                for rollback in reversed(rollbacks):
 
748
                    try:
 
749
                        rollback()
 
750
                    except Exception, e:
 
751
                        exc_info = sys.exc_info()
 
752
                if exc_info:
 
753
                    raise exc_info[0], exc_info[1], exc_info[2]
 
754
 
 
755
            # perform the disk move first - its the most likely failure point.
 
756
            if move_file:
 
757
                from_rel_abs = self.abspath(from_rel)
 
758
                to_rel_abs = self.abspath(to_rel)
 
759
                try:
 
760
                    osutils.rename(from_rel_abs, to_rel_abs)
 
761
                except OSError, e:
 
762
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
763
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
764
            try:
 
765
                # perform the rename in the inventory next if needed: its easy
 
766
                # to rollback
 
767
                if update_inventory:
 
768
                    # rename the entry
 
769
                    from_entry = inv[from_id]
 
770
                    current_parent = from_entry.parent_id
 
771
                    inv.rename(from_id, to_dir_id, from_tail)
 
772
                    rollbacks.append(
 
773
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
774
                # finally do the rename in the dirstate, which is a little
 
775
                # tricky to rollback, but least likely to need it.
 
776
                old_block_index, old_entry_index, dir_present, file_present = \
 
777
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
778
                old_block = state._dirblocks[old_block_index][1]
 
779
                old_entry = old_block[old_entry_index]
 
780
                from_key, old_entry_details = old_entry
 
781
                cur_details = old_entry_details[0]
 
782
                # remove the old row
 
783
                to_key = ((to_block[0],) + from_key[1:3])
 
784
                minikind = cur_details[0]
 
785
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
786
                         minikind=minikind,
 
787
                         executable=cur_details[3],
 
788
                         fingerprint=cur_details[1],
 
789
                         packed_stat=cur_details[4],
 
790
                         size=cur_details[2],
 
791
                         to_block=to_block,
 
792
                         to_key=to_key,
 
793
                         to_path_utf8=to_rel_utf8)
 
794
 
 
795
                if minikind == 'd':
 
796
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
797
                        """Recursively update all entries in this dirblock."""
 
798
                        if from_dir == '':
 
799
                            raise AssertionError("renaming root not supported")
 
800
                        from_key = (from_dir, '')
 
801
                        from_block_idx, present = \
 
802
                            state._find_block_index_from_key(from_key)
 
803
                        if not present:
 
804
                            # This is the old record, if it isn't present, then
 
805
                            # there is theoretically nothing to update.
 
806
                            # (Unless it isn't present because of lazy loading,
 
807
                            # but we don't do that yet)
 
808
                            return
 
809
                        from_block = state._dirblocks[from_block_idx]
 
810
                        to_block_index, to_entry_index, _, _ = \
 
811
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
812
                        to_block_index = state._ensure_block(
 
813
                            to_block_index, to_entry_index, to_dir_utf8)
 
814
                        to_block = state._dirblocks[to_block_index]
 
815
 
 
816
                        # Grab a copy since move_one may update the list.
 
817
                        for entry in from_block[1][:]:
 
818
                            if not (entry[0][0] == from_dir):
 
819
                                raise AssertionError()
 
820
                            cur_details = entry[1][0]
 
821
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
822
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
823
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
824
                            minikind = cur_details[0]
 
825
                            if minikind in 'ar':
 
826
                                # Deleted children of a renamed directory
 
827
                                # Do not need to be updated.
 
828
                                # Children that have been renamed out of this
 
829
                                # directory should also not be updated
 
830
                                continue
 
831
                            move_one(entry, from_path_utf8=from_path_utf8,
 
832
                                     minikind=minikind,
 
833
                                     executable=cur_details[3],
 
834
                                     fingerprint=cur_details[1],
 
835
                                     packed_stat=cur_details[4],
 
836
                                     size=cur_details[2],
 
837
                                     to_block=to_block,
 
838
                                     to_key=to_key,
 
839
                                     to_path_utf8=to_path_utf8)
 
840
                            if minikind == 'd':
 
841
                                # We need to move all the children of this
 
842
                                # entry
 
843
                                update_dirblock(from_path_utf8, to_key,
 
844
                                                to_path_utf8)
 
845
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
846
            except:
 
847
                rollback_rename()
 
848
                raise
 
849
            result.append((from_rel, to_rel))
 
850
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
851
            self._make_dirty(reset_inventory=False)
 
852
 
 
853
        return result
 
854
 
 
855
    def _must_be_locked(self):
 
856
        if not self._control_files._lock_count:
 
857
            raise errors.ObjectNotLocked(self)
 
858
 
 
859
    def _new_tree(self):
 
860
        """Initialize the state in this tree to be a new tree."""
 
861
        self._dirty = True
 
862
 
 
863
    @needs_read_lock
 
864
    def path2id(self, path):
 
865
        """Return the id for path in this tree."""
 
866
        path = path.strip('/')
 
867
        entry = self._get_entry(path=path)
 
868
        if entry == (None, None):
 
869
            return None
 
870
        return entry[0][2]
 
871
 
 
872
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
873
        """See Tree.paths2ids().
 
874
 
 
875
        This specialisation fast-paths the case where all the trees are in the
 
876
        dirstate.
 
877
        """
 
878
        if paths is None:
 
879
            return None
 
880
        parents = self.get_parent_ids()
 
881
        for tree in trees:
 
882
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
883
                parents):
 
884
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
885
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
886
        # -- make all paths utf8 --
 
887
        paths_utf8 = set()
 
888
        for path in paths:
 
889
            paths_utf8.add(path.encode('utf8'))
 
890
        paths = paths_utf8
 
891
        # -- paths is now a utf8 path set --
 
892
        # -- get the state object and prepare it.
 
893
        state = self.current_dirstate()
 
894
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
895
            and '' not in paths):
 
896
            paths2ids = self._paths2ids_using_bisect
 
897
        else:
 
898
            paths2ids = self._paths2ids_in_memory
 
899
        return paths2ids(paths, search_indexes,
 
900
                         require_versioned=require_versioned)
 
901
 
 
902
    def _paths2ids_in_memory(self, paths, search_indexes,
 
903
                             require_versioned=True):
 
904
        state = self.current_dirstate()
 
905
        state._read_dirblocks_if_needed()
 
906
        def _entries_for_path(path):
 
907
            """Return a list with all the entries that match path for all ids.
 
908
            """
 
909
            dirname, basename = os.path.split(path)
 
910
            key = (dirname, basename, '')
 
911
            block_index, present = state._find_block_index_from_key(key)
 
912
            if not present:
 
913
                # the block which should contain path is absent.
 
914
                return []
 
915
            result = []
 
916
            block = state._dirblocks[block_index][1]
 
917
            entry_index, _ = state._find_entry_index(key, block)
 
918
            # we may need to look at multiple entries at this path: walk while the paths match.
 
919
            while (entry_index < len(block) and
 
920
                block[entry_index][0][0:2] == key[0:2]):
 
921
                result.append(block[entry_index])
 
922
                entry_index += 1
 
923
            return result
 
924
        if require_versioned:
 
925
            # -- check all supplied paths are versioned in a search tree. --
 
926
            all_versioned = True
 
927
            for path in paths:
 
928
                path_entries = _entries_for_path(path)
 
929
                if not path_entries:
 
930
                    # this specified path is not present at all: error
 
931
                    all_versioned = False
 
932
                    break
 
933
                found_versioned = False
 
934
                # for each id at this path
 
935
                for entry in path_entries:
 
936
                    # for each tree.
 
937
                    for index in search_indexes:
 
938
                        if entry[1][index][0] != 'a': # absent
 
939
                            found_versioned = True
 
940
                            # all good: found a versioned cell
 
941
                            break
 
942
                if not found_versioned:
 
943
                    # none of the indexes was not 'absent' at all ids for this
 
944
                    # path.
 
945
                    all_versioned = False
 
946
                    break
 
947
            if not all_versioned:
 
948
                raise errors.PathsNotVersionedError(paths)
 
949
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
950
        search_paths = osutils.minimum_path_selection(paths)
 
951
        # sketch: 
 
952
        # for all search_indexs in each path at or under each element of
 
953
        # search_paths, if the detail is relocated: add the id, and add the
 
954
        # relocated path as one to search if its not searched already. If the
 
955
        # detail is not relocated, add the id.
 
956
        searched_paths = set()
 
957
        found_ids = set()
 
958
        def _process_entry(entry):
 
959
            """Look at search_indexes within entry.
 
960
 
 
961
            If a specific tree's details are relocated, add the relocation
 
962
            target to search_paths if not searched already. If it is absent, do
 
963
            nothing. Otherwise add the id to found_ids.
 
964
            """
 
965
            for index in search_indexes:
 
966
                if entry[1][index][0] == 'r': # relocated
 
967
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
968
                        search_paths.add(entry[1][index][1])
 
969
                elif entry[1][index][0] != 'a': # absent
 
970
                    found_ids.add(entry[0][2])
 
971
        while search_paths:
 
972
            current_root = search_paths.pop()
 
973
            searched_paths.add(current_root)
 
974
            # process the entries for this containing directory: the rest will be
 
975
            # found by their parents recursively.
 
976
            root_entries = _entries_for_path(current_root)
 
977
            if not root_entries:
 
978
                # this specified path is not present at all, skip it.
 
