/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Martin Pool
  • Date: 2007-10-12 08:00:07 UTC
  • mto: This revision was merged to the branch mainline in revision 2913.
  • Revision ID: mbp@sourcefrog.net-20071012080007-vf80woayyom8s8e1
Rename update_to_one_parent_via_delta to more wieldy update_basis_by_delta

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    debug,
 
47
    delta,
 
48
    dirstate,
 
49
    errors,
 
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    trace,
 
59
    transform,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
 
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
 
67
""")
 
68
 
 
69
from bzrlib import symbol_versioning
 
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
 
76
from bzrlib.osutils import (
 
77
    file_kind,
 
78
    isdir,
 
79
    normpath,
 
80
    pathjoin,
 
81
    rand_chars,
 
82
    realpath,
 
83
    safe_unicode,
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
 
87
from bzrlib.transport.local import LocalTransport
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
 
120
    def __init__(self, basedir,
 
121
                 branch,
 
122
                 _control_files=None,
 
123
                 _format=None,
 
124
                 _bzrdir=None):
 
125
        """Construct a WorkingTree for basedir.
 
126
 
 
127
        If the branch is not supplied, it is opened automatically.
 
128
        If the branch is supplied, it must be the branch for this basedir.
 
129
        (branch.base is not cross checked, because for remote branches that
 
130
        would be meaningless).
 
131
        """
 
132
        self._format = _format
 
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
 
136
        basedir = safe_unicode(basedir)
 
137
        mutter("opening working tree %r", basedir)
 
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
 
141
        self.basedir = realpath(basedir)
 
142
        # if branch is at our basedir and is a format 6 or less
 
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
 
146
        self._control_files = _control_files
 
147
        self._dirty = None
 
148
        #-------------
 
149
        # during a read or write lock these objects are set, and are
 
150
        # None the rest of the time.
 
151
        self._dirstate = None
 
152
        self._inventory = None
 
153
        #-------------
 
154
 
 
155
    @needs_tree_write_lock
 
156
    def _add(self, files, ids, kinds):
 
157
        """See MutableTree._add."""
 
158
        state = self.current_dirstate()
 
159
        for f, file_id, kind in zip(files, ids, kinds):
 
160
            f = f.strip('/')
 
161
            if self.path2id(f):
 
162
                # special case tree root handling.
 
163
                if f == '' and self.path2id(f) == ROOT_ID:
 
164
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
165
                continue
 
166
            if file_id is None:
 
167
                file_id = generate_ids.gen_file_id(f)
 
168
            # deliberately add the file with no cached stat or sha1
 
169
            # - on the first access it will be gathered, and we can
 
170
            # always change this once tests are all passing.
 
171
            state.add(f, file_id, kind, None, '')
 
172
        self._make_dirty(reset_inventory=True)
 
173
 
 
174
    def _make_dirty(self, reset_inventory):
 
175
        """Make the tree state dirty.
 
176
 
 
177
        :param reset_inventory: True if the cached inventory should be removed
 
178
            (presuming there is one).
 
179
        """
 
180
        self._dirty = True
 
181
        if reset_inventory and self._inventory is not None:
 
182
            self._inventory = None
 
183
 
 
184
    @needs_tree_write_lock
 
185
    def add_reference(self, sub_tree):
 
186
        # use standard implementation, which calls back to self._add
 
187
        # 
 
188
        # So we don't store the reference_revision in the working dirstate,
 
189
        # it's just recorded at the moment of commit. 
 
190
        self._add_reference(sub_tree)
 
191
 
 
192
    def break_lock(self):
 
193
        """Break a lock if one is present from another instance.
 
194
 
 
195
        Uses the ui factory to ask for confirmation if the lock may be from
 
196
        an active process.
 
197
 
 
198
        This will probe the repository for its lock as well.
 
199
        """
 
200
        # if the dirstate is locked by an active process, reject the break lock
 
201
        # call.
 
202
        try:
 
203
            if self._dirstate is None:
 
204
                clear = True
 
205
            else:
 
206
                clear = False
 
207
            state = self._current_dirstate()
 
208
            if state._lock_token is not None:
 
209
                # we already have it locked. sheese, cant break our own lock.
 
210
                raise errors.LockActive(self.basedir)
 
211
            else:
 
212
                try:
 
213
                    # try for a write lock - need permission to get one anyhow
 
214
                    # to break locks.
 
215
                    state.lock_write()
 
216
                except errors.LockContention:
 
217
                    # oslocks fail when a process is still live: fail.
 
218
                    # TODO: get the locked lockdir info and give to the user to
 
219
                    # assist in debugging.
 
220
                    raise errors.LockActive(self.basedir)
 
221
                else:
 
222
                    state.unlock()
 
223
        finally:
 
224
            if clear:
 
225
                self._dirstate = None
 
226
        self._control_files.break_lock()
 
227
        self.branch.break_lock()
 
228
 
 
229
    def _comparison_data(self, entry, path):
 
230
        kind, executable, stat_value = \
 
231
            WorkingTree3._comparison_data(self, entry, path)
 
232
        # it looks like a plain directory, but it's really a reference -- see
 
233
        # also kind()
 
234
        if (self._repo_supports_tree_reference and
 
235
            kind == 'directory' and
 
236
            self._directory_is_tree_reference(path)):
 
237
            kind = 'tree-reference'
 
238
        return kind, executable, stat_value
 
239
 
 
240
    @needs_write_lock
 
241
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
242
        # mark the tree as dirty post commit - commit
 
243
        # can change the current versioned list by doing deletes.
 
244
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
245
        self._make_dirty(reset_inventory=True)
 
246
        return result
 
247
 
 
248
    def current_dirstate(self):
 
249
        """Return the current dirstate object.
 
250
 
 
251
        This is not part of the tree interface and only exposed for ease of
 
252
        testing.
 
253
 
 
254
        :raises errors.NotWriteLocked: when not in a lock.
 
255
        """
 
256
        self._must_be_locked()
 
257
        return self._current_dirstate()
 
258
 
 
259
    def _current_dirstate(self):
 
260
        """Internal function that does not check lock status.
 
261
 
 
262
        This is needed for break_lock which also needs the dirstate.
 
263
        """
 
264
        if self._dirstate is not None:
 
265
            return self._dirstate
 
266
        local_path = self.bzrdir.get_workingtree_transport(None
 
267
            ).local_abspath('dirstate')
 
268
        self._dirstate = dirstate.DirState.on_file(local_path)
 
269
        return self._dirstate
 
270
 
 
271
    def filter_unversioned_files(self, paths):
 
272
        """Filter out paths that are versioned.
 
273
 
 
274
        :return: set of paths.
 
275
        """
 
276
        # TODO: make a generic multi-bisect routine roughly that should list
 
277
        # the paths, then process one half at a time recursively, and feed the
 
278
        # results of each bisect in further still
 
279
        paths = sorted(paths)
 
280
        result = set()
 
281
        state = self.current_dirstate()
 
282
        # TODO we want a paths_to_dirblocks helper I think
 
283
        for path in paths:
 
284
            dirname, basename = os.path.split(path.encode('utf8'))
 
285
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
286
                dirname, basename, 0)
 
287
            if not path_is_versioned:
 
288
                result.add(path)
 
289
        return result
 
290
 
 
291
    def flush(self):
 
292
        """Write all cached data to disk."""
 
293
        if self._control_files._lock_mode != 'w':
 
294
            raise errors.NotWriteLocked(self)
 
295
        self.current_dirstate().save()
 
296
        self._inventory = None
 
297
        self._dirty = False
 
298
 
 
299
    @needs_tree_write_lock
 
300
    def _gather_kinds(self, files, kinds):
 
301
        """See MutableTree._gather_kinds."""
 
302
        for pos, f in enumerate(files):
 
303
            if kinds[pos] is None:
 
304
                kinds[pos] = self._kind(f)
 
305
 
 
306
    def _generate_inventory(self):
 
307
        """Create and set self.inventory from the dirstate object.
 
308
        
 
309
        This is relatively expensive: we have to walk the entire dirstate.
 
310
        Ideally we would not, and can deprecate this function.
 
311
        """
 
312
        #: uncomment to trap on inventory requests.
 
313
        # import pdb;pdb.set_trace()
 
314
        state = self.current_dirstate()
 
315
        state._read_dirblocks_if_needed()
 
316
        root_key, current_entry = self._get_entry(path='')
 
317
        current_id = root_key[2]
 
318
        assert current_entry[0][0] == 'd' # directory
 
319
        inv = Inventory(root_id=current_id)
 
320
        # Turn some things into local variables
 
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
322
        factory = entry_factory
 
323
        utf8_decode = cache_utf8._utf8_decode
 
324
        inv_byid = inv._byid
 
325
        # we could do this straight out of the dirstate; it might be fast
 
326
        # and should be profiled - RBC 20070216
 
327
        parent_ies = {'' : inv.root}
 
328
        for block in state._dirblocks[1:]: # skip the root
 
329
            dirname = block[0]
 
330
            try:
 
331
                parent_ie = parent_ies[dirname]
 
332
            except KeyError:
 
333
                # all the paths in this block are not versioned in this tree
 
334
                continue
 
335
            for key, entry in block[1]:
 
336
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
337
                if minikind in ('a', 'r'): # absent, relocated
 
338
                    # a parent tree only entry
 
339
                    continue
 
340
                name = key[1]
 
341
                name_unicode = utf8_decode(name)[0]
 
342
                file_id = key[2]
 
343
                kind = minikind_to_kind[minikind]
 
344
                inv_entry = factory[kind](file_id, name_unicode,
 
345
                                          parent_ie.file_id)
 
346
                if kind == 'file':
 
347
                    # This is only needed on win32, where this is the only way
 
348
                    # we know the executable bit.
 
349
                    inv_entry.executable = executable
 
350
                    # not strictly needed: working tree
 
351
                    #inv_entry.text_size = size
 
352
                    #inv_entry.text_sha1 = sha1
 
353
                elif kind == 'directory':
 
354
                    # add this entry to the parent map.
 
355
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
356
                elif kind == 'tree-reference':
 
357
                    assert self._repo_supports_tree_reference, \
 
358
                        "repository of %r " \
 
359
                        "doesn't support tree references " \
 
360
                        "required by entry %r" \
 
361
                        % (self, name)
 
362
                    inv_entry.reference_revision = link_or_sha1 or None
 
363
                elif kind != 'symlink':
 
364
                    raise AssertionError("unknown kind %r" % kind)
 
365
                # These checks cost us around 40ms on a 55k entry tree
 
366
                assert file_id not in inv_byid, ('file_id %s already in'
 
367
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                assert name_unicode not in parent_ie.children
 
369
                inv_byid[file_id] = inv_entry
 
370
                parent_ie.children[name_unicode] = inv_entry
 
371
        self._inventory = inv
 
372
 
 
373
    def _get_entry(self, file_id=None, path=None):
 
374
        """Get the dirstate row for file_id or path.
 
375
 
 
376
        If either file_id or path is supplied, it is used as the key to lookup.
 
377
        If both are supplied, the fastest lookup is used, and an error is
 
378
        raised if they do not both point at the same row.
 
379
        
 
380
        :param file_id: An optional unicode file_id to be looked up.
 
381
        :param path: An optional unicode path to be looked up.
 
382
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
383
        """
 
384
        if file_id is None and path is None:
 
385
            raise errors.BzrError('must supply file_id or path')
 
386
        state = self.current_dirstate()
 
387
        if path is not None:
 
388
            path = path.encode('utf8')
 
389
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
390
 
 
391
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
392
        # check file id is valid unconditionally.
 
