~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2007-07-25 00:52:21 UTC
  • mfrom: (2650 +trunk)
  • mto: This revision was merged to the branch mainline in revision 2651.
  • Revision ID: robertc@robertcollins.net-20070725005221-0ysm6il5mqnme3wz
Merge bzr.dev.

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
 
47
    dirstate,
 
48
    errors,
 
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    transform,
 
59
    urlutils,
 
60
    xml5,
 
61
    xml6,
 
62
    )
 
63
import bzrlib.branch
 
64
from bzrlib.transport import get_transport
 
65
import bzrlib.ui
 
66
""")
 
67
 
 
68
from bzrlib import symbol_versioning
 
69
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
70
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
71
from bzrlib.lockable_files import LockableFiles, TransportLock
 
72
from bzrlib.lockdir import LockDir
 
73
import bzrlib.mutabletree
 
74
from bzrlib.mutabletree import needs_tree_write_lock
 
75
from bzrlib.osutils import (
 
76
    file_kind,
 
77
    isdir,
 
78
    normpath,
 
79
    pathjoin,
 
80
    rand_chars,
 
81
    realpath,
 
82
    safe_unicode,
 
83
    splitpath,
 
84
    )
 
85
from bzrlib.trace import mutter, note
 
86
from bzrlib.transport.local import LocalTransport
 
87
from bzrlib.tree import InterTree
 
88
from bzrlib.progress import DummyProgress, ProgressPhase
 
89
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
90
from bzrlib.rio import RioReader, rio_file, Stanza
 
91
from bzrlib.symbol_versioning import (deprecated_passed,
 
92
        deprecated_method,
 
93
        deprecated_function,
 
94
        DEPRECATED_PARAMETER,
 
95
        )
 
96
from bzrlib.tree import Tree
 
97
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
98
 
 
99
 
 
100
# This is the Windows equivalent of ENOTDIR
 
101
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
102
# just an error code.
 
103
ERROR_PATH_NOT_FOUND = 3
 
104
ERROR_DIRECTORY = 267
 
105
 
 
106
 
 
107
class WorkingTree4(WorkingTree3):
 
108
    """This is the Format 4 working tree.
 
109
 
 
110
    This differs from WorkingTree3 by:
 
111
     - Having a consolidated internal dirstate, stored in a
 
112
       randomly-accessible sorted file on disk.
 
113
     - Not having a regular inventory attribute.  One can be synthesized 
 
114
       on demand but this is expensive and should be avoided.
 
115
 
 
116
    This is new in bzr 0.15.
 
117
    """
 
118
 
 
119
    def __init__(self, basedir,
 
120
                 branch,
 
121
                 _control_files=None,
 
122
                 _format=None,
 
123
                 _bzrdir=None):
 
124
        """Construct a WorkingTree for basedir.
 
125
 
 
126
        If the branch is not supplied, it is opened automatically.
 
127
        If the branch is supplied, it must be the branch for this basedir.
 
128
        (branch.base is not cross checked, because for remote branches that
 
129
        would be meaningless).
 
130
        """
 
131
        self._format = _format
 
132
        self.bzrdir = _bzrdir
 
133
        from bzrlib.trace import note, mutter
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
 
136
        basedir = safe_unicode(basedir)
 
137
        mutter("opening working tree %r", basedir)
 
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
 
141
        self.basedir = realpath(basedir)
 
142
        # if branch is at our basedir and is a format 6 or less
 
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
 
146
        self._control_files = _control_files
 
147
        self._dirty = None
 
148
        #-------------
 
149
        # during a read or write lock these objects are set, and are
 
150
        # None the rest of the time.
 
151
        self._dirstate = None
 
152
        self._inventory = None
 
153
        #-------------
 
154
 
 
155
    @needs_tree_write_lock
 
156
    def _add(self, files, ids, kinds):
 
157
        """See MutableTree._add."""
 
158
        state = self.current_dirstate()
 
159
        for f, file_id, kind in zip(files, ids, kinds):
 
160
            f = f.strip('/')
 
161
            assert '//' not in f
 
162
            assert '..' not in f
 
163
            if self.path2id(f):
 
164
                # special case tree root handling.
 
165
                if f == '' and self.path2id(f) == ROOT_ID:
 
166
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
167
                continue
 
168
            if file_id is None:
 
169
                file_id = generate_ids.gen_file_id(f)
 
170
            # deliberately add the file with no cached stat or sha1
 
171
            # - on the first access it will be gathered, and we can
 
172
            # always change this once tests are all passing.
 
173
            state.add(f, file_id, kind, None, '')
 
174
        self._make_dirty(reset_inventory=True)
 
175
 
 
176
    def _make_dirty(self, reset_inventory):
 
177
        """Make the tree state dirty.
 
178
 
 
179
        :param reset_inventory: True if the cached inventory should be removed
 
180
            (presuming there is one).
 