979
                continue
 
980
            for entry in root_entries:
 
981
                _process_entry(entry)
 
982
            initial_key = (current_root, '', '')
 
983
            block_index, _ = state._find_block_index_from_key(initial_key)
 
984
            while (block_index < len(state._dirblocks) and
 
985
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
986
                for entry in state._dirblocks[block_index][1]:
 
987
                    _process_entry(entry)
 
988
                block_index += 1
 
989
        return found_ids
 
990
 
 
991
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
992
                                require_versioned=True):
 
993
        state = self.current_dirstate()
 
994
        found_ids = set()
 
995
 
 
996
        split_paths = sorted(osutils.split(p) for p in paths)
 
997
        found = state._bisect_recursive(split_paths)
 
998
 
 
999
        if require_versioned:
 
1000
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
1001
            for dir_name in split_paths:
 
1002
                if dir_name not in found_dir_names:
 
1003
                    raise errors.PathsNotVersionedError(paths)
 
1004
 
 
1005
        for dir_name_id, trees_info in found.iteritems():
 
1006
            for index in search_indexes:
 
1007
                if trees_info[index][0] not in ('r', 'a'):
 
1008
                    found_ids.add(dir_name_id[2])
 
1009
        return found_ids
 
1010
 
 
1011
    def read_working_inventory(self):
 
1012
        """Read the working inventory.
 
1013
        
 
1014
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1015
        """
 
1016
        return self.inventory
 
1017
 
 
1018
    @needs_read_lock
 
1019
    def revision_tree(self, revision_id):
 
1020
        """See Tree.revision_tree.
 
1021
 
 
1022
        WorkingTree4 supplies revision_trees for any basis tree.
 
1023
        """
 
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
        parents = self.get_parent_ids()
 
1037
        if new_revision in (NULL_REVISION, None):
 
1038
            if len(parents) >= 2:
 
1039
                raise AssertionError(
 
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
        trees = []
 
1061
        for revision_id in revision_ids:
 
1062
            try:
 
1063
                revtree = self.branch.repository.revision_tree(revision_id)
 
1064
                # TODO: jam 20070213 KnitVersionedFile raises
 
1065
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1066
                #       given revision_id is not present. Should Repository be
 
1067
                #       catching it and re-raising NoSuchRevision?
 
1068
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1069
                revtree = None
 
1070
            trees.append((revision_id, revtree))
 
1071
        self.set_parent_trees(trees,
 
1072
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1073
 
 
1074
    @needs_tree_write_lock
 
1075
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1076
        """Set the parents of the working tree.
 
1077
 
 
1078
        :param parents_list: A list of (revision_id, tree) tuples.
 
1079
            If tree is None, then that element is treated as an unreachable
 
1080
            parent tree - i.e. a ghost.
 
1081
        """
 
1082
        dirstate = self.current_dirstate()
 
1083
        if len(parents_list) > 0:
 
1084
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1085
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1086
        real_trees = []
 
1087
        ghosts = []
 
1088
 
 
1089
        parent_ids = [rev_id for rev_id, tree in parents_list]
 
1090
        graph = self.branch.repository.get_graph()
 
1091
        heads = graph.heads(parent_ids)
 
1092
        accepted_revisions = set()
 
1093
 
 
1094
        # convert absent trees to the null tree, which we convert back to
 
1095
        # missing on access.
 
1096
        for rev_id, tree in parents_list:
 
1097
            if len(accepted_revisions) > 0:
 
1098
                # we always accept the first tree
 
1099
                if rev_id in accepted_revisions or rev_id not in heads:
 
1100
                    # We have already included either this tree, or its
 
1101
                    # descendent, so we skip it.
 
1102
                    continue
 
1103
            _mod_revision.check_not_reserved_id(rev_id)
 
1104
            if tree is not None:
 
1105
                real_trees.append((rev_id, tree))
 
1106
            else:
 
1107
                real_trees.append((rev_id,
 
1108
                    self.branch.repository.revision_tree(None)))
 
1109
                ghosts.append(rev_id)
 
1110
            accepted_revisions.add(rev_id)
 
1111
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1112
        self._make_dirty(reset_inventory=False)
 
1113
 
 
1114
    def _set_root_id(self, file_id):
 
1115
        """See WorkingTree.set_root_id."""
 
1116
        state = self.current_dirstate()
 
1117
        state.set_path_id('', file_id)
 
1118
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1119
            self._make_dirty(reset_inventory=True)
 
1120
 
 
1121
    def _sha_from_stat(self, path, stat_result):
 
1122
        """Get a sha digest from the tree's stat cache.
 
1123
 
 
1124
        The default implementation assumes no stat cache is present.
 
1125
 
 
1126
        :param path: The path.
 
1127
        :param stat_result: The stat result being looked up.
 
1128
        """
 
1129
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1130
 
 
1131
    @needs_read_lock
 
1132
    def supports_tree_reference(self):
 
1133
        return self._repo_supports_tree_reference
 
1134
 
 
1135
    def unlock(self):
 
1136
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1137
        # do non-implementation specific cleanup
 
1138
        self._cleanup()
 
1139
 
 
1140
        if self._control_files._lock_count == 1:
 
1141
            # eventually we should do signature checking during read locks for
 
1142
            # dirstate updates.
 
1143
            if self._control_files._lock_mode == 'w':
 
1144
                if self._dirty:
 
1145
                    self.flush()
 
1146
            if self._dirstate is not None:
 
1147
                # This is a no-op if there are no modifications.
 
1148
                self._dirstate.save()
 
1149
                self._dirstate.unlock()
 
1150
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1151
            #       point. Instead, it could check if the header has been
 
1152
            #       modified when it is locked, and if not, it can hang on to
 
1153
            #       the data it has in memory.
 
1154
            self._dirstate = None
 
1155
            self._inventory = None
 
1156
        # reverse order of locking.
 
1157
        try:
 
1158
            return self._control_files.unlock()
 
1159
        finally:
 
1160
            self.branch.unlock()
 
1161
 
 
1162
    @needs_tree_write_lock
 
1163
    def unversion(self, file_ids):
 
1164
        """Remove the file ids in file_ids from the current versioned set.
 
1165
 
 
1166
        When a file_id is unversioned, all of its children are automatically
 
1167
        unversioned.
 
1168
 
 
1169
        :param file_ids: The file ids to stop versioning.
 
1170
        :raises: NoSuchId if any fileid is not currently versioned.
 
1171
        """
 
1172
        if not file_ids:
 
1173
            return
 
1174
        state = self.current_dirstate()
 
1175
        state._read_dirblocks_if_needed()
 
1176
        ids_to_unversion = set(file_ids)
 
1177
        paths_to_unversion = set()
 
1178
        # sketch:
 
1179
        # check if the root is to be unversioned, if so, assert for now.
 
1180
        # walk the state marking unversioned things as absent.
 
1181
        # if there are any un-unversioned ids at the end, raise
 
1182
        for key, details in state._dirblocks[0][1]:
 
1183
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1184
                key[2] in ids_to_unversion):
 
1185
                # I haven't written the code to unversion / yet - it should be
 
1186
                # supported.
 
1187
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1188
        block_index = 0
 
1189
        while block_index < len(state._dirblocks):
 
1190
            # process one directory at a time.
 
1191
            block = state._dirblocks[block_index]
 
1192
            # first check: is the path one to remove - it or its children
 
1193
            delete_block = False
 
1194
            for path in paths_to_unversion:
 
1195
                if (block[0].startswith(path) and
 
1196
                    (len(block[0]) == len(path) or
 
1197
                     block[0][len(path)] == '/')):
 
1198
                    # this entire block should be deleted - its the block for a
 
1199
                    # path to unversion; or the child of one
 
1200
                    delete_block = True
 
1201
                    break
 
1202
            # TODO: trim paths_to_unversion as we pass by paths
 
1203
            if delete_block:
 
1204
                # this block is to be deleted: process it.
 
1205
                # TODO: we can special case the no-parents case and
 
1206
                # just forget the whole block.
 
1207
                entry_index = 0
 
1208
                while entry_index < len(block[1]):
 
1209
                    # Mark this file id as having been removed
 
1210
                    entry = block[1][entry_index]
 
1211
                    ids_to_unversion.discard(entry[0][2])
 
1212
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1213
                                               # entries
 
1214
                        or not state._make_absent(entry)):
 
1215
                        entry_index += 1
 
1216
                # go to the next block. (At the moment we dont delete empty
 
1217
                # dirblocks)
 
1218
                block_index += 1
 
1219
                continue
 
1220
            entry_index = 0
 
1221
            while entry_index < len(block[1]):
 
1222
                entry = block[1][entry_index]
 
1223
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1224
                    # ^ some parent row.
 
1225
                    entry[0][2] not in ids_to_unversion):
 
1226
                    # ^ not an id to unversion
 
1227
                    entry_index += 1
 
1228
                    continue
 
1229
                if entry[1][0][0] == 'd':
 
1230
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1231
                if not state._make_absent(entry):
 
1232
                    entry_index += 1
 
1233
                # we have unversioned this id
 
1234
                ids_to_unversion.remove(entry[0][2])
 
1235
            block_index += 1
 
1236
        if ids_to_unversion:
 
1237
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1238
        self._make_dirty(reset_inventory=False)
 
1239
        # have to change the legacy inventory too.
 