393
        entry = self._get_entry(file_id=file_id, path=path)
 
394
        if entry[0] is None:
 
395
            raise errors.NoSuchId(self, file_id)
 
396
        if path is None:
 
397
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
398
 
 
399
        file_abspath = self.abspath(path)
 
400
        state = self.current_dirstate()
 
401
        if stat_value is None:
 
402
            try:
 
403
                stat_value = os.lstat(file_abspath)
 
404
            except OSError, e:
 
405
                if e.errno == errno.ENOENT:
 
406
                    return None
 
407
                else:
 
408
                    raise
 
409
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
410
                                          stat_value=stat_value)
 
411
        if entry[1][0][0] == 'f':
 
412
            return link_or_sha1
 
413
        return None
 
414
 
 
415
    def _get_inventory(self):
 
416
        """Get the inventory for the tree. This is only valid within a lock."""
 
417
        if 'evil' in debug.debug_flags:
 
418
            trace.mutter_callsite(2,
 
419
                "accessing .inventory forces a size of tree translation.")
 
420
        if self._inventory is not None:
 
421
            return self._inventory
 
422
        self._must_be_locked()
 
423
        self._generate_inventory()
 
424
        return self._inventory
 
425
 
 
426
    inventory = property(_get_inventory,
 
427
                         doc="Inventory of this Tree")
 
428
 
 
429
    @needs_read_lock
 
430
    def get_parent_ids(self):
 
431
        """See Tree.get_parent_ids.
 
432
        
 
433
        This implementation requests the ids list from the dirstate file.
 
434
        """
 
435
        return self.current_dirstate().get_parent_ids()
 
436
 
 
437
    def get_reference_revision(self, file_id, path=None):
 
438
        # referenced tree's revision is whatever's currently there
 
439
        return self.get_nested_tree(file_id, path).last_revision()
 
440
 
 
441
    def get_nested_tree(self, file_id, path=None):
 
442
        if path is None:
 
443
            path = self.id2path(file_id)
 
444
        # else: check file_id is at path?
 
445
        return WorkingTree.open(self.abspath(path))
 
446
 
 
447
    @needs_read_lock
 
448
    def get_root_id(self):
 
449
        """Return the id of this trees root"""
 
450
        return self._get_entry(path='')[0][2]
 
451
 
 
452
    def has_id(self, file_id):
 
453
        state = self.current_dirstate()
 
454
        row, parents = self._get_entry(file_id=file_id)
 
455
        if row is None:
 
456
            return False
 
457
        return osutils.lexists(pathjoin(
 
458
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
459
 
 
460
    @needs_read_lock
 
461
    def id2path(self, file_id):
 
462
        "Convert a file-id to a path."
 
463
        state = self.current_dirstate()
 
464
        entry = self._get_entry(file_id=file_id)
 
465
        if entry == (None, None):
 
466
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
467
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
468
        return path_utf8.decode('utf8')
 
469
 
 
470
    if not osutils.supports_executable():
 
471
        def is_executable(self, file_id, path=None):
 
472
            """Test if a file is executable or not.
 
473
 
 
474
            Note: The caller is expected to take a read-lock before calling this.
 
475
            """
 
476
            entry = self._get_entry(file_id=file_id, path=path)
 
477
            if entry == (None, None):
 
478
                return False
 
479
            return entry[1][0][3]
 
480
    else:
 
481
        def is_executable(self, file_id, path=None):
 
482
            """Test if a file is executable or not.
 
483
 
 
484
            Note: The caller is expected to take a read-lock before calling this.
 
485
            """
 
486
            if not path:
 
487
                path = self.id2path(file_id)
 
488
            mode = os.lstat(self.abspath(path)).st_mode
 
489
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
490
 
 
491
    @needs_read_lock
 
492
    def __iter__(self):
 
493
        """Iterate through file_ids for this tree.
 
494
 
 
495
        file_ids are in a WorkingTree if they are in the working inventory
 
496
        and the working file exists.
 
497
        """
 
498
        result = []
 
499
        for key, tree_details in self.current_dirstate()._iter_entries():
 
500
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
501
                # not relevant to the working tree
 
502
                continue
 
503
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
504
            if osutils.lexists(path):
 
505
                result.append(key[2])
 
506
        return iter(result)
 
507
 
 
508
    def iter_references(self):
 
509
        for key, tree_details in self.current_dirstate()._iter_entries():
 
510
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
511
                # not relevant to the working tree
 
512
                continue
 
513
            if not key[1]:
 
514
                # the root is not a reference.
 
515
                continue
 
516
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
517
            try:
 
518
                if self._kind(path) == 'tree-reference':
 
519
                    yield path, key[2]
 
520
            except errors.NoSuchFile:
 
521
                # path is missing on disk.
 
522
                continue
 
523
 
 
524
    def kind(self, file_id):
 
525
        """Return the kind of a file.
 
526
 
 
527
        This is always the actual kind that's on disk, regardless of what it
 
528
        was added as.
 
529
 
 
530
        Note: The caller is expected to take a read-lock before calling this.
 
531
        """
 
532
        relpath = self.id2path(file_id)
 
533
        assert relpath != None, \
 
534
            "path for id {%s} is None!" % file_id
 
535
        return self._kind(relpath)
 
536
 
 
537
    def _kind(self, relpath):
 
538
        abspath = self.abspath(relpath)
 
539
        kind = file_kind(abspath)
 
540
        if (self._repo_supports_tree_reference and
 
541
            kind == 'directory' and
 
542
            self._directory_is_tree_reference(relpath)):
 
543
            kind = 'tree-reference'
 
544
        return kind
 
545
 
 
546
    @needs_read_lock
 
547
    def _last_revision(self):
 
548
        """See Mutable.last_revision."""
 
549
        parent_ids = self.current_dirstate().get_parent_ids()
 
550
        if parent_ids:
 
551
            return parent_ids[0]
 
552
        else:
 
553
            return _mod_revision.NULL_REVISION
 
554
 
 
555
    def lock_read(self):
 
556
        """See Branch.lock_read, and WorkingTree.unlock."""
 
557
        self.branch.lock_read()
 
558
        try:
 
559
            self._control_files.lock_read()
 
560
            try:
 
561
                state = self.current_dirstate()
 
562
                if not state._lock_token:
 
563
                    state.lock_read()
 
564
                # set our support for tree references from the repository in
 
565
                # use.
 
566
                self._repo_supports_tree_reference = getattr(
 
567
                    self.branch.repository._format, "supports_tree_reference",
 
568
                    False)
 
569
            except:
 
570
                self._control_files.unlock()
 
571
                raise
 
572
        except:
 
573
            self.branch.unlock()
 
574
            raise
 
575
 
 
576
    def _lock_self_write(self):
 
577
        """This should be called after the branch is locked."""
 
578
        try:
 
579
            self._control_files.lock_write()
 
580
            try:
 
581
                state = self.current_dirstate()
 
582
                if not state._lock_token:
 
583
                    state.lock_write()
 
584
                # set our support for tree references from the repository in
 
585
                # use.
 
586
                self._repo_supports_tree_reference = getattr(
 
587
                    self.branch.repository._format, "supports_tree_reference",
 
588
                    False)
 
589
            except:
 
590
                self._control_files.unlock()
 
591
                raise
 
592
        except:
 
593
            self.branch.unlock()
 
594
            raise
 
595
 
 
596
    def lock_tree_write(self):
 
597
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
598
        self.branch.lock_read()
 
599
        self._lock_self_write()
 
600
 
 
601
    def lock_write(self):
 
602
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
603
        self.branch.lock_write()
 
604
        self._lock_self_write()
 
605
 
 
606
    @needs_tree_write_lock
 
607
    def move(self, from_paths, to_dir, after=False):
 
608
        """See WorkingTree.move()."""
 
609
        result = []
 
610
        if not from_paths:
 
611
            return result
 
612
 
 
613
        state = self.current_dirstate()
 
614
 
 
615
        assert not isinstance(from_paths, basestring)
 
616
        to_dir_utf8 = to_dir.encode('utf8')
 
617
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
618
        id_index = state._get_id_index()
 
619
        # check destination directory
 
620
        # get the details for it
 
621
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
622
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
623
        if not entry_present:
 
624
            raise errors.BzrMoveFailedError('', to_dir,
 
625
                errors.NotVersionedError(to_dir))
 
626
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
627
        # get a handle on the block itself.
 
628
        to_block_index = state._ensure_block(
 
629
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
630
        to_block = state._dirblocks[to_block_index]
 
631
        to_abs = self.abspath(to_dir)
 
632
        if not isdir(to_abs):
 
633
            raise errors.BzrMoveFailedError('',to_dir,
 
634
                errors.NotADirectory(to_abs))
 
635
 
 
636
        if to_entry[1][0][0] != 'd':
 
637
            raise errors.BzrMoveFailedError('',to_dir,
 
638
                errors.NotADirectory(to_abs))
 
639
 
 
640
        if self._inventory is not None:
 
641
            update_inventory = True
 
642
            inv = self.inventory
 
643
            to_dir_ie = inv[to_dir_id]
 
644
            to_dir_id = to_entry[0][2]
 
645
        else:
 
646
            update_inventory = False
 
647
 
 
648
        rollbacks = []
 
649
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
650
                     fingerprint, packed_stat, size,
 
651
                     to_block, to_key, to_path_utf8):
 
652
            state._make_absent(old_entry)
 
653
            from_key = old_entry[0]
 
654
            rollbacks.append(
 
655
                lambda:state.update_minimal(from_key,
 
656
                    minikind,
 
657
                    executable=executable,
 
658
                    fingerprint=fingerprint,
 
659
                    packed_stat=packed_stat,
 
660
                    size=size,
 
661
                    path_utf8=from_path_utf8))
 
662
            state.update_minimal(to_key,
 
663
                    minikind,
 
664
                    executable=executable,
 
665
                    fingerprint=fingerprint,
 
666
                    packed_stat=packed_stat,
 
667
                    size=size,
 
668
                    path_utf8=to_path_utf8)
 
669
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
670
            new_entry = to_block[1][added_entry_index]
 
671
            rollbacks.append(lambda:state._make_absent(new_entry))
 
672
 
 
673
        for from_rel in from_paths:
 
674
            # from_rel is 'pathinroot/foo/bar'
 
675
            from_rel_utf8 = from_rel.encode('utf8')
 
676
            from_dirname, from_tail = osutils.split(from_rel)
 
677
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
678
            from_entry = self._get_entry(path=from_rel)
 
679
            if from_entry == (None, None):
 
680
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
681
                    errors.NotVersionedError(path=str(from_rel)))
 
682
 
 
683
            from_id = from_entry[0][2]
 
684
            to_rel = pathjoin(to_dir, from_tail)
 
685
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
686
            item_to_entry = self._get_entry(path=to_rel)
 
687
            if item_to_entry != (None, None):
 
688
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
689
                    "Target is already versioned.")
 
690
 
 
691
            if from_rel == to_rel:
 
692
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
693
                    "Source and target are identical.")
 
694
 
 
695
            from_missing = not self.has_filename(from_rel)
 
696
            to_missing = not self.has_filename(to_rel)
 
697
            if after:
 
698
                move_file = False
 
699
            else:
 
700
                move_file = True
 
701
            if to_missing:
 
702
                if not move_file:
 
703
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
704
                        errors.NoSuchFile(path=to_rel,
 
705
                        extra="New file has not been created yet"))
 
706
                elif from_missing:
 
707
                    # neither path exists
 
708
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
709
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
710
            else:
 
711
                if from_missing: # implicitly just update our path mapping
 
712
                    move_file = False
 
713
                elif not after:
 
714
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
715
                        extra="(Use --after to update the Bazaar id)")
 
716
 
 
717
            rollbacks = []
 
718
            def rollback_rename():
 
719
                """A single rename has failed, roll it back."""
 
720
                # roll back everything, even if we encounter trouble doing one
 
721
                # of them.
 