181
        """
 
182
        self._dirty = True
 
183
        if reset_inventory and self._inventory is not None:
 
184
            self._inventory = None
 
185
 
 
186
    @needs_tree_write_lock
 
187
    def add_reference(self, sub_tree):
 
188
        # use standard implementation, which calls back to self._add
 
189
        # 
 
190
        # So we don't store the reference_revision in the working dirstate,
 
191
        # it's just recorded at the moment of commit. 
 
192
        self._add_reference(sub_tree)
 
193
 
 
194
    def break_lock(self):
 
195
        """Break a lock if one is present from another instance.
 
196
 
 
197
        Uses the ui factory to ask for confirmation if the lock may be from
 
198
        an active process.
 
199
 
 
200
        This will probe the repository for its lock as well.
 
201
        """
 
202
        # if the dirstate is locked by an active process, reject the break lock
 
203
        # call.
 
204
        try:
 
205
            if self._dirstate is None:
 
206
                clear = True
 
207
            else:
 
208
                clear = False
 
209
            state = self._current_dirstate()
 
210
            if state._lock_token is not None:
 
211
                # we already have it locked. sheese, cant break our own lock.
 
212
                raise errors.LockActive(self.basedir)
 
213
            else:
 
214
                try:
 
215
                    # try for a write lock - need permission to get one anyhow
 
216
                    # to break locks.
 
217
                    state.lock_write()
 
218
                except errors.LockContention:
 
219
                    # oslocks fail when a process is still live: fail.
 
220
                    # TODO: get the locked lockdir info and give to the user to
 
221
                    # assist in debugging.
 
222
                    raise errors.LockActive(self.basedir)
 
223
                else:
 
224
                    state.unlock()
 
225
        finally:
 
226
            if clear:
 
227
                self._dirstate = None
 
228
        self._control_files.break_lock()
 
229
        self.branch.break_lock()
 
230
 
 
231
    def _comparison_data(self, entry, path):
 
232
        kind, executable, stat_value = \
 
233
            WorkingTree3._comparison_data(self, entry, path)
 
234
        # it looks like a plain directory, but it's really a reference -- see
 
235
        # also kind()
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
 
239
            kind = 'tree-reference'
 
240
        return kind, executable, stat_value
 
241
 
 
242
    @needs_write_lock
 
243
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
244
        # mark the tree as dirty post commit - commit
 
245
        # can change the current versioned list by doing deletes.
 
246
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
247
        self._make_dirty(reset_inventory=True)
 
248
        return result
 
249
 
 
250
    def current_dirstate(self):
 
251
        """Return the current dirstate object.
 
252
 
 
253
        This is not part of the tree interface and only exposed for ease of
 
254
        testing.
 
255
 
 
256
        :raises errors.NotWriteLocked: when not in a lock.
 
257
        """
 
258
        self._must_be_locked()
 
259
        return self._current_dirstate()
 
260
 
 
261
    def _current_dirstate(self):
 
262
        """Internal function that does not check lock status.
 
263
 
 
264
        This is needed for break_lock which also needs the dirstate.
 
265
        """
 
266
        if self._dirstate is not None:
 
267
            return self._dirstate
 
268
        local_path = self.bzrdir.get_workingtree_transport(None
 
269
            ).local_abspath('dirstate')
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
 
271
        return self._dirstate
 
272
 
 
273
    def _directory_is_tree_reference(self, relpath):
 
274
        # as a special case, if a directory contains control files then 
 
275
        # it's a tree reference, except that the root of the tree is not
 
276
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
277
        # TODO: We could ask all the control formats whether they
 
278
        # recognize this directory, but at the moment there's no cheap api
 
279
        # to do that.  Since we probably can only nest bzr checkouts and
 
280
        # they always use this name it's ok for now.  -- mbp 20060306
 
281
        #
 
282
        # FIXME: There is an unhandled case here of a subdirectory
 
283
        # containing .bzr but not a branch; that will probably blow up
 
284
        # when you try to commit it.  It might happen if there is a
 
285
        # checkout in a subdirectory.  This can be avoided by not adding
 
286
        # it.  mbp 20070306
 
287
 
 
288
    def filter_unversioned_files(self, paths):
 
289
        """Filter out paths that are versioned.
 
290
 
 
291
        :return: set of paths.
 
292
        """
 
293
        # TODO: make a generic multi-bisect routine roughly that should list
 
294
        # the paths, then process one half at a time recursively, and feed the
 
295
        # results of each bisect in further still
 
296
        paths = sorted(paths)
 
297
        result = set()
 
298
        state = self.current_dirstate()
 
299
        # TODO we want a paths_to_dirblocks helper I think
 
300
        for path in paths:
 
301
            dirname, basename = os.path.split(path.encode('utf8'))
 
302
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
303
                dirname, basename, 0)
 
304
            if not path_is_versioned:
 
305
                result.add(path)
 
306
        return result
 
307
 
 
308
    def flush(self):
 
309
        """Write all cached data to disk."""
 