1240
        if self._inventory is not None:
 
1241
            for file_id in file_ids:
 
1242
                self._inventory.remove_recursive_id(file_id)
 
1243
 
 
1244
    @needs_tree_write_lock
 
1245
    def rename_one(self, from_rel, to_rel, after=False):
 
1246
        """See WorkingTree.rename_one"""
 
1247
        self.flush()
 
1248
        WorkingTree.rename_one(self, from_rel, to_rel, after)
 
1249
 
 
1250
    @needs_tree_write_lock
 
1251
    def apply_inventory_delta(self, changes):
 
1252
        """See MutableTree.apply_inventory_delta"""
 
1253
        state = self.current_dirstate()
 
1254
        state.update_by_delta(changes)
 
1255
        self._make_dirty(reset_inventory=True)
 
1256
 
 
1257
    def update_basis_by_delta(self, new_revid, delta):
 
1258
        """See MutableTree.update_basis_by_delta."""
 
1259
        if self.last_revision() == new_revid:
 
1260
            raise AssertionError()
 
1261
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1262
 
 
1263
    @needs_read_lock
 
1264
    def _validate(self):
 
1265
        self._dirstate._validate()
 
1266
 
 
1267
    @needs_tree_write_lock
 
1268
    def _write_inventory(self, inv):
 
1269
        """Write inventory as the current inventory."""
 
1270
        if self._dirty:
 
1271
            raise AssertionError("attempting to write an inventory when the "
 
1272
                "dirstate is dirty will lose pending changes")
 
1273
        self.current_dirstate().set_state_from_inventory(inv)
 
1274
        self._make_dirty(reset_inventory=False)
 
1275
        if self._inventory is not None:
 
1276
            self._inventory = inv
 
1277
        self.flush()
 
1278
 
 
1279
 
 
1280
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1281
    """The first consolidated dirstate working tree format.
 
1282
 
 
1283
    This format:
 
1284
        - exists within a metadir controlling .bzr
 
1285
        - includes an explicit version marker for the workingtree control
 
1286
          files, separate from the BzrDir format
 
1287
        - modifies the hash cache format
 
1288
        - is new in bzr 0.15
 
1289
        - uses a LockDir to guard access to it.
 
1290
    """
 
1291
 
 
1292
    upgrade_recommended = False
 
1293
 
 
1294
    def get_format_string(self):
 
1295
        """See WorkingTreeFormat.get_format_string()."""
 
1296
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1297
 
 
1298
    def get_format_description(self):
 
1299
        """See WorkingTreeFormat.get_format_description()."""
 
1300
        return "Working tree format 4"
 
1301
 
 
1302
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1303
                   accelerator_tree=None, hardlink=False):
 
1304
        """See WorkingTreeFormat.initialize().
 
1305
 
 
1306
        :param revision_id: allows creating a working tree at a different
 
1307
        revision than the branch is at.
 
1308
        :param accelerator_tree: A tree which can be used for retrieving file
 
1309
            contents more quickly than the revision tree, i.e. a workingtree.
 
1310
            The revision tree will be used for cases where accelerator_tree's
 
1311
            content is different.
 
1312
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1313
            where possible.
 
1314
 
 
1315
        These trees get an initial random root id, if their repository supports
 
1316
        rich root data, TREE_ROOT otherwise.
 
1317
        """
 
1318
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1319
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1320
        transport = a_bzrdir.get_workingtree_transport(self)
 
1321
        control_files = self._open_control_files(a_bzrdir)
 
1322
        control_files.create_lock()
 
1323
        control_files.lock_write()
 
1324
        transport.put_bytes('format', self.get_format_string(),
 
1325
            mode=a_bzrdir._get_file_mode())
 
1326
        if from_branch is not None:
 
1327
            branch = from_branch
 
1328
        else:
 
1329
            branch = a_bzrdir.open_branch()
 
1330
        if revision_id is None:
 
1331
            revision_id = branch.last_revision()
 
1332
        local_path = transport.local_abspath('dirstate')
 
1333
        # write out new dirstate (must exist when we create the tree)
 
1334
        state = dirstate.DirState.initialize(local_path)
 
1335
        state.unlock()
 
1336
        del state
 
1337
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1338
                         branch,
 
1339
                         _format=self,
 
1340
                         _bzrdir=a_bzrdir,
 
1341
                         _control_files=control_files)
 
1342
        wt._new_tree()
 
1343
        wt.lock_tree_write()
 
1344
        try:
 
1345
            if revision_id in (None, NULL_REVISION):
 
1346
                if branch.repository.supports_rich_root():
 
1347
                    wt._set_root_id(generate_ids.gen_root_id())
 
1348
                else:
 
1349
                    wt._set_root_id(ROOT_ID)
 
1350
                wt.flush()
 
1351
            basis = None
 
1352
            # frequently, we will get here due to branching.  The accelerator
 
1353
            # tree will be the tree from the branch, so the desired basis
 
1354
            # tree will often be a parent of the accelerator tree.
 
1355
            if accelerator_tree is not None:
 
1356
                try:
 
1357
                    basis = accelerator_tree.revision_tree(revision_id)
 
1358
                except errors.NoSuchRevision:
 
1359
                    pass
 
1360
            if basis is None:
 
1361
                basis = branch.repository.revision_tree(revision_id)
 
1362
            if revision_id == NULL_REVISION:
 
1363
                parents_list = []
 
1364
            else:
 
1365
                parents_list = [(revision_id, basis)]
 
1366
            basis.lock_read()
 
1367
            try:
 
1368
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1369
                wt.flush()
 
1370
                # if the basis has a root id we have to use that; otherwise we
 
1371
                # use a new random one
 
1372
                basis_root_id = basis.get_root_id()
 
1373
                if basis_root_id is not None:
 
1374
                    wt._set_root_id(basis_root_id)
 
1375
                    wt.flush()
 
1376
                # delta_from_tree is safe even for DirStateRevisionTrees,
 
1377
                # because wt4.apply_inventory_delta does not mutate the input
 
1378
                # inventory entries.
 
1379
                transform.build_tree(basis, wt, accelerator_tree,
 
1380
                                     hardlink=hardlink, delta_from_tree=True)
 
1381
            finally:
 
1382
                basis.unlock()
 
1383
        finally:
 
1384
            control_files.unlock()
 
1385
            wt.unlock()
 
1386
        return wt
 
1387
 
 
1388
    def _open(self, a_bzrdir, control_files):
 
1389
        """Open the tree itself.
 
1390
 
 
1391
        :param a_bzrdir: the dir for the tree.
 
1392
        :param control_files: the control files for the tree.
 
1393
        """
 
1394
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1395
                           branch=a_bzrdir.open_branch(),
 
1396
                           _format=self,
 
1397
                           _bzrdir=a_bzrdir,
 
1398
                           _control_files=control_files)
 
1399
 
 
1400
    def __get_matchingbzrdir(self):
 
1401
        # please test against something that will let us do tree references
 
1402
        return bzrdir.format_registry.make_bzrdir(
 
1403
            'dirstate-with-subtree')
 
1404
 
 
1405
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1406
 
 
1407
 
 
1408
class DirStateRevisionTree(Tree):
 
1409
    """A revision tree pulling the inventory from a dirstate."""
 
1410
 
 
1411
    def __init__(self, dirstate, revision_id, repository):
 
1412
        self._dirstate = dirstate
 
1413
        self._revision_id = revision_id
 
1414
        self._repository = repository
 
1415
        self._inventory = None
 
1416
        self._locked = 0
 
1417
        self._dirstate_locked = False
 
1418
 
 
1419
    def __repr__(self):
 
1420
        return "<%s of %s in %s>" % \
 
1421
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1422
 
 
1423
    def annotate_iter(self, file_id,
 
1424
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1425
        """See Tree.annotate_iter"""
 
1426
        w = self._get_weave(file_id)
 
1427
        return w.annotate(self.inventory[file_id].revision)
 
1428
 
 
1429
    def _get_ancestors(self, default_revision):
 
1430
        return set(self._repository.get_ancestry(self._revision_id,
 
1431
                                                 topo_sorted=False))
 
1432
    def _comparison_data(self, entry, path):
 
1433
        """See Tree._comparison_data."""
 
1434
        if entry is None:
 
1435
            return None, False, None
 
1436
        # trust the entry as RevisionTree does, but this may not be
 
1437
        # sensible: the entry might not have come from us?
 
1438
        return entry.kind, entry.executable, None
 
1439
 
 
1440
    def _file_size(self, entry, stat_value):
 
1441
        return entry.text_size
 
1442
 
 
1443
    def filter_unversioned_files(self, paths):
 
1444
        """Filter out paths that are not versioned.
 
1445
 
 
1446
        :return: set of paths.
 
1447
        """
 
1448
        pred = self.has_filename
 
1449
        return set((p for p in paths if not pred(p)))
 
1450
 
 
1451
    def get_root_id(self):
 
1452
        return self.path2id('')
 
1453
 
 
1454
    def id2path(self, file_id):
 
1455
        "Convert a file-id to a path."
 
1456
        entry = self._get_entry(file_id=file_id)
 
1457
        if entry == (None, None):
 
1458
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1459
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1460
        return path_utf8.decode('utf8')
 
1461
 
 
1462
    def _get_parent_index(self):
 
1463
        """Return the index in the dirstate referenced by this tree."""
 