722
                #
 
723
                # TODO: at least log the other exceptions rather than just
 
724
                # losing them mbp 20070307
 
725
                exc_info = None
 
726
                for rollback in reversed(rollbacks):
 
727
                    try:
 
728
                        rollback()
 
729
                    except Exception, e:
 
730
                        exc_info = sys.exc_info()
 
731
                if exc_info:
 
732
                    raise exc_info[0], exc_info[1], exc_info[2]
 
733
 
 
734
            # perform the disk move first - its the most likely failure point.
 
735
            if move_file:
 
736
                from_rel_abs = self.abspath(from_rel)
 
737
                to_rel_abs = self.abspath(to_rel)
 
738
                try:
 
739
                    osutils.rename(from_rel_abs, to_rel_abs)
 
740
                except OSError, e:
 
741
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
742
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
743
            try:
 
744
                # perform the rename in the inventory next if needed: its easy
 
745
                # to rollback
 
746
                if update_inventory:
 
747
                    # rename the entry
 
748
                    from_entry = inv[from_id]
 
749
                    current_parent = from_entry.parent_id
 
750
                    inv.rename(from_id, to_dir_id, from_tail)
 
751
                    rollbacks.append(
 
752
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
753
                # finally do the rename in the dirstate, which is a little
 
754
                # tricky to rollback, but least likely to need it.
 
755
                old_block_index, old_entry_index, dir_present, file_present = \
 
756
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
757
                old_block = state._dirblocks[old_block_index][1]
 
758
                old_entry = old_block[old_entry_index]
 
759
                from_key, old_entry_details = old_entry
 
760
                cur_details = old_entry_details[0]
 
761
                # remove the old row
 
762
                to_key = ((to_block[0],) + from_key[1:3])
 
763
                minikind = cur_details[0]
 
764
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
765
                         minikind=minikind,
 
766
                         executable=cur_details[3],
 
767
                         fingerprint=cur_details[1],
 
768
                         packed_stat=cur_details[4],
 
769
                         size=cur_details[2],
 
770
                         to_block=to_block,
 
771
                         to_key=to_key,
 
772
                         to_path_utf8=to_rel_utf8)
 
773
 
 
774
                if minikind == 'd':
 
775
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
776
                        """Recursively update all entries in this dirblock."""
 
777
                        assert from_dir != '', "renaming root not supported"
 
778
                        from_key = (from_dir, '')
 
779
                        from_block_idx, present = \
 
780
                            state._find_block_index_from_key(from_key)
 
781
                        if not present:
 
782
                            # This is the old record, if it isn't present, then
 
783
                            # there is theoretically nothing to update.
 
784
                            # (Unless it isn't present because of lazy loading,
 
785
                            # but we don't do that yet)
 
786
                            return
 
787
                        from_block = state._dirblocks[from_block_idx]
 
788
                        to_block_index, to_entry_index, _, _ = \
 
789
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
790
                        to_block_index = state._ensure_block(
 
791
                            to_block_index, to_entry_index, to_dir_utf8)
 
792
                        to_block = state._dirblocks[to_block_index]
 
793
 
 
794
                        # Grab a copy since move_one may update the list.
 
795
                        for entry in from_block[1][:]:
 
796
                            assert entry[0][0] == from_dir
 
797
                            cur_details = entry[1][0]
 
798
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
799
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
800
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
801
                            minikind = cur_details[0]
 
802
                            if minikind in 'ar':
 
803
                                # Deleted children of a renamed directory
 
804
                                # Do not need to be updated.
 
805
                                # Children that have been renamed out of this
 
806
                                # directory should also not be updated
 
807
                                continue
 
808
                            move_one(entry, from_path_utf8=from_path_utf8,
 
809
                                     minikind=minikind,
 
810
                                     executable=cur_details[3],
 
811
                                     fingerprint=cur_details[1],
 
812
                                     packed_stat=cur_details[4],
 
813
                                     size=cur_details[2],
 
814
                                     to_block=to_block,
 
815
                                     to_key=to_key,
 
816
                                     to_path_utf8=to_path_utf8)
 
817
                            if minikind == 'd':
 
818
                                # We need to move all the children of this
 
819
                                # entry
 
820
                                update_dirblock(from_path_utf8, to_key,
 
821
                                                to_path_utf8)
 
822
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
823
            except:
 
824
                rollback_rename()
 
825
                raise
 
826
            result.append((from_rel, to_rel))
 
827
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
828
            self._make_dirty(reset_inventory=False)
 
829
 
 
830
        return result
 
831
 
 
832
    def _must_be_locked(self):
 
833
        if not self._control_files._lock_count:
 
834
            raise errors.ObjectNotLocked(self)
 
835
 
 
836
    def _new_tree(self):
 
837
        """Initialize the state in this tree to be a new tree."""
 
838
        self._dirty = True
 
839
 
 
840
    @needs_read_lock
 
841
    def path2id(self, path):
 
842
        """Return the id for path in this tree."""
 
843
        path = path.strip('/')
 
844
        entry = self._get_entry(path=path)
 
845
        if entry == (None, None):
 
846
            return None
 
847
        return entry[0][2]
 
848
 
 
849
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
850
        """See Tree.paths2ids().
 
851
 
 
852
        This specialisation fast-paths the case where all the trees are in the
 
853
        dirstate.
 
854
        """
 
855
        if paths is None:
 
856
            return None
 
857
        parents = self.get_parent_ids()
 
858
        for tree in trees:
 
859
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
860
                parents):
 
861
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
862
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
863
        # -- make all paths utf8 --
 
864
        paths_utf8 = set()
 
865
        for path in paths:
 
866
            paths_utf8.add(path.encode('utf8'))
 
867
        paths = paths_utf8
 
868
        # -- paths is now a utf8 path set --
 
869
        # -- get the state object and prepare it.
 
870
        state = self.current_dirstate()
 
871
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
872
            and '' not in paths):
 
873
            paths2ids = self._paths2ids_using_bisect
 
874
        else:
 
875
            paths2ids = self._paths2ids_in_memory
 
876
        return paths2ids(paths, search_indexes,
 
877
                         require_versioned=require_versioned)
 
878
 
 
879
    def _paths2ids_in_memory(self, paths, search_indexes,
 
880
                             require_versioned=True):
 
881
        state = self.current_dirstate()
 
882
        state._read_dirblocks_if_needed()
 
883
        def _entries_for_path(path):
 
884
            """Return a list with all the entries that match path for all ids.
 
885
            """
 
886
            dirname, basename = os.path.split(path)
 
887
            key = (dirname, basename, '')
 
888
            block_index, present = state._find_block_index_from_key(key)
 
889
            if not present:
 
890
                # the block which should contain path is absent.
 
891
                return []
 
892
            result = []
 
893
            block = state._dirblocks[block_index][1]
 
894
            entry_index, _ = state._find_entry_index(key, block)
 
895
            # we may need to look at multiple entries at this path: walk while the paths match.
 
896
            while (entry_index < len(block) and
 
897
                block[entry_index][0][0:2] == key[0:2]):
 
898
                result.append(block[entry_index])
 
899
                entry_index += 1
 
900
            return result
 
901
        if require_versioned:
 
902
            # -- check all supplied paths are versioned in a search tree. --
 
903
            all_versioned = True
 
904
            for path in paths:
 
905
                path_entries = _entries_for_path(path)
 
906
                if not path_entries:
 
907
                    # this specified path is not present at all: error
 
908
                    all_versioned = False
 
909
                    break
 
910
                found_versioned = False
 
911
                # for each id at this path
 
912
                for entry in path_entries:
 
913
                    # for each tree.
 
914
                    for index in search_indexes:
 
915
                        if entry[1][index][0] != 'a': # absent
 
916
                            found_versioned = True
 
917
                            # all good: found a versioned cell
 
918
                            break
 
919
                if not found_versioned:
 
920
                    # none of the indexes was not 'absent' at all ids for this
 
921
                    # path.
 
922
                    all_versioned = False
 
923
                    break
 
924
            if not all_versioned:
 
925
                raise errors.PathsNotVersionedError(paths)
 
926
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
927
        search_paths = osutils.minimum_path_selection(paths)
 
928
        # sketch: 
 
929
        # for all search_indexs in each path at or under each element of
 
930
        # search_paths, if the detail is relocated: add the id, and add the
 
931
        # relocated path as one to search if its not searched already. If the
 
932
        # detail is not relocated, add the id.
 
933
        searched_paths = set()
 
934
        found_ids = set()
 
935
        def _process_entry(entry):
 
936
            """Look at search_indexes within entry.
 
937
 
 
938
            If a specific tree's details are relocated, add the relocation
 
939
            target to search_paths if not searched already. If it is absent, do
 
940
            nothing. Otherwise add the id to found_ids.
 
941
            """
 
942
            for index in search_indexes:
 
943
                if entry[1][index][0] == 'r': # relocated
 
944
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
945
                        search_paths.add(entry[1][index][1])
 
946
                elif entry[1][index][0] != 'a': # absent
 
947
                    found_ids.add(entry[0][2])
 
948
        while search_paths:
 
949
            current_root = search_paths.pop()
 
950
            searched_paths.add(current_root)
 
951
            # process the entries for this containing directory: the rest will be
 
952
            # found by their parents recursively.
 
953
            root_entries = _entries_for_path(current_root)
 
954
            if not root_entries:
 
955
                # this specified path is not present at all, skip it.
 
956
                continue
 
957
            for entry in root_entries:
 
958
                _process_entry(entry)
 
959
            initial_key = (current_root, '', '')
 
960
            block_index, _ = state._find_block_index_from_key(initial_key)
 
961
            while (block_index < len(state._dirblocks) and
 
962
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
963
                for entry in state._dirblocks[block_index][1]:
 
964
                    _process_entry(entry)
 
965
                block_index += 1
 
966
        return found_ids
 
967
 
 
968
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
969
                                require_versioned=True):
 
970
        state = self.current_dirstate()
 
971
        found_ids = set()
 
972
 
 
973
        split_paths = sorted(osutils.split(p) for p in paths)
 
974
        found = state._bisect_recursive(split_paths)
 
975
 
 
976
        if require_versioned:
 
977
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
978
            for dir_name in split_paths:
 
979
                if dir_name not in found_dir_names:
 
980
                    raise errors.PathsNotVersionedError(paths)
 
981
 
 
982
        for dir_name_id, trees_info in found.iteritems():
 
983
            for index in search_indexes:
 
984
                if trees_info[index][0] not in ('r', 'a'):
 
985
                    found_ids.add(dir_name_id[2])
 
986
        return found_ids
 
987
 
 
988
    def read_working_inventory(self):
 
989
        """Read the working inventory.
 
990
        
 
991
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
992
        """
 
993
        return self.inventory
 
994
 
 
995
    @needs_read_lock
 
996
    def revision_tree(self, revision_id):
 
997
        """See Tree.revision_tree.
 
998
 
 
999
        WorkingTree4 supplies revision_trees for any basis tree.
 
1000
        """
 
1001
        dirstate = self.current_dirstate()
 
1002
        parent_ids = dirstate.get_parent_ids()
 
1003
        if revision_id not in parent_ids:
 
1004
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1005
        if revision_id in dirstate.get_ghosts():
 
1006
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1007
        return DirStateRevisionTree(dirstate, revision_id,
 
1008
            self.branch.repository)
 
1009
 
 
1010
    @needs_tree_write_lock
 
1011
    def set_last_revision(self, new_revision):
 
1012
        """Change the last revision in the working tree."""
 
1013
        parents = self.get_parent_ids()
 
1014
        if new_revision in (NULL_REVISION, None):
 
1015
            assert len(parents) < 2, (
 
1016
                "setting the last parent to none with a pending merge is "
 
1017
                "unsupported.")
 
1018
            self.set_parent_ids([])
 
1019
        else:
 
1020
            self.set_parent_ids([new_revision] + parents[1:],
 
1021
                allow_leftmost_as_ghost=True)
 
1022
 
 
1023
    @needs_tree_write_lock
 
1024
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1025
        """Set the parent ids to revision_ids.
 