1464
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1465
 
 
1466
    def _get_entry(self, file_id=None, path=None):
 
1467
        """Get the dirstate row for file_id or path.
 
1468
 
 
1469
        If either file_id or path is supplied, it is used as the key to lookup.
 
1470
        If both are supplied, the fastest lookup is used, and an error is
 
1471
        raised if they do not both point at the same row.
 
1472
        
 
1473
        :param file_id: An optional unicode file_id to be looked up.
 
1474
        :param path: An optional unicode path to be looked up.
 
1475
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1476
        """
 
1477
        if file_id is None and path is None:
 
1478
            raise errors.BzrError('must supply file_id or path')
 
1479
        if path is not None:
 
1480
            path = path.encode('utf8')
 
1481
        parent_index = self._get_parent_index()
 
1482
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1483
 
 
1484
    def _generate_inventory(self):
 
1485
        """Create and set self.inventory from the dirstate object.
 
1486
 
 
1487
        (So this is only called the first time the inventory is requested for
 
1488
        this tree; it then remains in memory until it's out of date.)
 
1489
 
 
1490
        This is relatively expensive: we have to walk the entire dirstate.
 
1491
        """
 
1492
        if not self._locked:
 
1493
            raise AssertionError(
 
1494
                'cannot generate inventory of an unlocked '
 
1495
                'dirstate revision tree')
 
1496
        # separate call for profiling - makes it clear where the costs are.
 
1497
        self._dirstate._read_dirblocks_if_needed()
 
1498
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1499
            raise AssertionError(
 
1500
                'parent %s has disappeared from %s' % (
 
1501
                self._revision_id, self._dirstate.get_parent_ids()))
 
1502
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1503
        # This is identical now to the WorkingTree _generate_inventory except
 
1504
        # for the tree index use.
 
1505
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1506
        current_id = root_key[2]
 
1507
        if current_entry[parent_index][0] != 'd':
 
1508
            raise AssertionError()
 
1509
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1510
        inv.root.revision = current_entry[parent_index][4]
 
1511
        # Turn some things into local variables
 
1512
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1513
        factory = entry_factory
 
1514
        utf8_decode = cache_utf8._utf8_decode
 
1515
        inv_byid = inv._byid
 
1516
        # we could do this straight out of the dirstate; it might be fast
 
1517
        # and should be profiled - RBC 20070216
 
1518
        parent_ies = {'' : inv.root}
 
1519
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1520
            dirname = block[0]
 
1521
            try:
 
1522
                parent_ie = parent_ies[dirname]
 
1523
            except KeyError:
 
1524
                # all the paths in this block are not versioned in this tree
 
1525
                continue
 
1526
            for key, entry in block[1]:
 
1527
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1528
                if minikind in ('a', 'r'): # absent, relocated
 
1529
                    # not this tree
 
1530
                    continue
 
1531
                name = key[1]
 
1532
                name_unicode = utf8_decode(name)[0]
 
1533
                file_id = key[2]
 
1534
                kind = minikind_to_kind[minikind]
 
1535
                inv_entry = factory[kind](file_id, name_unicode,
 
1536
                                          parent_ie.file_id)
 
1537
                inv_entry.revision = revid
 
1538
                if kind == 'file':
 
1539
                    inv_entry.executable = executable
 
1540
                    inv_entry.text_size = size
 
1541
                    inv_entry.text_sha1 = fingerprint
 
1542
                elif kind == 'directory':
 
1543
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1544
                elif kind == 'symlink':
 
1545
                    inv_entry.executable = False
 
1546
                    inv_entry.text_size = None
 
1547
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1548
                elif kind == 'tree-reference':
 
1549
                    inv_entry.reference_revision = fingerprint or None
 
1550
                else:
 
1551
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1552
                            % entry)
 
1553
                # These checks cost us around 40ms on a 55k entry tree
 
1554
                if file_id in inv_byid:
 
1555
                    raise AssertionError('file_id %s already in'
 
1556
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1557
                if name_unicode in parent_ie.children:
 
1558
                    raise AssertionError('name %r already in parent'
 
1559
                        % (name_unicode,))
 
1560
                inv_byid[file_id] = inv_entry
 
1561
                parent_ie.children[name_unicode] = inv_entry
 
1562
        self._inventory = inv
 
1563
 
 
1564
    def get_file_mtime(self, file_id, path=None):
 
1565
        """Return the modification time for this record.
 
1566
 
 
1567
        We return the timestamp of the last-changed revision.
 
1568
        """
 
1569
        # Make sure the file exists
 
1570
        entry = self._get_entry(file_id, path=path)
 
1571
        if entry == (None, None): # do we raise?
 
1572
            return None
 
1573
        parent_index = self._get_parent_index()
 
1574
        last_changed_revision = entry[1][parent_index][4]
 
1575
        return self._repository.get_revision(last_changed_revision).timestamp
 
1576
 
 
1577
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1578
        entry = self._get_entry(file_id=file_id, path=path)
 
1579
        parent_index = self._get_parent_index()
 
1580
        parent_details = entry[1][parent_index]
 
1581
        if parent_details[0] == 'f':
 
1582
            return parent_details[1]
 
1583
        return None
 
1584
 
 
1585
    def _get_weave(self, file_id):
 
1586
        return self._repository.weave_store.get_weave(file_id,
 
1587
                self._repository.get_transaction())
 
1588
 
 
1589
    def get_file(self, file_id, path=None):
 
1590
        return StringIO(self.get_file_text(file_id))
 
1591
 
 
1592
    def get_file_lines(self, file_id):
 
1593
        entry = self._get_entry(file_id=file_id)[1]
 
1594
        if entry is None:
 
1595
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1596
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1597
 
 
1598
    def get_file_size(self, file_id):
 
1599
        """See Tree.get_file_size"""
 
1600
        return self.inventory[file_id].text_size
 
1601
 
 
1602
    def get_file_text(self, file_id):
 
1603
        return ''.join(self.get_file_lines(file_id))
 
1604
 
 
1605
    def get_reference_revision(self, file_id, path=None):
 
1606
        return self.inventory[file_id].reference_revision
 
1607
 
 
1608
    def iter_files_bytes(self, desired_files):
 
1609
        """See Tree.iter_files_bytes.
 
1610
 
 
1611
        This version is implemented on top of Repository.iter_files_bytes"""
 
1612
        parent_index = self._get_parent_index()
 
1613
        repo_desired_files = []
 
1614
        for file_id, identifier in desired_files:
 
1615
            entry = self._get_entry(file_id)
 
1616
            if entry == (None, None):
 
1617
                raise errors.NoSuchId(self, file_id)
 
1618
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1619
                                       identifier))
 
1620
        return self._repository.iter_files_bytes(repo_desired_files)
 
1621
 
 
1622
    def get_symlink_target(self, file_id):
 
1623
        entry = self._get_entry(file_id=file_id)
 
1624
        parent_index = self._get_parent_index()
 
1625
        if entry[1][parent_index][0] != 'l':
 
1626
            return None
 
1627
        else:
 
1628
            # At present, none of the tree implementations supports non-ascii
 
1629
            # symlink targets. So we will just assume that the dirstate path is
 
1630
            # correct.
 
1631
            return entry[1][parent_index][1]
 
1632
 
 
1633
    def get_revision_id(self):
 
1634
        """Return the revision id for this tree."""
 
1635
        return self._revision_id
 
1636
 
 
1637
    def _get_inventory(self):
 
1638
        if self._inventory is not None:
 
1639
            return self._inventory
 
1640
        self._must_be_locked()
 
1641
        self._generate_inventory()
 
1642
        return self._inventory
 
1643
 
 
1644
    inventory = property(_get_inventory,
 
1645
                         doc="Inventory of this Tree")
 
1646
 
 
1647
    def get_parent_ids(self):
 
1648
        """The parents of a tree in the dirstate are not cached."""
 
1649
        return self._repository.get_revision(self._revision_id).parent_ids
 
1650
 
 
1651
    def has_filename(self, filename):
 
1652
        return bool(self.path2id(filename))
 
1653
 
 
1654
    def kind(self, file_id):
 
1655
        entry = self._get_entry(file_id=file_id)[1]
 
1656
        if entry is None:
 
1657
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1658
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1659
 
 
1660
    def stored_kind(self, file_id):
 
1661
        """See Tree.stored_kind"""
 
1662
        return self.kind(file_id)
 
1663
 
 
1664
    def path_content_summary(self, path):
 
1665
        """See Tree.path_content_summary."""
 
1666
        id = self.inventory.path2id(path)
 
1667
        if id is None:
 
1668
            return ('missing', None, None, None)
 
1669
        entry = self._inventory[id]
 
1670
        kind = entry.kind
 
1671
        if kind == 'file':
 
1672
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1673
        elif kind == 'symlink':
 
1674
            return (kind, None, None, entry.symlink_target)
 
1675
        else:
 
1676
            return (kind, None, None, None)
 
1677
 
 
1678
    def is_executable(self, file_id, path=None):
 
1679
        ie = self.inventory[file_id]
 
1680
        if ie.kind != "file":
 
1681
            return None
 
1682
        return ie.executable
 
1683
 
 
1684
    def list_files(self, include_root=False):
 
1685
        # We use a standard implementation, because DirStateRevisionTree is
 
1686
        # dealing with one of the parents of the current state
 
1687
        inv = self._get_inventory()
 
1688
        entries = inv.iter_entries()
 
1689
        if self.inventory.root is not None and not include_root:
 
1690
            entries.next()
 
1691
        for path, entry in entries:
 
1692
            yield path, 'V', entry.kind, entry.file_id, entry
 
1693
 
 
1694
    def lock_read(self):
 
1695
        """Lock the tree for a set of operations."""
 
1696
        if not self._locked:
 
1697
            self._repository.lock_read()
 
1698
            if self._dirstate._lock_token is None:
 
1699
                self._dirstate.lock_read()
 
1700
                self._dirstate_locked = True
 
1701
        self._locked += 1
 
1702
 
 
1703
    def _must_be_locked(self):
 
1704
        if not self._locked:
 
1705
            raise errors.ObjectNotLocked(self)
 
1706
 
 
1707
    @needs_read_lock
 
1708
    def path2id(self, path):
 
1709
        """Return the id for path in this tree."""
 
1710
        # lookup by path: faster than splitting and walking the ivnentory.
 
1711
        entry = self._get_entry(path=path)
 
1712
        if entry == (None, None):
 
1713
            return None
 
1714
        return entry[0][2]
 
1715
 
 
1716
    def unlock(self):
 
1717
        """Unlock, freeing any cache memory used during the lock."""
 
1718
        # outside of a lock, the inventory is suspect: release it.
 
1719
        self._locked -=1
 
1720
        if not self._locked:
 
1721
            self._inventory = None
 
1722
            self._locked = 0
 
1723
            if self._dirstate_locked:
 
1724
                self._dirstate.unlock()
 
1725
                self._dirstate_locked = False
 
1726
            self._repository.unlock()
 
1727
 
 
1728
    def walkdirs(self, prefix=""):
 
1729
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1730
        # implementation based on an inventory.  
 
1731
        # This should be cleaned up to use the much faster Dirstate code
 
1732
        # So for now, we just build up the parent inventory, and extract
 
1733
        # it the same way RevisionTree does.
 
1734
        _directory = 'directory'
 
1735
        inv = self._get_inventory()
 
1736
        top_id = inv.path2id(prefix)
 
1737
        if top_id is None:
 
1738
            pending = []
 
1739
        else:
 
1740
            pending = [(prefix, top_id)]
 
1741
        while pending:
 
1742
            dirblock = []
 
1743
            relpath, file_id = pending.pop()
 
1744
            # 0 - relpath, 1- file-id
 
1745
            if relpath:
 
1746
                relroot = relpath + '/'
 
1747
            else:
 
1748
                relroot = ""
 
1749
            # FIXME: stash the node in pending
 
1750
            entry = inv[file_id]
 
1751
            for name, child in entry.sorted_children():
 
1752
                toppath = relroot + name
 
1753
                dirblock.append((toppath, name, child.kind, None,
 
1754
                    child.file_id, child.kind
 
1755
                    ))
 
1756
            yield (relpath, entry.file_id), dirblock
 
1757
            # push the user specified dirs from dirblock
 
1758
            for dir in reversed(dirblock):
 
1759
                if dir[2] == _directory:
 
1760
                    pending.append((dir[0], dir[4]))
 
1761
 
 
1762
 
 
1763
class InterDirStateTree(InterTree):
 
1764
    """Fast path optimiser for changes_from with dirstate trees.
 
1765
    
 
1766
    This is used only when both trees are in the dirstate working file, and 
 
1767
    the source is any parent within the dirstate, and the destination is 
 
1768
    the current working tree of the same dirstate.
 
1769
    """
 
1770
    # this could be generalized to allow comparisons between any trees in the
 
1771
    # dirstate, and possibly between trees stored in different dirstates.
 
1772
 
 
1773
    def __init__(self, source, target):
 
1774
        super(InterDirStateTree, self).__init__(source, target)
 
1775
        if not InterDirStateTree.is_compatible(source, target):
 
1776
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1777
 
 
1778
    @staticmethod
 
1779
    def make_source_parent_tree(source, target):
 
1780
        """Change the source tree into a parent of the target."""
 
1781
        revid = source.commit('record tree')
 
1782
        target.branch.repository.fetch(source.branch.repository, revid)
 
1783
        target.set_parent_ids([revid])
 
1784
        return target.basis_tree(), target
 
1785
 
 
1786
    _matching_from_tree_format = WorkingTreeFormat4()
 
1787
    _matching_to_tree_format = WorkingTreeFormat4()
 
1788
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1789
 
 
1790
    def iter_changes(self, include_unchanged=False,
 
1791
                      specific_files=None, pb=None, extra_trees=[],
 
1792
                      require_versioned=True, want_unversioned=False):
 
1793
        """Return the changes from source to target.
 
1794
 
 
1795
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1796
            for details.
 
1797
        :param specific_files: An optional list of file paths to restrict the
 
1798
            comparison to. When mapping filenames to ids, all matches in all
 
1799
            trees (including optional extra_trees) are used, and all children of
 
1800
            matched directories are included.
 
1801
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1802
            unchanged entries in the result.
 
1803
        :param extra_trees: An optional list of additional trees to use when
 
1804
            mapping the contents of specific_files (paths) to file_ids.
 
1805
        :param require_versioned: If True, all files in specific_files must be
 
1806
            versioned in one of source, target, extra_trees or
 
1807
            PathsNotVersionedError is raised.
 
1808
        :param want_unversioned: Should unversioned files be returned in the
 
1809
            output. An unversioned file is defined as one with (False, False)
 
1810
            for the versioned pair.
 
1811
        """
 
1812
        utf8_decode = cache_utf8._utf8_decode
 
1813
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1814
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1815
        # NB: show_status depends on being able to pass in non-versioned files
 
1816
        # and report them as unknown
 
1817
        # TODO: handle extra trees in the dirstate.
 
1818
        if (extra_trees or specific_files == []):
 
1819
            # we can't fast-path these cases (yet)
 
1820
            for f in super(InterDirStateTree, self).iter_changes(
 
1821
                include_unchanged, specific_files, pb, extra_trees,
 
1822
                require_versioned, want_unversioned=want_unversioned):
 
1823
                yield f
 
1824
            return
 
1825
        parent_ids = self.target.get_parent_ids()
 
1826
        if not (self.source._revision_id in parent_ids
 
1827
                or self.source._revision_id == NULL_REVISION):
 
1828
            raise AssertionError(
 
1829
                "revision {%s} is not stored in {%s}, but %s "
 
1830
                "can only be used for trees stored in the dirstate"
 
1831
                % (self.source._revision_id, self.target, self.iter_changes))
 
1832
        target_index = 0
 
1833
        if self.source._revision_id == NULL_REVISION:
 
1834
            source_index = None
 
1835
            indices = (target_index,)
 
1836
        else:
 
1837
            if not (self.source._revision_id in parent_ids):
 
1838
                raise AssertionError(
 
1839
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1840
                    self.source._revision_id, parent_ids))
 
1841
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1842
            indices = (source_index, target_index)
 
1843
        # -- make all specific_files utf8 --
 
1844
        if specific_files:
 
1845
            specific_files_utf8 = set()
 
1846
            for path in specific_files:
 
1847
                specific_files_utf8.add(path.encode('utf8'))
 
1848
            specific_files = specific_files_utf8
 
1849
        else:
 
1850
            specific_files = set([''])
 
1851
        # -- specific_files is now a utf8 path set --
 
1852
        # -- get the state object and prepare it.
 
1853
        state = self.target.current_dirstate()
 
1854
        state._read_dirblocks_if_needed()
 
1855
        def _entries_for_path(path):
 
1856
            """Return a list with all the entries that match path for all ids.
 
1857
            """
 
1858
            dirname, basename = os.path.split(path)
 
1859
            key = (dirname, basename, '')
 
1860
            block_index, present = state._find_block_index_from_key(key)
 
1861
            if not present:
 
1862
                # the block which should contain path is absent.
 
1863
                return []
 
1864
            result = []
 
1865
            block = state._dirblocks[block_index][1]
 
1866
            entry_index, _ = state._find_entry_index(key, block)
 
1867
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1868
            while (entry_index < len(block) and
 
1869
                block[entry_index][0][0:2] == key[0:2]):
 
1870
                result.append(block[entry_index])
 
1871
                entry_index += 1
 
1872
            return result
 
1873
        if require_versioned:
 
1874
            # -- check all supplied paths are versioned in a search tree. --
 
1875
            all_versioned = True
 
1876
            for path in specific_files:
 
1877
                path_entries = _entries_for_path(path)
 
1878
                if not path_entries:
 
1879
                    # this specified path is not present at all: error
 
1880
                    all_versioned = False
 
1881
                    break
 
1882
                found_versioned = False
 
1883
                # for each id at this path
 
1884
                for entry in path_entries:
 
1885
                    # for each tree.
 
1886
                    for index in indices:
 
1887
                        if entry[1][index][0] != 'a': # absent
 
1888
                            found_versioned = True
 
1889
                            # all good: found a versioned cell
 
1890
                            break
 
1891
                if not found_versioned:
 
1892
                    # none of the indexes was not 'absent' at all ids for this
 
1893
                    # path.
 