1026
        
 
1027
        See also set_parent_trees. This api will try to retrieve the tree data
 
1028
        for each element of revision_ids from the trees repository. If you have
 
1029
        tree data already available, it is more efficient to use
 
1030
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1031
        an easier API to use.
 
1032
 
 
1033
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1034
            working tree. Any of these may be ghosts.
 
1035
        """
 
1036
        trees = []
 
1037
        for revision_id in revision_ids:
 
1038
            try:
 
1039
                revtree = self.branch.repository.revision_tree(revision_id)
 
1040
                # TODO: jam 20070213 KnitVersionedFile raises
 
1041
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1042
                #       given revision_id is not present. Should Repository be
 
1043
                #       catching it and re-raising NoSuchRevision?
 
1044
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1045
                revtree = None
 
1046
            trees.append((revision_id, revtree))
 
1047
        self.set_parent_trees(trees,
 
1048
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1049
 
 
1050
    @needs_tree_write_lock
 
1051
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1052
        """Set the parents of the working tree.
 
1053
 
 
1054
        :param parents_list: A list of (revision_id, tree) tuples.
 
1055
            If tree is None, then that element is treated as an unreachable
 
1056
            parent tree - i.e. a ghost.
 
1057
        """
 
1058
        dirstate = self.current_dirstate()
 
1059
        if len(parents_list) > 0:
 
1060
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1061
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1062
        real_trees = []
 
1063
        ghosts = []
 
1064
        # convert absent trees to the null tree, which we convert back to
 
1065
        # missing on access.
 
1066
        for rev_id, tree in parents_list:
 
1067
            _mod_revision.check_not_reserved_id(rev_id)
 
1068
            if tree is not None:
 
1069
                real_trees.append((rev_id, tree))
 
1070
            else:
 
1071
                real_trees.append((rev_id,
 
1072
                    self.branch.repository.revision_tree(None)))
 
1073
                ghosts.append(rev_id)
 
1074
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1075
        self._make_dirty(reset_inventory=False)
 
1076
 
 
1077
    def _set_root_id(self, file_id):
 
1078
        """See WorkingTree.set_root_id."""
 
1079
        state = self.current_dirstate()
 
1080
        state.set_path_id('', file_id)
 
1081
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1082
            self._make_dirty(reset_inventory=True)
 
1083
 
 
1084
    def _sha_from_stat(self, path, stat_result):
 
1085
        """Get a sha digest from the tree's stat cache.
 
1086
 
 
1087
        The default implementation assumes no stat cache is present.
 
1088
 
 
1089
        :param path: The path.
 
1090
        :param stat_result: The stat result being looked up.
 
1091
        """
 
1092
        state = self.current_dirstate()
 
1093
        # XXX: should we make the path be passed in as utf8 ?
 
1094
        entry = state._get_entry(0, path_utf8=cache_utf8.encode(path))
 
1095
        tree_details = entry[1][0]
 
1096
        packed_stat = dirstate.pack_stat(stat_result)
 
1097
        if tree_details[4] == packed_stat:
 
1098
            return tree_details[1]
 
1099
        else:
 
1100
            return None
 
1101
 
 
1102
    @needs_read_lock
 
1103
    def supports_tree_reference(self):
 
1104
        return self._repo_supports_tree_reference
 
1105
 
 
1106
    def unlock(self):
 
1107
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1108
        # do non-implementation specific cleanup
 
1109
        self._cleanup()
 
1110
 
 
1111
        if self._control_files._lock_count == 1:
 
1112
            # eventually we should do signature checking during read locks for
 
1113
            # dirstate updates.
 
1114
            if self._control_files._lock_mode == 'w':
 
1115
                if self._dirty:
 
1116
                    self.flush()
 
1117
            if self._dirstate is not None:
 
1118
                # This is a no-op if there are no modifications.
 
1119
                self._dirstate.save()
 
1120
                self._dirstate.unlock()
 
1121
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1122
            #       point. Instead, it could check if the header has been
 
1123
            #       modified when it is locked, and if not, it can hang on to
 
1124
            #       the data it has in memory.
 
1125
            self._dirstate = None
 
1126
            self._inventory = None
 
1127
        # reverse order of locking.
 
1128
        try:
 
1129
            return self._control_files.unlock()
 
1130
        finally:
 
1131
            self.branch.unlock()
 
1132
 
 
1133
    @needs_tree_write_lock
 
1134
    def unversion(self, file_ids):
 
1135
        """Remove the file ids in file_ids from the current versioned set.
 
1136
 
 
1137
        When a file_id is unversioned, all of its children are automatically
 
1138
        unversioned.
 
1139
 
 
1140
        :param file_ids: The file ids to stop versioning.
 
1141
        :raises: NoSuchId if any fileid is not currently versioned.
 
1142
        """
 
1143
        if not file_ids:
 
1144
            return
 
1145
        state = self.current_dirstate()
 
1146
        state._read_dirblocks_if_needed()
 
1147
        ids_to_unversion = set(file_ids)
 
1148
        paths_to_unversion = set()
 
1149
        # sketch:
 
1150
        # check if the root is to be unversioned, if so, assert for now.
 
1151
        # walk the state marking unversioned things as absent.
 
1152
        # if there are any un-unversioned ids at the end, raise
 
1153
        for key, details in state._dirblocks[0][1]:
 
1154
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1155
                key[2] in ids_to_unversion):
 
1156
                # I haven't written the code to unversion / yet - it should be
 
1157
                # supported.
 
1158
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1159
        block_index = 0
 
1160
        while block_index < len(state._dirblocks):
 
1161
            # process one directory at a time.
 
1162
            block = state._dirblocks[block_index]
 
1163
            # first check: is the path one to remove - it or its children
 
1164
            delete_block = False
 
1165
            for path in paths_to_unversion:
 
1166
                if (block[0].startswith(path) and
 
1167
                    (len(block[0]) == len(path) or
 
1168
                     block[0][len(path)] == '/')):
 
1169
                    # this entire block should be deleted - its the block for a
 
1170
                    # path to unversion; or the child of one
 
1171
                    delete_block = True
 
1172
                    break
 
1173
            # TODO: trim paths_to_unversion as we pass by paths
 
1174
            if delete_block:
 
1175
                # this block is to be deleted: process it.
 
1176
                # TODO: we can special case the no-parents case and
 
1177
                # just forget the whole block.
 
1178
                entry_index = 0
 
1179
                while entry_index < len(block[1]):
 
1180
                    # Mark this file id as having been removed
 
1181
                    entry = block[1][entry_index]
 
1182
                    ids_to_unversion.discard(entry[0][2])
 
1183
                    if (entry[1][0][0] == 'a'
 
1184
                        or not state._make_absent(entry)):
 
1185
                        entry_index += 1
 
1186
                # go to the next block. (At the moment we dont delete empty
 
1187
                # dirblocks)
 
1188
                block_index += 1
 
1189
                continue
 
1190
            entry_index = 0
 
1191
            while entry_index < len(block[1]):
 
1192
                entry = block[1][entry_index]
 
1193
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1194
                    # ^ some parent row.
 
1195
                    entry[0][2] not in ids_to_unversion):
 
1196
                    # ^ not an id to unversion
 
1197
                    entry_index += 1
 
1198
                    continue
 
1199
                if entry[1][0][0] == 'd':
 
1200
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1201
                if not state._make_absent(entry):
 
1202
                    entry_index += 1
 
1203
                # we have unversioned this id
 
1204
                ids_to_unversion.remove(entry[0][2])
 
1205
            block_index += 1
 
1206
        if ids_to_unversion:
 
1207
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1208
        self._make_dirty(reset_inventory=False)
 
1209
        # have to change the legacy inventory too.
 
1210
        if self._inventory is not None:
 
1211
            for file_id in file_ids:
 
1212
                self._inventory.remove_recursive_id(file_id)
 
1213
 
 
1214
    @needs_read_lock
 
1215
    def _validate(self):
 
1216
        self._dirstate._validate()
 
1217
 
 
1218
    @needs_tree_write_lock
 
1219
    def _write_inventory(self, inv):
 
1220
        """Write inventory as the current inventory."""
 
1221
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1222
        self.current_dirstate().set_state_from_inventory(inv)
 
1223
        self._make_dirty(reset_inventory=False)
 
1224
        if self._inventory is not None:
 
1225
            self._inventory = inv
 
1226
        self.flush()
 
1227
 
 
1228
 
 
1229
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1230
    """The first consolidated dirstate working tree format.
 
1231
 
 
1232
    This format:
 
1233
        - exists within a metadir controlling .bzr
 
1234
        - includes an explicit version marker for the workingtree control
 
1235
          files, separate from the BzrDir format
 
1236
        - modifies the hash cache format
 
1237
        - is new in bzr 0.15
 
1238
        - uses a LockDir to guard access to it.
 
1239
    """
 
1240
 
 
1241
    upgrade_recommended = False
 
1242
 
 
1243
    def get_format_string(self):
 
1244
        """See WorkingTreeFormat.get_format_string()."""
 
1245
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1246
 
 
1247
    def get_format_description(self):
 
1248
        """See WorkingTreeFormat.get_format_description()."""
 
1249
        return "Working tree format 4"
 
1250
 
 
1251
    def initialize(self, a_bzrdir, revision_id=None):
 
1252
        """See WorkingTreeFormat.initialize().
 
1253
 
 
1254
        :param revision_id: allows creating a working tree at a different
 
1255
        revision than the branch is at.
 
1256
 
 
1257
        These trees get an initial random root id, if their repository supports
 
1258
        rich root data, TREE_ROOT otherwise.
 
1259
        """
 
1260
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1261
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1262
        transport = a_bzrdir.get_workingtree_transport(self)
 
1263
        control_files = self._open_control_files(a_bzrdir)
 
1264
        control_files.create_lock()
 
1265
        control_files.lock_write()
 
1266
        control_files.put_utf8('format', self.get_format_string())
 
1267
        branch = a_bzrdir.open_branch()
 
1268
        if revision_id is None:
 
1269
            revision_id = branch.last_revision()
 
1270
        local_path = transport.local_abspath('dirstate')
 
1271
        # write out new dirstate (must exist when we create the tree)
 
1272
        state = dirstate.DirState.initialize(local_path)
 
1273
        state.unlock()
 
1274
        del state
 
1275
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1276
                         branch,
 
1277
                         _format=self,
 
1278
                         _bzrdir=a_bzrdir,
 
1279
                         _control_files=control_files)
 
1280
        wt._new_tree()
 
1281
        wt.lock_tree_write()
 
1282
        try:
 
1283
            if revision_id in (None, NULL_REVISION):
 
1284
                if branch.repository.supports_rich_root():
 
1285
                    wt._set_root_id(generate_ids.gen_root_id())
 
1286
                else:
 
1287
                    wt._set_root_id(ROOT_ID)
 
1288
                wt.flush()
 
1289
            wt.set_last_revision(revision_id)
 
1290
            wt.flush()
 
1291
            basis = wt.basis_tree()
 
1292
            basis.lock_read()
 
1293
            # if the basis has a root id we have to use that; otherwise we use
 
1294
            # a new random one
 
1295
            basis_root_id = basis.get_root_id()
 
1296
            if basis_root_id is not None:
 
1297
                wt._set_root_id(basis_root_id)
 
1298
                wt.flush()
 
1299
            transform.build_tree(basis, wt)
 
1300
            basis.unlock()
 
1301
        finally:
 
1302
            control_files.unlock()
 
1303
            wt.unlock()
 
1304
        return wt
 
1305
 
 
1306
    def _open(self, a_bzrdir, control_files):
 
1307
        """Open the tree itself.
 
1308
 
 
1309
        :param a_bzrdir: the dir for the tree.
 
1310
        :param control_files: the control files for the tree.
 
1311
        """
 
1312
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1313
                           branch=a_bzrdir.open_branch(),
 
1314
                           _format=self,
 
1315
                           _bzrdir=a_bzrdir,
 
1316
                           _control_files=control_files)
 
1317
 
 
1318
    def __get_matchingbzrdir(self):
 
1319
        # please test against something that will let us do tree references
 
1320
        return bzrdir.format_registry.make_bzrdir(
 
1321
            'dirstate-with-subtree')
 
1322
 
 
1323
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1324
 
 
1325
 
 
1326
class DirStateRevisionTree(Tree):
 
1327
    """A revision tree pulling the inventory from a dirstate."""
 
1328
 
 
1329
    def __init__(self, dirstate, revision_id, repository):
 
1330
        self._dirstate = dirstate
 
1331
        self._revision_id = revision_id
 
1332
        self._repository = repository
 
1333
        self._inventory = None
 
1334
        self._locked = 0
 
1335
        self._dirstate_locked = False
 
1336
 
 
1337
    def __repr__(self):
 
1338
        return "<%s of %s in %s>" % \
 
1339
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1340
 
 
1341
    def annotate_iter(self, file_id,
 
1342
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1343
        """See Tree.annotate_iter"""
 
1344
        w = self._get_weave(file_id)
 
1345
        return w.annotate_iter(self.inventory[file_id].revision)
 
1346
 
 
1347
    def _get_ancestors(self, default_revision):
 
1348
        return set(self._repository.get_ancestry(self._revision_id,
 
1349
                                                 topo_sorted=False))
 
1350
    def _comparison_data(self, entry, path):
 
1351
        """See Tree._comparison_data."""
 
1352
        if entry is None:
 
1353
            return None, False, None
 
1354
        # trust the entry as RevisionTree does, but this may not be
 
1355
        # sensible: the entry might not have come from us?
 
1356
        return entry.kind, entry.executable, None
 
1357
 
 
1358
    def _file_size(self, entry, stat_value):
 
1359
        return entry.text_size
 
1360
 
 
1361
    def filter_unversioned_files(self, paths):
 
1362
        """Filter out paths that are not versioned.
 
1363
 
 
1364
        :return: set of paths.
 
1365
        """
 
1366
        pred = self.has_filename
 
1367
        return set((p for p in paths if not pred(p)))
 
1368
 
 
1369
    def get_root_id(self):
 
1370
        return self.path2id('')
 
1371
 
 
1372
    def _get_parent_index(self):
 
1373
        """Return the index in the dirstate referenced by this tree."""
 
1374
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1375
 
 
1376
    def _get_entry(self, file_id=None, path=None):
 
1377
        """Get the dirstate row for file_id or path.
 
1378
 
 
1379
        If either file_id or path is supplied, it is used as the key to lookup.
 
1380
        If both are supplied, the fastest lookup is used, and an error is
 
1381
        raised if they do not both point at the same row.
 
1382
        
 
1383
        :param file_id: An optional unicode file_id to be looked up.
 
1384
        :param path: An optional unicode path to be looked up.
 
1385
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1386
        """
 
1387
        if file_id is None and path is None:
 
1388
            raise errors.BzrError('must supply file_id or path')
 
1389
        if path is not None:
 
1390
            path = path.encode('utf8')
 
1391
        parent_index = self._get_parent_index()
 
1392
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1393
 
 
1394
    def _generate_inventory(self):
 
1395
        """Create and set self.inventory from the dirstate object.
 
1396
 
 
1397
        (So this is only called the first time the inventory is requested for
 
1398
        this tree; it then remains in memory until it's out of date.)
 
1399
 
 
1400
        This is relatively expensive: we have to walk the entire dirstate.
 
1401
        """
 
1402
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1403
            'dirstate revision tree'
 
1404
        # separate call for profiling - makes it clear where the costs are.
 
1405
        self._dirstate._read_dirblocks_if_needed()
 
1406
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1407
            'parent %s has disappeared from %s' % (
 
1408
            self._revision_id, self._dirstate.get_parent_ids())
 
1409
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1410
        # This is identical now to the WorkingTree _generate_inventory except
 
1411
        # for the tree index use.
 
1412
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1413
        current_id = root_key[2]
 
1414
        assert current_entry[parent_index][0] == 'd'
 
1415
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1416
        inv.root.revision = current_entry[parent_index][4]
 
1417
        # Turn some things into local variables
 
1418
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1419
        factory = entry_factory
 
1420
        utf8_decode = cache_utf8._utf8_decode
 
1421
        inv_byid = inv._byid
 
1422
        # we could do this straight out of the dirstate; it might be fast
 
1423
        # and should be profiled - RBC 20070216
 
1424
        parent_ies = {'' : inv.root}
 
1425
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1426
            dirname = block[0]
 
1427
            try:
 
1428
                parent_ie = parent_ies[dirname]
 
1429
            except KeyError:
 
1430
                # all the paths in this block are not versioned in this tree
 
1431
                continue
 
1432
            for key, entry in block[1]:
 
1433
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1434
                if minikind in ('a', 'r'): # absent, relocated
 
1435
                    # not this tree
 
1436
                    continue
 
1437
                name = key[1]
 
1438
                name_unicode = utf8_decode(name)[0]
 
1439
                file_id = key[2]
 
1440
                kind = minikind_to_kind[minikind]
 
1441
                inv_entry = factory[kind](file_id, name_unicode,
 
1442
                                          parent_ie.file_id)
 
1443
                inv_entry.revision = revid
 
1444
                if kind == 'file':
 
1445
                    inv_entry.executable = executable
 
1446
                    inv_entry.text_size = size
 
1447
                    inv_entry.text_sha1 = fingerprint
 
1448
                elif kind == 'directory':
 
1449
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1450
                elif kind == 'symlink':
 
1451
                    inv_entry.executable = False
 
1452
                    inv_entry.text_size = None
 
1453
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1454
                elif kind == 'tree-reference':
 
1455
                    inv_entry.reference_revision = fingerprint or None
 
1456
                else:
 
1457
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1458
                            % entry)
 
1459
                # These checks cost us around 40ms on a 55k entry tree
 
1460
                assert file_id not in inv_byid
 
1461
                assert name_unicode not in parent_ie.children
 
1462
                inv_byid[file_id] = inv_entry
 
1463
                parent_ie.children[name_unicode] = inv_entry
 
1464
        self._inventory = inv
 
1465
 
 
1466
    def get_file_mtime(self, file_id, path=None):
 
1467
        """Return the modification time for this record.
 
1468
 
 
1469
        We return the timestamp of the last-changed revision.
 
1470
        """
 
1471
        # Make sure the file exists
 
1472
        entry = self._get_entry(file_id, path=path)
 
1473
        if entry == (None, None): # do we raise?
 
1474
            return None
 
1475
        parent_index = self._get_parent_index()
 
1476
        last_changed_revision = entry[1][parent_index][4]
 
1477
        return self._repository.get_revision(last_changed_revision).timestamp
 
1478
 
 
1479
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1480
        entry = self._get_entry(file_id=file_id, path=path)
 
1481
        parent_index = self._get_parent_index()
 
1482
        parent_details = entry[1][parent_index]
 
1483
        if parent_details[0] == 'f':
 
1484
            return parent_details[1]
 
1485
        return None
 
1486
 
 
1487
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1488
    def get_weave(self, file_id):
 
1489
        return self._get_weave(file_id)
 
1490
 
 
1491
    def _get_weave(self, file_id):
 
1492
        return self._repository.weave_store.get_weave(file_id,
 
1493
                self._repository.get_transaction())
 
1494
 
 
1495
    def get_file(self, file_id, path=None):
 
1496
        return StringIO(self.get_file_text(file_id))
 
1497
 
 
1498
    def get_file_lines(self, file_id):
 
1499
        ie = self.inventory[file_id]
 
1500
        return self._get_weave(file_id).get_lines(ie.revision)
 
1501
 
 
1502
    def get_file_size(self, file_id):
 
1503
        return self.inventory[file_id].text_size
 
1504
 
 
1505
    def get_file_text(self, file_id):
 
1506
        return ''.join(self.get_file_lines(file_id))
 
1507
 
 
1508
    def get_reference_revision(self, file_id, path=None):
 
1509
        return self.inventory[file_id].reference_revision
 
1510
 
 
1511
    def iter_files_bytes(self, desired_files):
 
1512
        """See Tree.iter_files_bytes.
 
1513
 
 
1514
        This version is implemented on top of Repository.iter_files_bytes"""
 
1515
        parent_index = self._get_parent_index()
 
1516
        repo_desired_files = []
 
1517
        for file_id, identifier in desired_files:
 
1518
            entry = self._get_entry(file_id)
 
1519
            if entry == (None, None):
 
1520
                raise errors.NoSuchId(self, file_id)
 
1521
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1522
                                       identifier))
 
1523
        return self._repository.iter_files_bytes(repo_desired_files)
 
1524
 
 
1525
    def get_symlink_target(self, file_id):
 
1526
        entry = self._get_entry(file_id=file_id)
 
1527
        parent_index = self._get_parent_index()
 
1528
        if entry[1][parent_index][0] != 'l':
 
1529
            return None
 
1530
        else:
 
1531
            # At present, none of the tree implementations supports non-ascii
 
1532
            # symlink targets. So we will just assume that the dirstate path is
 
1533
            # correct.
 
1534
            return entry[1][parent_index][1]
 
1535
 
 
1536
    def get_revision_id(self):
 
1537
        """Return the revision id for this tree."""
 
1538
        return self._revision_id
 
1539
 
 
1540
    def _get_inventory(self):
 
1541
        if self._inventory is not None:
 
1542
            return self._inventory
 
1543
        self._must_be_locked()
 
1544
        self._generate_inventory()
 
1545
        return self._inventory
 
1546
 
 
1547
    inventory = property(_get_inventory,
 
1548
                         doc="Inventory of this Tree")
 
1549
 
 
1550
    def get_parent_ids(self):
 
1551
        """The parents of a tree in the dirstate are not cached."""
 
1552
        return self._repository.get_revision(self._revision_id).parent_ids
 
1553
 
 
1554
    def has_filename(self, filename):
 
1555
        return bool(self.path2id(filename))
 
1556
 
 
1557
    def kind(self, file_id):
 
1558
        return self.inventory[file_id].kind
 
1559
 
 
1560
    def path_content_summary(self, path):
 
1561
        """See Tree.path_content_summary."""
 
1562
        id = self.inventory.path2id(path)
 
1563
        if id is None:
 
1564
            return ('missing', None, None, None)
 
1565
        entry = self._inventory[id]
 
1566
        kind = entry.kind
 
1567
        if kind == 'file':
 
1568
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1569
        elif kind == 'symlink':
 
1570
            return (kind, None, None, entry.symlink_target)
 
1571
        else:
 
1572
            return (kind, None, None, None)
 
1573
 
 
1574
    def is_executable(self, file_id, path=None):
 
1575
        ie = self.inventory[file_id]
 
1576
        if ie.kind != "file":
 
1577
            return None
 
1578
        return ie.executable
 
1579
 
 
1580
    def list_files(self, include_root=False):
 
1581
        # We use a standard implementation, because DirStateRevisionTree is
 
1582
        # dealing with one of the parents of the current state
 
1583
        inv = self._get_inventory()
 
1584
        entries = inv.iter_entries()
 
1585
        if self.inventory.root is not None and not include_root:
 
1586
            entries.next()
 
1587
        for path, entry in entries:
 
1588
            yield path, 'V', entry.kind, entry.file_id, entry
 
1589
 
 
1590
    def lock_read(self):
 
1591
        """Lock the tree for a set of operations."""
 
1592
        if not self._locked:
 
1593
            self._repository.lock_read()
 
1594
            if self._dirstate._lock_token is None:
 
1595
                self._dirstate.lock_read()
 
1596
                self._dirstate_locked = True
 
1597
        self._locked += 1
 
1598
 
 
1599
    def _must_be_locked(self):
 
1600
        if not self._locked:
 
1601
            raise errors.ObjectNotLocked(self)
 
1602
 
 
1603
    @needs_read_lock
 
1604
    def path2id(self, path):
 
1605
        """Return the id for path in this tree."""
 