1894
                    all_versioned = False
 
1895
                    break
 
1896
            if not all_versioned:
 
1897
                raise errors.PathsNotVersionedError(specific_files)
 
1898
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1899
        search_specific_files = set()
 
1900
        for path in specific_files:
 
1901
            other_specific_files = specific_files.difference(set([path]))
 
1902
            if not osutils.is_inside_any(other_specific_files, path):
 
1903
                # this is a top level path, we must check it.
 
1904
                search_specific_files.add(path)
 
1905
        # sketch: 
 
1906
        # compare source_index and target_index at or under each element of search_specific_files.
 
1907
        # follow the following comparison table. Note that we only want to do diff operations when
 
1908
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1909
        # for the target.
 
1910
        # cases:
 
1911
        # 
 
1912
        # Source | Target | disk | action
 
1913
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1914
        #        |        |      | diff check on source-target
 
1915
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1916
        #        |        |      | ???
 
1917
        #   r    |  a     |      | add source to search
 
1918
        #   r    |  a     |  a   | 
 
1919
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1920
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1921
        #   a    | fdlt   |      | add new id
 
1922
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1923
        #   a    |  a     |      | not present in either tree, skip
 
1924
        #   a    |  a     |  a   | not present in any tree, skip
 
1925
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1926
        #        |        |      | may not be selected by the users list of paths.
 
1927
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1928
        #        |        |      | may not be selected by the users list of paths.
 
1929
        #  fdlt  | fdlt   |      | content in both: diff them
 
1930
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1931
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1932
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1933
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1934
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1935
        #        |        |      | this id at the other path.
 
1936
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1937
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1938
        #        |        |      | this id at the other path.
 
1939
 
 
1940
        # for all search_indexs in each path at or under each element of
 
1941
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1942
        # relocated path as one to search if its not searched already. If the
 
1943
        # detail is not relocated, add the id.
 
1944
        searched_specific_files = set()
 
1945
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1946
        # Using a list so that we can access the values and change them in
 
1947
        # nested scope. Each one is [path, file_id, entry]
 
1948
        last_source_parent = [None, None]
 
1949
        last_target_parent = [None, None]
 
1950
 
 
1951
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1952
 
 
1953
        # Just a sentry, so that _process_entry can say that this
 
1954
        # record is handled, but isn't interesting to process (unchanged)
 
1955
        uninteresting = object()
 
1956
 
 
1957
 
 
1958
        old_dirname_to_file_id = {}
 
1959
        new_dirname_to_file_id = {}
 
1960
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1961
        #       keeping a cache of directories that we have seen.
 
1962
 
 
1963
        def _process_entry(entry, path_info):
 
1964
            """Compare an entry and real disk to generate delta information.
 
1965
 
 
1966
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1967
                the path of entry. If None, then the path is considered absent.
 
1968
                (Perhaps we should pass in a concrete entry for this ?)
 
1969
                Basename is returned as a utf8 string because we expect this
 
1970
                tuple will be ignored, and don't want to take the time to
 
1971
                decode.
 
1972
            :return: None if these don't match
 
1973
                     A tuple of information about the change, or
 
1974
                     the object 'uninteresting' if these match, but are
 
1975
                     basically identical.
 
1976
            """
 
1977
            if source_index is None:
 
1978
                source_details = NULL_PARENT_DETAILS
 
1979
            else:
 
1980
                source_details = entry[1][source_index]
 
1981
            target_details = entry[1][target_index]
 
1982
            target_minikind = target_details[0]
 
1983
            if path_info is not None and target_minikind in 'fdlt':
 
1984
                if not (target_index == 0):
 
1985
                    raise AssertionError()
 
1986
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1987
                                                  stat_value=path_info[3])
 
1988
                # The entry may have been modified by update_entry
 
1989
                target_details = entry[1][target_index]
 
1990
                target_minikind = target_details[0]
 
1991
            else:
 
1992
                link_or_sha1 = None
 
1993
            file_id = entry[0][2]
 
1994
            source_minikind = source_details[0]
 
1995
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1996
                # claimed content in both: diff
 
1997
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1998
                #        |        |      | diff check on source-target
 
1999
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
2000
                #        |        |      | ???
 
2001
                if source_minikind in 'r':
 
2002
                    # add the source to the search path to find any children it
 
2003
                    # has.  TODO ? : only add if it is a container ?
 
2004
                    if not osutils.is_inside_any(searched_specific_files,
 
2005
                                                 source_details[1]):
 
2006
                        search_specific_files.add(source_details[1])
 
2007
                    # generate the old path; this is needed for stating later
 
2008
                    # as well.
 
2009
                    old_path = source_details[1]
 
2010
                    old_dirname, old_basename = os.path.split(old_path)
 
2011
                    path = pathjoin(entry[0][0], entry[0][1])
 
2012
                    old_entry = state._get_entry(source_index,
 
2013
                                                 path_utf8=old_path)
 
2014
                    # update the source details variable to be the real
 
2015
                    # location.
 
2016
                    if old_entry == (None, None):
 
2017
                        raise errors.CorruptDirstate(state._filename,
 
2018
                            "entry '%s/%s' is considered renamed from %r"
 
2019
                            " but source does not exist\n"
 
2020
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
2021
                    source_details = old_entry[1][source_index]
 
2022
                    source_minikind = source_details[0]
 
2023
                else:
 
2024
                    old_dirname = entry[0][0]
 
2025
                    old_basename = entry[0][1]
 
2026
                    old_path = path = None
 
2027
                if path_info is None:
 
2028
                    # the file is missing on disk, show as removed.
 
2029
                    content_change = True
 
2030
                    target_kind = None
 
2031
                    target_exec = False
 
2032
                else:
 
2033
                    # source and target are both versioned and disk file is present.
 
2034
                    target_kind = path_info[2]
 
2035
                    if target_kind == 'directory':
 
2036
                        if path is None:
 
2037
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2038
                        new_dirname_to_file_id[path] = file_id
 
2039
                        if source_minikind != 'd':
 
2040
                            content_change = True
 
2041
                        else:
 
2042
                            # directories have no fingerprint
 
2043
                            content_change = False
 
2044
                        target_exec = False
 
2045
                    elif target_kind == 'file':
 
2046
                        if source_minikind != 'f':
 
2047
                            content_change = True
 
2048
                        else:
 
2049
                            # We could check the size, but we already have the
 
2050
                            # sha1 hash.
 
2051
                            content_change = (link_or_sha1 != source_details[1])
 
2052
                        # Target details is updated at update_entry time
 
2053
                        if use_filesystem_for_exec:
 
2054
                            # We don't need S_ISREG here, because we are sure
 
2055
                            # we are dealing with a file.
 
2056
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2057
                        else:
 
2058
                            target_exec = target_details[3]
 
2059
                    elif target_kind == 'symlink':
 
2060
                        if source_minikind != 'l':
 
2061
                            content_change = True
 
2062
                        else:
 
2063
                            content_change = (link_or_sha1 != source_details[1])
 
2064
                        target_exec = False
 
2065
                    elif target_kind == 'tree-reference':
 
2066
                        if source_minikind != 't':
 
2067
                            content_change = True
 
2068
                        else:
 
2069
                            content_change = False
 
2070
                        target_exec = False
 
2071
                    else:
 
2072
                        raise Exception, "unknown kind %s" % path_info[2]
 
2073
                if source_minikind == 'd':
 
2074
                    if path is None:
 
2075
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2076
                    old_dirname_to_file_id[old_path] = file_id
 
2077
                # parent id is the entry for the path in the target tree
 
2078
                if old_dirname == last_source_parent[0]:
 
2079
                    source_parent_id = last_source_parent[1]
 
2080
                else:
 
2081
                    try:
 
2082
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2083
                    except KeyError:
 
2084
                        source_parent_entry = state._get_entry(source_index,
 
2085
                                                               path_utf8=old_dirname)
 
2086
                        source_parent_id = source_parent_entry[0][2]
 
2087
                    if source_parent_id == entry[0][2]:
 
2088
                        # This is the root, so the parent is None
 
2089
                        source_parent_id = None
 
2090
                    else:
 
2091
                        last_source_parent[0] = old_dirname
 
2092
                        last_source_parent[1] = source_parent_id
 
2093
                new_dirname = entry[0][0]
 
2094
                if new_dirname == last_target_parent[0]:
 
2095
                    target_parent_id = last_target_parent[1]
 
2096
                else:
 
2097
                    try:
 
2098
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2099
                    except KeyError:
 
2100
                        # TODO: We don't always need to do the lookup, because the
 
2101
                        #       parent entry will be the same as the source entry.
 