1606
        # lookup by path: faster than splitting and walking the ivnentory.
 
1607
        entry = self._get_entry(path=path)
 
1608
        if entry == (None, None):
 
1609
            return None
 
1610
        return entry[0][2]
 
1611
 
 
1612
    def unlock(self):
 
1613
        """Unlock, freeing any cache memory used during the lock."""
 
1614
        # outside of a lock, the inventory is suspect: release it.
 
1615
        self._locked -=1
 
1616
        if not self._locked:
 
1617
            self._inventory = None
 
1618
            self._locked = 0
 
1619
            if self._dirstate_locked:
 
1620
                self._dirstate.unlock()
 
1621
                self._dirstate_locked = False
 
1622
            self._repository.unlock()
 
1623
 
 
1624
    def walkdirs(self, prefix=""):
 
1625
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1626
        # implementation based on an inventory.  
 
1627
        # This should be cleaned up to use the much faster Dirstate code
 
1628
        # So for now, we just build up the parent inventory, and extract
 
1629
        # it the same way RevisionTree does.
 
1630
        _directory = 'directory'
 
1631
        inv = self._get_inventory()
 
1632
        top_id = inv.path2id(prefix)
 
1633
        if top_id is None:
 
1634
            pending = []
 
1635
        else:
 
1636
            pending = [(prefix, top_id)]
 
1637
        while pending:
 
1638
            dirblock = []
 
1639
            relpath, file_id = pending.pop()
 
1640
            # 0 - relpath, 1- file-id
 
1641
            if relpath:
 
1642
                relroot = relpath + '/'
 
1643
            else:
 
1644
                relroot = ""
 
1645
            # FIXME: stash the node in pending
 
1646
            entry = inv[file_id]
 
1647
            for name, child in entry.sorted_children():
 
1648
                toppath = relroot + name
 
1649
                dirblock.append((toppath, name, child.kind, None,
 
1650
                    child.file_id, child.kind
 
1651
                    ))
 
1652
            yield (relpath, entry.file_id), dirblock
 
1653
            # push the user specified dirs from dirblock
 
1654
            for dir in reversed(dirblock):
 
1655
                if dir[2] == _directory:
 
1656
                    pending.append((dir[0], dir[4]))
 
1657
 
 
1658
 
 
1659
class InterDirStateTree(InterTree):
 
1660
    """Fast path optimiser for changes_from with dirstate trees.
 
1661
    
 
1662
    This is used only when both trees are in the dirstate working file, and 
 
1663
    the source is any parent within the dirstate, and the destination is 
 
1664
    the current working tree of the same dirstate.
 
1665
    """
 
1666
    # this could be generalized to allow comparisons between any trees in the
 
1667
    # dirstate, and possibly between trees stored in different dirstates.
 
1668
 
 
1669
    def __init__(self, source, target):
 
1670
        super(InterDirStateTree, self).__init__(source, target)
 
1671
        if not InterDirStateTree.is_compatible(source, target):
 
1672
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1673
 
 
1674
    @staticmethod
 
1675
    def make_source_parent_tree(source, target):
 
1676
        """Change the source tree into a parent of the target."""
 
1677
        revid = source.commit('record tree')
 
1678
        target.branch.repository.fetch(source.branch.repository, revid)
 
1679
        target.set_parent_ids([revid])
 
1680
        return target.basis_tree(), target
 
1681
 
 
1682
    _matching_from_tree_format = WorkingTreeFormat4()
 
1683
    _matching_to_tree_format = WorkingTreeFormat4()
 
1684
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1685
 
 
1686
    def _iter_changes(self, include_unchanged=False,
 
1687
                      specific_files=None, pb=None, extra_trees=[],
 
1688
                      require_versioned=True, want_unversioned=False):
 
1689
        """Return the changes from source to target.
 
1690
 
 
1691
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1692
            for details.
 
1693
        :param specific_files: An optional list of file paths to restrict the
 
1694
            comparison to. When mapping filenames to ids, all matches in all
 
1695
            trees (including optional extra_trees) are used, and all children of
 
1696
            matched directories are included.
 
1697
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1698
            unchanged entries in the result.
 
1699
        :param extra_trees: An optional list of additional trees to use when
 
1700
            mapping the contents of specific_files (paths) to file_ids.
 
1701
        :param require_versioned: If True, all files in specific_files must be
 
1702
            versioned in one of source, target, extra_trees or
 
1703
            PathsNotVersionedError is raised.
 
1704
        :param want_unversioned: Should unversioned files be returned in the
 
1705
            output. An unversioned file is defined as one with (False, False)
 
1706
            for the versioned pair.
 
1707
        """
 
1708
        utf8_decode = cache_utf8._utf8_decode
 
1709
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1710
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1711
        # NB: show_status depends on being able to pass in non-versioned files
 
1712
        # and report them as unknown
 
1713
        # TODO: handle extra trees in the dirstate.
 
1714
        # TODO: handle comparisons as an empty tree as a different special
 
1715
        # case? mbp 20070226
 
1716
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1717
            or specific_files == []):
 
1718
            # we can't fast-path these cases (yet)
 
1719
            for f in super(InterDirStateTree, self)._iter_changes(
 
1720
                include_unchanged, specific_files, pb, extra_trees,
 
1721
                require_versioned, want_unversioned=want_unversioned):
 
1722
                yield f
 
1723
            return
 
1724
        parent_ids = self.target.get_parent_ids()
 
1725
        assert (self.source._revision_id in parent_ids), \
 
1726
                "revision {%s} is not stored in {%s}, but %s " \
 
1727
                "can only be used for trees stored in the dirstate" \
 
1728
                % (self.source._revision_id, self.target, self._iter_changes)
 
1729
        target_index = 0
 
1730
        if self.source._revision_id == NULL_REVISION:
 
1731
            source_index = None
 
1732
            indices = (target_index,)
 
1733
        else:
 
1734
            assert (self.source._revision_id in parent_ids), \
 
1735
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1736
                self.source._revision_id, parent_ids)
 
1737
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1738
            indices = (source_index,target_index)
 
1739
        # -- make all specific_files utf8 --
 
1740
        if specific_files:
 
1741
            specific_files_utf8 = set()
 
1742
            for path in specific_files:
 
1743
                specific_files_utf8.add(path.encode('utf8'))
 
1744
            specific_files = specific_files_utf8
 
1745
        else:
 
1746
            specific_files = set([''])
 
1747
        # -- specific_files is now a utf8 path set --
 
1748
        # -- get the state object and prepare it.
 
1749
        state = self.target.current_dirstate()
 
1750
        state._read_dirblocks_if_needed()
 
1751
        def _entries_for_path(path):
 
1752
            """Return a list with all the entries that match path for all ids.
 
1753
            """
 
1754
            dirname, basename = os.path.split(path)
 
1755
            key = (dirname, basename, '')
 
1756
            block_index, present = state._find_block_index_from_key(key)
 
1757
            if not present:
 
1758
                # the block which should contain path is absent.
 
1759
                return []
 
1760
            result = []
 
1761
            block = state._dirblocks[block_index][1]
 
1762
            entry_index, _ = state._find_entry_index(key, block)
 
1763
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1764
            while (entry_index < len(block) and
 
1765
                block[entry_index][0][0:2] == key[0:2]):
 
1766
                result.append(block[entry_index])
 
1767
                entry_index += 1
 
1768
            return result
 
1769
        if require_versioned:
 
1770
            # -- check all supplied paths are versioned in a search tree. --
 
1771
            all_versioned = True
 
1772
            for path in specific_files:
 
1773
                path_entries = _entries_for_path(path)
 
1774
                if not path_entries:
 
1775
                    # this specified path is not present at all: error
 
1776
                    all_versioned = False
 
1777
                    break
 
1778
                found_versioned = False
 
1779
                # for each id at this path
 
1780
                for entry in path_entries:
 
1781
                    # for each tree.
 
1782
                    for index in indices:
 
1783
                        if entry[1][index][0] != 'a': # absent
 
1784
                            found_versioned = True
 
1785
                            # all good: found a versioned cell
 
1786
                            break
 
1787
                if not found_versioned:
 
1788
                    # none of the indexes was not 'absent' at all ids for this
 
1789
                    # path.
 
1790
                    all_versioned = False
 
1791
                    break
 
1792
            if not all_versioned:
 
1793
                raise errors.PathsNotVersionedError(specific_files)
 
1794
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1795
        search_specific_files = set()
 
1796
        for path in specific_files:
 
1797
            other_specific_files = specific_files.difference(set([path]))
 
1798
            if not osutils.is_inside_any(other_specific_files, path):
 
1799
                # this is a top level path, we must check it.
 
1800
                search_specific_files.add(path)
 
1801
        # sketch: 
 
1802
        # compare source_index and target_index at or under each element of search_specific_files.
 
1803
        # follow the following comparison table. Note that we only want to do diff operations when
 
1804
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1805
        # for the target.
 
1806
        # cases:
 
1807
        # 
 
1808
        # Source | Target | disk | action
 
1809
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1810
        #        |        |      | diff check on source-target
 
1811
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1812
        #        |        |      | ???
 
1813
        #   r    |  a     |      | add source to search
 
1814
        #   r    |  a     |  a   | 
 
1815
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1816
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1817
        #   a    | fdlt   |      | add new id
 
1818
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1819
        #   a    |  a     |      | not present in either tree, skip
 
1820
        #   a    |  a     |  a   | not present in any tree, skip
 
1821
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1822
        #        |        |      | may not be selected by the users list of paths.
 
1823
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1824
        #        |        |      | may not be selected by the users list of paths.
 
1825
        #  fdlt  | fdlt   |      | content in both: diff them
 
1826
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1827
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1828
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1829
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1830
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1831
        #        |        |      | this id at the other path.
 
1832
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1833
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1834
        #        |        |      | this id at the other path.
 
1835
 
 
1836
        # for all search_indexs in each path at or under each element of
 
1837
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1838
        # relocated path as one to search if its not searched already. If the
 
1839
        # detail is not relocated, add the id.
 
1840
        searched_specific_files = set()
 
1841
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1842
        # Using a list so that we can access the values and change them in
 
1843
        # nested scope. Each one is [path, file_id, entry]
 
1844
        last_source_parent = [None, None]
 
1845
        last_target_parent = [None, None]
 
1846
 
 
1847
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1848
 
 
1849
        # Just a sentry, so that _process_entry can say that this
 
1850
        # record is handled, but isn't interesting to process (unchanged)
 
1851
        uninteresting = object()
 
1852
 
 
1853
 
 
1854
        old_dirname_to_file_id = {}
 
1855
        new_dirname_to_file_id = {}
 
1856
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1857
        #       keeping a cache of directories that we have seen.
 
1858
 
 
1859
        def _process_entry(entry, path_info):
 
1860
            """Compare an entry and real disk to generate delta information.
 
1861
 
 
1862
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1863
                the path of entry. If None, then the path is considered absent.
 
1864
                (Perhaps we should pass in a concrete entry for this ?)
 
1865
                Basename is returned as a utf8 string because we expect this
 
1866
                tuple will be ignored, and don't want to take the time to
 
1867
                decode.
 
1868
            :return: None if these don't match
 
1869
                     A tuple of information about the change, or
 
1870
                     the object 'uninteresting' if these match, but are
 
1871
                     basically identical.
 
1872
            """
 
1873
            if source_index is None:
 
1874
                source_details = NULL_PARENT_DETAILS
 
1875
            else:
 
1876
                source_details = entry[1][source_index]
 
1877
            target_details = entry[1][target_index]
 
1878
            target_minikind = target_details[0]
 
1879
            if path_info is not None and target_minikind in 'fdlt':
 
1880
                assert target_index == 0
 
1881
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1882
                                                  stat_value=path_info[3])
 
1883
                # The entry may have been modified by update_entry
 
1884
                target_details = entry[1][target_index]
 
1885
                target_minikind = target_details[0]
 
1886
            else:
 
1887
                link_or_sha1 = None
 
1888
            file_id = entry[0][2]
 
1889
            source_minikind = source_details[0]
 
1890
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1891
                # claimed content in both: diff
 
1892
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1893
                #        |        |      | diff check on source-target
 
1894
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1895
                #        |        |      | ???
 
1896
                if source_minikind in 'r':
 
1897
                    # add the source to the search path to find any children it
 
1898
                    # has.  TODO ? : only add if it is a container ?
 
1899
                    if not osutils.is_inside_any(searched_specific_files,
 
1900
                                                 source_details[1]):
 
1901
                        search_specific_files.add(source_details[1])
 
1902
                    # generate the old path; this is needed for stating later
 
1903
                    # as well.
 
1904
                    old_path = source_details[1]
 
1905
                    old_dirname, old_basename = os.path.split(old_path)
 
1906
                    path = pathjoin(entry[0][0], entry[0][1])
 
1907
                    old_entry = state._get_entry(source_index,
 
1908
                                                 path_utf8=old_path)
 
1909
                    # update the source details variable to be the real
 
1910
                    # location.
 
1911
                    source_details = old_entry[1][source_index]
 
1912
                    source_minikind = source_details[0]
 
1913
                else:
 
1914
                    old_dirname = entry[0][0]
 
1915
                    old_basename = entry[0][1]
 
1916
                    old_path = path = None
 
1917
                if path_info is None:
 
1918
                    # the file is missing on disk, show as removed.
 
1919
                    content_change = True
 
1920
                    target_kind = None
 
1921
                    target_exec = False
 
1922
                else:
 
1923
                    # source and target are both versioned and disk file is present.
 
1924
                    target_kind = path_info[2]
 
1925
                    if target_kind == 'directory':
 
1926
                        if path is None:
 
1927
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1928
                        new_dirname_to_file_id[path] = file_id
 
1929
                        if source_minikind != 'd':
 
1930
                            content_change = True
 
1931
                        else:
 
1932
                            # directories have no fingerprint
 
1933
                            content_change = False
 
1934
                        target_exec = False
 
1935
                    elif target_kind == 'file':
 
1936
                        if source_minikind != 'f':
 
1937
                            content_change = True
 
1938
                        else:
 
1939
                            # We could check the size, but we already have the
 
1940
                            # sha1 hash.
 
1941
                            content_change = (link_or_sha1 != source_details[1])
 
1942
                        # Target details is updated at update_entry time
 
1943
                        if use_filesystem_for_exec:
 
1944
                            # We don't need S_ISREG here, because we are sure
 
1945
                            # we are dealing with a file.
 
1946
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1947
                        else:
 
1948
                            target_exec = target_details[3]
 
1949
                    elif target_kind == 'symlink':
 
1950
                        if source_minikind != 'l':
 
1951
                            content_change = True
 
1952
                        else:
 
1953
                            content_change = (link_or_sha1 != source_details[1])
 
1954
                        target_exec = False
 
1955
                    elif target_kind == 'tree-reference':
 
1956
                        if source_minikind != 't':
 
1957
                            content_change = True
 
1958
                        else:
 
1959
                            content_change = False
 
1960
                        target_exec = False
 
1961
                    else:
 
1962
                        raise Exception, "unknown kind %s" % path_info[2]
 
1963
                if source_minikind == 'd':
 
1964
                    if path is None:
 
1965
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1966
                    old_dirname_to_file_id[old_path] = file_id
 
1967
                # parent id is the entry for the path in the target tree
 
1968
                if old_dirname == last_source_parent[0]:
 
1969
                    source_parent_id = last_source_parent[1]
 
1970
                else:
 
1971
                    try:
 
1972
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1973
                    except KeyError:
 
1974
                        source_parent_entry = state._get_entry(source_index,
 
1975
                                                               path_utf8=old_dirname)
 
1976
                        source_parent_id = source_parent_entry[0][2]
 
1977
                    if source_parent_id == entry[0][2]:
 
1978
                        # This is the root, so the parent is None
 
1979
                        source_parent_id = None
 
1980
                    else:
 
1981
                        last_source_parent[0] = old_dirname
 
1982
                        last_source_parent[1] = source_parent_id
 
1983
                new_dirname = entry[0][0]
 
1984
                if new_dirname == last_target_parent[0]:
 
1985
                    target_parent_id = last_target_parent[1]
 
1986
                else:
 
1987
                    try:
 
1988
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1989
                    except KeyError:
 
1990
                        # TODO: We don't always need to do the lookup, because the
 
1991
                        #       parent entry will be the same as the source entry.
 
1992
                        target_parent_entry = state._get_entry(target_index,
 
1993
                                                               path_utf8=new_dirname)
 
1994
                        assert target_parent_entry != (None, None), (
 
1995
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1996
                            % (new_dirname, entry))
 
1997
                        target_parent_id = target_parent_entry[0][2]
 
1998
                    if target_parent_id == entry[0][2]:
 
1999
                        # This is the root, so the parent is None
 
2000
                        target_parent_id = None
 
2001
                    else:
 
2002
                        last_target_parent[0] = new_dirname
 
2003
                        last_target_parent[1] = target_parent_id
 
2004
 
 
2005
                source_exec = source_details[3]
 
2006
                if (include_unchanged
 
2007
                    or content_change
 
2008
                    or source_parent_id != target_parent_id
 
2009
                    or old_basename != entry[0][1]
 
2010
                    or source_exec != target_exec
 
2011
                    ):
 
2012
                    if old_path is None:
 
2013
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2014
                        old_path_u = utf8_decode(old_path)[0]
 
2015
                        path_u = old_path_u
 
2016
                    else:
 
2017
                        old_path_u = utf8_decode(old_path)[0]
 
2018
                        if old_path == path:
 
2019
                            path_u = old_path_u
 
2020
                        else:
 
2021
                            path_u = utf8_decode(path)[0]
 
2022
                    source_kind = _minikind_to_kind[source_minikind]
 
2023
                    return (entry[0][2],
 
2024
                           (old_path_u, path_u),
 
2025
                           content_change,
 
2026
                           (True, True),
 
2027
                           (source_parent_id, target_parent_id),
 
2028
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2029
                           (source_kind, target_kind),
 
2030
                           (source_exec, target_exec))
 
2031
                else:
 
2032
                    return uninteresting
 
2033
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2034
                # looks like a new file
 
2035
                if path_info is not None:
 
2036
                    path = pathjoin(entry[0][0], entry[0][1])
 
2037
                    # parent id is the entry for the path in the target tree
 
2038
                    # TODO: these are the same for an entire directory: cache em.
 
2039
                    parent_id = state._get_entry(target_index,
 
2040
                                                 path_utf8=entry[0][0])[0][2]
 
2041
                    if parent_id == entry[0][2]:
 
2042
                        parent_id = None
 
2043
                    if use_filesystem_for_exec:
 
2044
                        # We need S_ISREG here, because we aren't sure if this
 
2045
                        # is a file or not.
 
2046
                        target_exec = bool(
 
2047
                            stat.S_ISREG(path_info[3].st_mode)
 
2048
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2049
                    else:
 
2050
                        target_exec = target_details[3]
 
2051
                    return (entry[0][2],
 
2052
                           (None, utf8_decode(path)[0]),
 
2053
                           True,
 
2054
                           (False, True),
 
2055
                           (None, parent_id),
 
2056
                           (None, utf8_decode(entry[0][1])[0]),
 
2057
                           (None, path_info[2]),
 
2058
                           (None, target_exec))
 
2059
                else:
 
2060
                    # but its not on disk: we deliberately treat this as just
 
2061
                    # never-present. (Why ?! - RBC 20070224)
 
2062
                    pass
 
2063
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2064
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2065
                # if its still on disk, *and* theres no other entry at this
 
2066
                # path [we dont know this in this routine at the moment -
 
2067
                # perhaps we should change this - then it would be an unknown.
 
2068
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2069
                # parent id is the entry for the path in the target tree
 
2070
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2071
                if parent_id == entry[0][2]:
 
2072
                    parent_id = None
 
2073
                return (entry[0][2],
 
2074
                       (utf8_decode(old_path)[0], None),
 
2075
                       True,
 
2076
                       (True, False),
 
2077
                       (parent_id, None),
 
2078
                       (utf8_decode(entry[0][1])[0], None),
 
2079
                       (_minikind_to_kind[source_minikind], None),
 
2080
                       (source_details[3], None))
 
2081
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2082
                # a rename; could be a true rename, or a rename inherited from
 
2083
                # a renamed parent. TODO: handle this efficiently. Its not
 
2084
                # common case to rename dirs though, so a correct but slow
 
2085
                # implementation will do.
 
2086
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2087
                    search_specific_files.add(target_details[1])
 
2088
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2089
                # neither of the selected trees contain this file,
 
2090
                # so skip over it. This is not currently directly tested, but
 
2091
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2092
                pass
 
2093
            else:
 
2094
                raise AssertionError("don't know how to compare "
 
2095
                    "source_minikind=%r, target_minikind=%r"
 
2096
                    % (source_minikind, target_minikind))
 
2097
                ## import pdb;pdb.set_trace()
 
2098
            return None
 
2099
 
 
2100
        while search_specific_files:
 
2101
            # TODO: the pending list should be lexically sorted?  the
 
2102
            # interface doesn't require it.
 
2103
            current_root = search_specific_files.pop()
 
2104
            current_root_unicode = current_root.decode('utf8')
 
2105
            searched_specific_files.add(current_root)
 
2106
            # process the entries for this containing directory: the rest will be
 
2107
            # found by their parents recursively.
 
2108
            root_entries = _entries_for_path(current_root)
 
2109
            root_abspath = self.target.abspath(current_root_unicode)
 
2110
            try:
 
2111
                root_stat = os.lstat(root_abspath)
 
2112
            except OSError, e:
 
2113
                if e.errno == errno.ENOENT:
 
2114
                    # the path does not exist: let _process_entry know that.
 
2115
                    root_dir_info = None
 
2116
                else:
 
2117
                    # some other random error: hand it up.
 
2118
                    raise
 
2119
            else:
 
2120
                root_dir_info = ('', current_root,
 
2121
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2122
                    root_abspath)
 
2123
                if root_dir_info[2] == 'directory':
 
2124
                    if self.target._directory_is_tree_reference(
 
2125
                        current_root.decode('utf8')):
 
2126
                        root_dir_info = root_dir_info[:2] + \
 
2127
                            ('tree-reference',) + root_dir_info[3:]
 
2128
 
 
2129
            if not root_entries and not root_dir_info:
 
2130
                # this specified path is not present at all, skip it.
 
2131
                continue
 
2132
            path_handled = False
 
2133
            for entry in root_entries:
 
2134
                result = _process_entry(entry, root_dir_info)
 
2135
                if result is not None:
 
2136
                    path_handled = True
 
2137
                    if result is not uninteresting:
 
2138
                        yield result
 
2139
            if want_unversioned and not path_handled and root_dir_info:
 
2140
                new_executable = bool(
 
2141
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2142
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2143
                yield (None,
 
2144
                       (None, current_root_unicode),
 
2145
                       True,
 
2146
                       (False, False),
 
2147
                       (None, None),
 
2148
                       (None, splitpath(current_root_unicode)[-1]),
 
2149
                       (None, root_dir_info[2]),
 
2150
                       (None, new_executable)
 
2151
                      )
 
2152
            initial_key = (current_root, '', '')
 
2153
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2154
            if block_index == 0:
 
2155
                # we have processed the total root already, but because the
 
2156
                # initial key matched it we should skip it here.
 