2102
                        target_parent_entry = state._get_entry(target_index,
 
2103
                                                               path_utf8=new_dirname)
 
2104
                        if target_parent_entry == (None, None):
 
2105
                            raise AssertionError(
 
2106
                                "Could not find target parent in wt: %s\nparent of: %s"
 
2107
                                % (new_dirname, entry))
 
2108
                        target_parent_id = target_parent_entry[0][2]
 
2109
                    if target_parent_id == entry[0][2]:
 
2110
                        # This is the root, so the parent is None
 
2111
                        target_parent_id = None
 
2112
                    else:
 
2113
                        last_target_parent[0] = new_dirname
 
2114
                        last_target_parent[1] = target_parent_id
 
2115
 
 
2116
                source_exec = source_details[3]
 
2117
                if (include_unchanged
 
2118
                    or content_change
 
2119
                    or source_parent_id != target_parent_id
 
2120
                    or old_basename != entry[0][1]
 
2121
                    or source_exec != target_exec
 
2122
                    ):
 
2123
                    if old_path is None:
 
2124
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2125
                        old_path_u = utf8_decode(old_path)[0]
 
2126
                        path_u = old_path_u
 
2127
                    else:
 
2128
                        old_path_u = utf8_decode(old_path)[0]
 
2129
                        if old_path == path:
 
2130
                            path_u = old_path_u
 
2131
                        else:
 
2132
                            path_u = utf8_decode(path)[0]
 
2133
                    source_kind = _minikind_to_kind[source_minikind]
 
2134
                    return (entry[0][2],
 
2135
                           (old_path_u, path_u),
 
2136
                           content_change,
 
2137
                           (True, True),
 
2138
                           (source_parent_id, target_parent_id),
 
2139
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2140
                           (source_kind, target_kind),
 
2141
                           (source_exec, target_exec))
 
2142
                else:
 
2143
                    return uninteresting
 
2144
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2145
                # looks like a new file
 
2146
                if path_info is not None:
 
2147
                    path = pathjoin(entry[0][0], entry[0][1])
 
2148
                    # parent id is the entry for the path in the target tree
 
2149
                    # TODO: these are the same for an entire directory: cache em.
 
2150
                    parent_id = state._get_entry(target_index,
 
2151
                                                 path_utf8=entry[0][0])[0][2]
 
2152
                    if parent_id == entry[0][2]:
 
2153
                        parent_id = None
 
2154
                    if use_filesystem_for_exec:
 
2155
                        # We need S_ISREG here, because we aren't sure if this
 
2156
                        # is a file or not.
 
2157
                        target_exec = bool(
 
2158
                            stat.S_ISREG(path_info[3].st_mode)
 
2159
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2160
                    else:
 
2161
                        target_exec = target_details[3]
 
2162
                    return (entry[0][2],
 
2163
                           (None, utf8_decode(path)[0]),
 
2164
                           True,
 
2165
                           (False, True),
 
2166
                           (None, parent_id),
 
2167
                           (None, utf8_decode(entry[0][1])[0]),
 
2168
                           (None, path_info[2]),
 
2169
                           (None, target_exec))
 
2170
                else:
 
2171
                    # but its not on disk: we deliberately treat this as just
 
2172
                    # never-present. (Why ?! - RBC 20070224)
 
2173
                    pass
 
2174
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2175
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2176
                # if its still on disk, *and* theres no other entry at this
 
2177
                # path [we dont know this in this routine at the moment -
 
2178
                # perhaps we should change this - then it would be an unknown.
 
2179
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2180
                # parent id is the entry for the path in the target tree
 
2181
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2182
                if parent_id == entry[0][2]:
 
2183
                    parent_id = None
 
2184
                return (entry[0][2],
 
2185
                       (utf8_decode(old_path)[0], None),
 
2186
                       True,
 
2187
                       (True, False),
 
2188
                       (parent_id, None),
 
2189
                       (utf8_decode(entry[0][1])[0], None),
 
2190
                       (_minikind_to_kind[source_minikind], None),
 
2191
                       (source_details[3], None))
 
2192
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2193
                # a rename; could be a true rename, or a rename inherited from
 
2194
                # a renamed parent. TODO: handle this efficiently. Its not
 
2195
                # common case to rename dirs though, so a correct but slow
 
2196
                # implementation will do.
 
2197
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2198
                    search_specific_files.add(target_details[1])
 
2199
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2200
                # neither of the selected trees contain this file,
 
2201
                # so skip over it. This is not currently directly tested, but
 
2202
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2203
                pass
 
2204
            else:
 
2205
                raise AssertionError("don't know how to compare "
 
2206
                    "source_minikind=%r, target_minikind=%r"
 
2207
                    % (source_minikind, target_minikind))
 
2208
                ## import pdb;pdb.set_trace()
 
2209
            return None
 
2210
 
 
2211
        while search_specific_files:
 
2212
            # TODO: the pending list should be lexically sorted?  the
 
2213
            # interface doesn't require it.
 
2214
            current_root = search_specific_files.pop()
 
2215
            current_root_unicode = current_root.decode('utf8')
 
2216
            searched_specific_files.add(current_root)
 
2217
            # process the entries for this containing directory: the rest will be
 
2218
            # found by their parents recursively.
 
2219
            root_entries = _entries_for_path(current_root)
 
2220
            root_abspath = self.target.abspath(current_root_unicode)
 
2221
            try:
 
2222
                root_stat = os.lstat(root_abspath)
 
2223
            except OSError, e:
 
2224
                if e.errno == errno.ENOENT:
 
2225
                    # the path does not exist: let _process_entry know that.
 
2226
                    root_dir_info = None
 
2227
                else:
 
2228
                    # some other random error: hand it up.
 
2229
                    raise
 
2230
            else:
 
2231
                root_dir_info = ('', current_root,
 
2232
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2233
                    root_abspath)
 
2234
                if root_dir_info[2] == 'directory':
 
2235
                    if self.target._directory_is_tree_reference(
 
2236
                        current_root.decode('utf8')):
 
2237
                        root_dir_info = root_dir_info[:2] + \
 
2238
                            ('tree-reference',) + root_dir_info[3:]
 
2239
 
 
2240
            if not root_entries and not root_dir_info:
 
2241
                # this specified path is not present at all, skip it.
 
2242
                continue
 
2243
            path_handled = False
 
2244
            for entry in root_entries:
 
2245
                result = _process_entry(entry, root_dir_info)
 
2246
                if result is not None:
 
2247
                    path_handled = True
 
2248
                    if result is not uninteresting:
 
2249
                        yield result
 
2250
            if want_unversioned and not path_handled and root_dir_info:
 
2251
                new_executable = bool(
 
2252
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2253
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2254
                yield (None,
 
2255
                       (None, current_root_unicode),
 
2256
                       True,
 
2257
                       (False, False),
 
2258
                       (None, None),
 
2259
                       (None, splitpath(current_root_unicode)[-1]),
 
2260
                       (None, root_dir_info[2]),
 
2261
                       (None, new_executable)
 
2262
                      )
 
2263
            initial_key = (current_root, '', '')
 
2264
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2265
            if block_index == 0:
 
2266
                # we have processed the total root already, but because the
 
2267
                # initial key matched it we should skip it here.
 
2268
                block_index +=1
 
2269
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2270
                current_dir_info = None
 
2271
            else:
 
2272
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2273
                try:
 
2274
                    current_dir_info = dir_iterator.next()
 
2275
                except OSError, e:
 
2276
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2277
                    # python 2.5 has e.errno == EINVAL,
 
2278
                    #            and e.winerror == ERROR_DIRECTORY
 
2279
                    e_winerror = getattr(e, 'winerror', None)
 
2280
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2281
                    # there may be directories in the inventory even though
 
2282
                    # this path is not a file on disk: so mark it as end of
 
2283
                    # iterator
 
2284
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2285
                        current_dir_info = None
 
2286
                    elif (sys.platform == 'win32'
 
2287
                          and (e.errno in win_errors
 
2288
                               or e_winerror in win_errors)):
 
2289
                        current_dir_info = None
 
2290
                    else:
 
2291
                        raise
 
2292
                else:
 
2293
                    if current_dir_info[0][0] == '':
 
2294
                        # remove .bzr from iteration
 
2295
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2296
                        if current_dir_info[1][bzr_index][0] != '.bzr':
 
2297
                            raise AssertionError()
 
2298
                        del current_dir_info[1][bzr_index]
 
2299
            # walk until both the directory listing and the versioned metadata
 
2300
            # are exhausted. 
 
2301
            if (block_index < len(state._dirblocks) and
 
2302
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2303
                current_block = state._dirblocks[block_index]
 
2304
            else:
 
2305
                current_block = None
 
2306
            while (current_dir_info is not None or
 
2307
                   current_block is not None):
 
2308
                if (current_dir_info and current_block
 
2309
                    and current_dir_info[0][0] != current_block[0]):
 
2310
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2311
                        # filesystem data refers to paths not covered by the dirblock.
 
2312
                        # this has two possibilities:
 
2313
                        # A) it is versioned but empty, so there is no block for it
 
2314
                        # B) it is not versioned.
 
2315
 
 
2316
                        # if (A) then we need to recurse into it to check for
 
2317
                        # new unknown files or directories.
 
2318
                        # if (B) then we should ignore it, because we don't
 
2319
                        # recurse into unknown directories.
 