2157
                block_index +=1
 
2158
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2159
                current_dir_info = None
 
2160
            else:
 
2161
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2162
                try:
 
2163
                    current_dir_info = dir_iterator.next()
 
2164
                except OSError, e:
 
2165
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2166
                    # python 2.5 has e.errno == EINVAL,
 
2167
                    #            and e.winerror == ERROR_DIRECTORY
 
2168
                    e_winerror = getattr(e, 'winerror', None)
 
2169
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2170
                    # there may be directories in the inventory even though
 
2171
                    # this path is not a file on disk: so mark it as end of
 
2172
                    # iterator
 
2173
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2174
                        current_dir_info = None
 
2175
                    elif (sys.platform == 'win32'
 
2176
                          and (e.errno in win_errors
 
2177
                               or e_winerror in win_errors)):
 
2178
                        current_dir_info = None
 
2179
                    else:
 
2180
                        raise
 
2181
                else:
 
2182
                    if current_dir_info[0][0] == '':
 
2183
                        # remove .bzr from iteration
 
2184
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2185
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2186
                        del current_dir_info[1][bzr_index]
 
2187
            # walk until both the directory listing and the versioned metadata
 
2188
            # are exhausted. 
 
2189
            if (block_index < len(state._dirblocks) and
 
2190
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2191
                current_block = state._dirblocks[block_index]
 
2192
            else:
 
2193
                current_block = None
 
2194
            while (current_dir_info is not None or
 
2195
                   current_block is not None):
 
2196
                if (current_dir_info and current_block
 
2197
                    and current_dir_info[0][0] != current_block[0]):
 
2198
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2199
                        # filesystem data refers to paths not covered by the dirblock.
 
2200
                        # this has two possibilities:
 
2201
                        # A) it is versioned but empty, so there is no block for it
 
2202
                        # B) it is not versioned.
 
2203
 
 
2204
                        # if (A) then we need to recurse into it to check for
 
2205
                        # new unknown files or directories.
 
2206
                        # if (B) then we should ignore it, because we don't
 
2207
                        # recurse into unknown directories.
 
2208
                        path_index = 0
 
2209
                        while path_index < len(current_dir_info[1]):
 
2210
                                current_path_info = current_dir_info[1][path_index]
 
2211
                                if want_unversioned:
 
2212
                                    if current_path_info[2] == 'directory':
 
2213
                                        if self.target._directory_is_tree_reference(
 
2214
                                            current_path_info[0].decode('utf8')):
 
2215
                                            current_path_info = current_path_info[:2] + \
 
2216
                                                ('tree-reference',) + current_path_info[3:]
 
2217
                                    new_executable = bool(
 
2218
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2219
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2220
                                    yield (None,
 
2221
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2222
                                        True,
 
2223
                                        (False, False),
 
2224
                                        (None, None),
 
2225
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2226
                                        (None, current_path_info[2]),
 
2227
                                        (None, new_executable))
 
2228
                                # dont descend into this unversioned path if it is
 
2229
                                # a dir
 
2230
                                if current_path_info[2] in ('directory',
 
2231
                                                            'tree-reference'):
 
2232
                                    del current_dir_info[1][path_index]
 
2233
                                    path_index -= 1
 
2234
                                path_index += 1
 
2235
 
 
2236
                        # This dir info has been handled, go to the next
 
2237
                        try:
 
2238
                            current_dir_info = dir_iterator.next()
 
2239
                        except StopIteration:
 
2240
                            current_dir_info = None
 
2241
                    else:
 
2242
                        # We have a dirblock entry for this location, but there
 
2243
                        # is no filesystem path for this. This is most likely
 
2244
                        # because a directory was removed from the disk.
 
2245
                        # We don't have to report the missing directory,
 
2246
                        # because that should have already been handled, but we
 
2247
                        # need to handle all of the files that are contained
 
2248
                        # within.
 
2249
                        for current_entry in current_block[1]:
 
2250
                            # entry referring to file not present on disk.
 
2251
                            # advance the entry only, after processing.
 
2252
                            result = _process_entry(current_entry, None)
 
2253
                            if result is not None:
 
2254
                                if result is not uninteresting:
 
2255
                                    yield result
 
2256
                        block_index +=1
 
2257
                        if (block_index < len(state._dirblocks) and
 
2258
                            osutils.is_inside(current_root,
 
2259
                                              state._dirblocks[block_index][0])):
 
2260
                            current_block = state._dirblocks[block_index]
 
2261
                        else:
 
2262
                            current_block = None
 
2263
                    continue
 
2264
                entry_index = 0
 
2265
                if current_block and entry_index < len(current_block[1]):
 
2266
                    current_entry = current_block[1][entry_index]
 
2267
                else:
 
2268
                    current_entry = None
 
2269
                advance_entry = True
 
2270
                path_index = 0
 
2271
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2272
                    current_path_info = current_dir_info[1][path_index]
 
2273
                    if current_path_info[2] == 'directory':
 
2274
                        if self.target._directory_is_tree_reference(
 
2275
                            current_path_info[0].decode('utf8')):
 
2276
                            current_path_info = current_path_info[:2] + \
 
2277
                                ('tree-reference',) + current_path_info[3:]
 
2278
                else:
 
2279
                    current_path_info = None
 
2280
                advance_path = True
 
2281
                path_handled = False
 
2282
                while (current_entry is not None or
 
2283
                    current_path_info is not None):
 
2284
                    if current_entry is None:
 
2285
                        # the check for path_handled when the path is adnvaced
 
2286
                        # will yield this path if needed.
 
2287
                        pass
 
2288
                    elif current_path_info is None:
 
2289
                        # no path is fine: the per entry code will handle it.
 
2290
                        result = _process_entry(current_entry, current_path_info)
 
2291
                        if result is not None:
 
2292
                            if result is not uninteresting:
 
2293
                                yield result
 
2294
                    elif (current_entry[0][1] != current_path_info[1]
 
2295
                          or current_entry[1][target_index][0] in 'ar'):
 
2296
                        # The current path on disk doesn't match the dirblock
 
2297
                        # record. Either the dirblock is marked as absent, or
 
2298
                        # the file on disk is not present at all in the
 
2299
                        # dirblock. Either way, report about the dirblock
 
2300
                        # entry, and let other code handle the filesystem one.
 
2301
 
 
2302
                        # Compare the basename for these files to determine
 
2303
                        # which comes first
 
2304
                        if current_path_info[1] < current_entry[0][1]:
 
2305
                            # extra file on disk: pass for now, but only
 
2306
                            # increment the path, not the entry
 
2307
                            advance_entry = False
 
2308
                        else:
 
2309
                            # entry referring to file not present on disk.
 
2310
                            # advance the entry only, after processing.
 
2311
                            result = _process_entry(current_entry, None)
 
2312
                            if result is not None:
 
2313
                                if result is not uninteresting:
 
2314
                                    yield result
 
2315
                            advance_path = False
 
2316
                    else:
 
2317
                        result = _process_entry(current_entry, current_path_info)
 
2318
                        if result is not None:
 
2319
                            path_handled = True
 
2320
                            if result is not uninteresting:
 
2321
                                yield result
 
2322
                    if advance_entry and current_entry is not None:
 
2323
                        entry_index += 1
 
2324
                        if entry_index < len(current_block[1]):
 
2325
                            current_entry = current_block[1][entry_index]
 
2326
                        else:
 
2327
                            current_entry = None
 
2328
                    else:
 
2329
                        advance_entry = True # reset the advance flaga
 
2330
                    if advance_path and current_path_info is not None:
 
2331
                        if not path_handled:
 
2332
                            # unversioned in all regards
 
2333
                            if want_unversioned:
 
2334
                                new_executable = bool(
 
2335
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2336
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2337
                                yield (None,
 
2338
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2339
                                    True,
 
2340
                                    (False, False),
 
2341
                                    (None, None),
 
2342
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2343
                                    (None, current_path_info[2]),
 
2344
                                    (None, new_executable))
 
2345
                            # dont descend into this unversioned path if it is
 
2346
                            # a dir
 
2347
                            if current_path_info[2] in ('directory'):
 
2348
                                del current_dir_info[1][path_index]
 
2349
                                path_index -= 1
 
2350
                        # dont descend the disk iterator into any tree 
 
2351
                        # paths.
 
2352
                        if current_path_info[2] == 'tree-reference':
 
2353
                            del current_dir_info[1][path_index]
 
2354
                            path_index -= 1
 
2355
                        path_index += 1
 
2356
                        if path_index < len(current_dir_info[1]):
 
2357
                            current_path_info = current_dir_info[1][path_index]
 
2358
                            if current_path_info[2] == 'directory':
 
2359
                                if self.target._directory_is_tree_reference(
 
2360
                                    current_path_info[0].decode('utf8')):
 
2361
                                    current_path_info = current_path_info[:2] + \
 
2362
                                        ('tree-reference',) + current_path_info[3:]
 
2363
                        else:
 
2364
                            current_path_info = None
 
2365
                        path_handled = False
 
2366
                    else:
 
2367
                        advance_path = True # reset the advance flagg.
 
2368
                if current_block is not None:
 
2369
                    block_index += 1
 
2370
                    if (block_index < len(state._dirblocks) and
 
2371
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2372
                        current_block = state._dirblocks[block_index]
 
2373
                    else:
 
2374
                        current_block = None
 
2375
                if current_dir_info is not None:
 
2376
                    try:
 
2377
                        current_dir_info = dir_iterator.next()
 
2378
                    except StopIteration:
 
2379
                        current_dir_info = None
 
2380
 
 
2381
 
 
2382
    @staticmethod
 
2383
    def is_compatible(source, target):
 
2384
        # the target must be a dirstate working tree
 
2385
        if not isinstance(target, WorkingTree4):
 
2386
            return False
 
2387
        # the source must be a revtreee or dirstate rev tree.
 
2388
        if not isinstance(source,
 
2389
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2390
            return False
 
2391
        # the source revid must be in the target dirstate
 
2392
        if not (source._revision_id == NULL_REVISION or
 
2393
            source._revision_id in target.get_parent_ids()):
 
2394
            # TODO: what about ghosts? it may well need to 
 
2395
            # check for them explicitly.
 
2396
            return False
 
2397
        return True
 
2398
 
 
2399
InterTree.register_optimiser(InterDirStateTree)
 
2400
 
 
2401
 
 
2402
class Converter3to4(object):
 
2403
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2404
 
 
2405
    def __init__(self):
 
2406
        self.target_format = WorkingTreeFormat4()
 
2407
 
 
2408
    def convert(self, tree):
 
2409
        # lock the control files not the tree, so that we dont get tree
 
2410
        # on-unlock behaviours, and so that noone else diddles with the 
 
2411
        # tree during upgrade.
 
2412
        tree._control_files.lock_write()
 
2413
        try:
 
2414
            tree.read_working_inventory()
 
2415
            self.create_dirstate_data(tree)
 
2416
            self.update_format(tree)
 
2417
            self.remove_xml_files(tree)
 
2418
        finally:
 
2419
            tree._control_files.unlock()
 
2420
 
 
2421
    def create_dirstate_data(self, tree):
 
2422
        """Create the dirstate based data for tree."""
 
2423
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2424
            ).local_abspath('dirstate')
 
2425
        state = dirstate.DirState.from_tree(tree, local_path)
 
2426
        state.save()
 
2427
        state.unlock()
 
2428
 
 
2429
    def remove_xml_files(self, tree):
 
2430
        """Remove the oldformat 3 data."""
 
2431
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2432
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2433
            'pending-merges', 'stat-cache']:
 
2434
            try:
 
2435
                transport.delete(path)
 
2436
            except errors.NoSuchFile:
 
2437
                # some files are optional - just deal.
 
2438
                pass
 
2439
 
 
2440
    def update_format(self, tree):
 
2441
        """Change the format marker."""
 
2442
        tree._control_files.put_utf8('format',
 
2443
            self.target_format.get_format_string())