2320
                        path_index = 0
 
2321
                        while path_index < len(current_dir_info[1]):
 
2322
                                current_path_info = current_dir_info[1][path_index]
 
2323
                                if want_unversioned:
 
2324
                                    if current_path_info[2] == 'directory':
 
2325
                                        if self.target._directory_is_tree_reference(
 
2326
                                            current_path_info[0].decode('utf8')):
 
2327
                                            current_path_info = current_path_info[:2] + \
 
2328
                                                ('tree-reference',) + current_path_info[3:]
 
2329
                                    new_executable = bool(
 
2330
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2331
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2332
                                    yield (None,
 
2333
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2334
                                        True,
 
2335
                                        (False, False),
 
2336
                                        (None, None),
 
2337
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2338
                                        (None, current_path_info[2]),
 
2339
                                        (None, new_executable))
 
2340
                                # dont descend into this unversioned path if it is
 
2341
                                # a dir
 
2342
                                if current_path_info[2] in ('directory',
 
2343
                                                            'tree-reference'):
 
2344
                                    del current_dir_info[1][path_index]
 
2345
                                    path_index -= 1
 
2346
                                path_index += 1
 
2347
 
 
2348
                        # This dir info has been handled, go to the next
 
2349
                        try:
 
2350
                            current_dir_info = dir_iterator.next()
 
2351
                        except StopIteration:
 
2352
                            current_dir_info = None
 
2353
                    else:
 
2354
                        # We have a dirblock entry for this location, but there
 
2355
                        # is no filesystem path for this. This is most likely
 
2356
                        # because a directory was removed from the disk.
 
2357
                        # We don't have to report the missing directory,
 
2358
                        # because that should have already been handled, but we
 
2359
                        # need to handle all of the files that are contained
 
2360
                        # within.
 
2361
                        for current_entry in current_block[1]:
 
2362
                            # entry referring to file not present on disk.
 
2363
                            # advance the entry only, after processing.
 
2364
                            result = _process_entry(current_entry, None)
 
2365
                            if result is not None:
 
2366
                                if result is not uninteresting:
 
2367
                                    yield result
 
2368
                        block_index +=1
 
2369
                        if (block_index < len(state._dirblocks) and
 
2370
                            osutils.is_inside(current_root,
 
2371
                                              state._dirblocks[block_index][0])):
 
2372
                            current_block = state._dirblocks[block_index]
 
2373
                        else:
 
2374
                            current_block = None
 
2375
                    continue
 
2376
                entry_index = 0
 
2377
                if current_block and entry_index < len(current_block[1]):
 
2378
                    current_entry = current_block[1][entry_index]
 
2379
                else:
 
2380
                    current_entry = None
 
2381
                advance_entry = True
 
2382
                path_index = 0
 
2383
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2384
                    current_path_info = current_dir_info[1][path_index]
 
2385
                    if current_path_info[2] == 'directory':
 
2386
                        if self.target._directory_is_tree_reference(
 
2387
                            current_path_info[0].decode('utf8')):
 
2388
                            current_path_info = current_path_info[:2] + \
 
2389
                                ('tree-reference',) + current_path_info[3:]
 
2390
                else:
 
2391
                    current_path_info = None
 
2392
                advance_path = True
 
2393
                path_handled = False
 
2394
                while (current_entry is not None or
 
2395
                    current_path_info is not None):
 
2396
                    if current_entry is None:
 
2397
                        # the check for path_handled when the path is adnvaced
 
2398
                        # will yield this path if needed.
 
2399
                        pass
 
2400
                    elif current_path_info is None:
 
2401
                        # no path is fine: the per entry code will handle it.
 
2402
                        result = _process_entry(current_entry, current_path_info)
 
2403
                        if result is not None:
 
2404
                            if result is not uninteresting:
 
2405
                                yield result
 
2406
                    elif (current_entry[0][1] != current_path_info[1]
 
2407
                          or current_entry[1][target_index][0] in 'ar'):
 
2408
                        # The current path on disk doesn't match the dirblock
 
2409
                        # record. Either the dirblock is marked as absent, or
 
2410
                        # the file on disk is not present at all in the
 
2411
                        # dirblock. Either way, report about the dirblock
 
2412
                        # entry, and let other code handle the filesystem one.
 
2413
 
 
2414
                        # Compare the basename for these files to determine
 
2415
                        # which comes first
 
2416
                        if current_path_info[1] < current_entry[0][1]:
 
2417
                            # extra file on disk: pass for now, but only
 
2418
                            # increment the path, not the entry
 
2419
                            advance_entry = False
 
2420
                        else:
 
2421
                            # entry referring to file not present on disk.
 
2422
                            # advance the entry only, after processing.
 
2423
                            result = _process_entry(current_entry, None)
 
2424
                            if result is not None:
 
2425
                                if result is not uninteresting:
 
2426
                                    yield result
 
2427
                            advance_path = False
 
2428
                    else:
 
2429
                        result = _process_entry(current_entry, current_path_info)
 
2430
                        if result is not None:
 
2431
                            path_handled = True
 
2432
                            if result is not uninteresting:
 
2433
                                yield result
 
2434
                    if advance_entry and current_entry is not None:
 
2435
                        entry_index += 1
 
2436
                        if entry_index < len(current_block[1]):
 
2437
                            current_entry = current_block[1][entry_index]
 
2438
                        else:
 
2439
                            current_entry = None
 
2440
                    else:
 
2441
                        advance_entry = True # reset the advance flaga
 
2442
                    if advance_path and current_path_info is not None:
 
2443
                        if not path_handled:
 
2444
                            # unversioned in all regards
 
2445
                            if want_unversioned:
 
2446
                                new_executable = bool(
 
2447
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2448
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2449
                                yield (None,
 
2450
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2451
                                    True,
 
2452
                                    (False, False),
 
2453
                                    (None, None),
 
2454
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2455
                                    (None, current_path_info[2]),
 
2456
                                    (None, new_executable))
 
2457
                            # dont descend into this unversioned path if it is
 
2458
                            # a dir
 
2459
                            if current_path_info[2] in ('directory'):
 
2460
                                del current_dir_info[1][path_index]
 
2461
                                path_index -= 1
 
2462
                        # dont descend the disk iterator into any tree 
 
2463
                        # paths.
 
2464
                        if current_path_info[2] == 'tree-reference':
 
2465
                            del current_dir_info[1][path_index]
 
2466
                            path_index -= 1
 
2467
                        path_index += 1
 
2468
                        if path_index < len(current_dir_info[1]):
 
2469
                            current_path_info = current_dir_info[1][path_index]
 
2470
                            if current_path_info[2] == 'directory':
 
2471
                                if self.target._directory_is_tree_reference(
 
2472
                                    current_path_info[0].decode('utf8')):
 
2473
                                    current_path_info = current_path_info[:2] + \
 
2474
                                        ('tree-reference',) + current_path_info[3:]
 
2475
                        else:
 
2476
                            current_path_info = None
 
2477
                        path_handled = False
 
2478
                    else:
 
2479
                        advance_path = True # reset the advance flagg.
 
2480
                if current_block is not None:
 
2481
                    block_index += 1
 
2482
                    if (block_index < len(state._dirblocks) and
 
2483
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2484
                        current_block = state._dirblocks[block_index]
 
2485
                    else:
 
2486
                        current_block = None
 
2487
                if current_dir_info is not None:
 
2488
                    try:
 
2489
                        current_dir_info = dir_iterator.next()
 
2490
                    except StopIteration:
 
2491
                        current_dir_info = None
 
2492
 
 
2493
 
 
2494
    @staticmethod
 
2495
    def is_compatible(source, target):
 
2496
        # the target must be a dirstate working tree
 
2497
        if not isinstance(target, WorkingTree4):
 
2498
            return False
 
2499
        # the source must be a revtreee or dirstate rev tree.
 
2500
        if not isinstance(source,
 
2501
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2502
            return False
 
2503
        # the source revid must be in the target dirstate
 
2504
        if not (source._revision_id == NULL_REVISION or
 
2505
            source._revision_id in target.get_parent_ids()):
 
2506
            # TODO: what about ghosts? it may well need to 
 
2507
            # check for them explicitly.
 
2508
            return False
 
2509
        return True
 
2510
 
 
2511
InterTree.register_optimiser(InterDirStateTree)
 
2512
 
 
2513
 
 
2514
class Converter3to4(object):
 
2515
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2516
 
 
2517
    def __init__(self):
 
2518
        self.target_format = WorkingTreeFormat4()
 
2519
 
 
2520
    def convert(self, tree):
 
2521
        # lock the control files not the tree, so that we dont get tree
 
2522
        # on-unlock behaviours, and so that noone else diddles with the 
 
2523
        # tree during upgrade.
 
2524
        tree._control_files.lock_write()
 
2525
        try:
 
2526
            tree.read_working_inventory()
 
2527
            self.create_dirstate_data(tree)
 
2528
            self.update_format(tree)
 
2529
            self.remove_xml_files(tree)
 
2530
        finally:
 
2531
            tree._control_files.unlock()
 
2532
 
 
2533
    def create_dirstate_data(self, tree):
 
2534
        """Create the dirstate based data for tree."""
 
2535
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2536
            ).local_abspath('dirstate')
 
2537
        state = dirstate.DirState.from_tree(tree, local_path)
 
2538
        state.save()
 
2539
        state.unlock()
 
2540
 
 
2541
    def remove_xml_files(self, tree):
 
2542
        """Remove the oldformat 3 data."""
 
2543
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2544
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2545
            'pending-merges', 'stat-cache']:
 
2546
            try:
 
2547
                transport.delete(path)
 
2548
            except errors.NoSuchFile:
 
2549
                # some files are optional - just deal.
 
2550
                pass
 
2551
 
 
2552
    def update_format(self, tree):
 
2553
        """Change the format marker."""
 
2554
        tree._transport.put_bytes('format',
 
2555
            self.target_format.get_format_string(),
 
2556
            mode=tree.bzrdir._get_file_mode())