310
        if self._control_files._lock_mode != 'w':
 
311
            raise errors.NotWriteLocked(self)
 
312
        self.current_dirstate().save()
 
313
        self._inventory = None
 
314
        self._dirty = False
 
315
 
 
316
    @needs_tree_write_lock
 
317
    def _gather_kinds(self, files, kinds):
 
318
        """See MutableTree._gather_kinds."""
 
319
        for pos, f in enumerate(files):
 
320
            if kinds[pos] is None:
 
321
                kinds[pos] = self._kind(f)
 
322
 
 
323
    def _generate_inventory(self):
 
324
        """Create and set self.inventory from the dirstate object.
 
325
        
 
326
        This is relatively expensive: we have to walk the entire dirstate.
 
327
        Ideally we would not, and can deprecate this function.
 
328
        """
 
329
        #: uncomment to trap on inventory requests.
 
330
        # import pdb;pdb.set_trace()
 
331
        state = self.current_dirstate()
 
332
        state._read_dirblocks_if_needed()
 
333
        root_key, current_entry = self._get_entry(path='')
 
334
        current_id = root_key[2]
 
335
        assert current_entry[0][0] == 'd' # directory
 
336
        inv = Inventory(root_id=current_id)
 
337
        # Turn some things into local variables
 
338
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
339
        factory = entry_factory
 
340
        utf8_decode = cache_utf8._utf8_decode
 
341
        inv_byid = inv._byid
 
342
        # we could do this straight out of the dirstate; it might be fast
 
343
        # and should be profiled - RBC 20070216
 
344
        parent_ies = {'' : inv.root}
 
345
        for block in state._dirblocks[1:]: # skip the root
 
346
            dirname = block[0]
 
347
            try:
 
348
                parent_ie = parent_ies[dirname]
 
349
            except KeyError:
 
350
                # all the paths in this block are not versioned in this tree
 
351
                continue
 
352
            for key, entry in block[1]:
 
353
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
354
                if minikind in ('a', 'r'): # absent, relocated
 
355
                    # a parent tree only entry
 
356
                    continue
 
357
                name = key[1]
 
358
                name_unicode = utf8_decode(name)[0]
 
359
                file_id = key[2]
 
360
                kind = minikind_to_kind[minikind]
 
361
                inv_entry = factory[kind](file_id, name_unicode,
 
362
                                          parent_ie.file_id)
 
363
                if kind == 'file':
 
364
                    # This is only needed on win32, where this is the only way
 
365
                    # we know the executable bit.
 
366
                    inv_entry.executable = executable
 
367
                    # not strictly needed: working tree
 
368
                    #inv_entry.text_size = size
 
369
                    #inv_entry.text_sha1 = sha1
 
370
                elif kind == 'directory':
 
371
                    # add this entry to the parent map.
 
372
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
373
                elif kind == 'tree-reference':
 
374
                    assert self._repo_supports_tree_reference, \
 
375
                        "repository of %r " \
 
376
                        "doesn't support tree references " \
 
377
                        "required by entry %r" \
 
378
                        % (self, name)
 
379
                    inv_entry.reference_revision = link_or_sha1 or None
 
380
                elif kind != 'symlink':
 
381
                    raise AssertionError("unknown kind %r" % kind)
 
382
                # These checks cost us around 40ms on a 55k entry tree
 
383
                assert file_id not in inv_byid, ('file_id %s already in'
 
384
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
385
                assert name_unicode not in parent_ie.children
 
386
                inv_byid[file_id] = inv_entry
 
387
                parent_ie.children[name_unicode] = inv_entry
 
388
        self._inventory = inv
 
389
 
 
390
    def _get_entry(self, file_id=None, path=None):
 
391
        """Get the dirstate row for file_id or path.
 
392
 
 
393
        If either file_id or path is supplied, it is used as the key to lookup.
 
394
        If both are supplied, the fastest lookup is used, and an error is
 
395
        raised if they do not both point at the same row.
 
396
        
 
397
        :param file_id: An optional unicode file_id to be looked up.
 
398
        :param path: An optional unicode path to be looked up.
 
399
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
400
        """
 
401
        if file_id is None and path is None:
 
402
            raise errors.BzrError('must supply file_id or path')
 
403
        state = self.current_dirstate()
 
404
        if path is not None:
 
405
            path = path.encode('utf8')
 
406
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
407
 
 
408
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
409
        # check file id is valid unconditionally.
 
410
        entry = self._get_entry(file_id=file_id, path=path)
 
411
        if entry[0] is None:
 
412
            raise errors.NoSuchId(self, file_id)
 
413
        if path is None:
 
414
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
415
 
 
416
        file_abspath = self.abspath(path)
 
417
        state = self.current_dirstate()
 
418
        if stat_value is None:
 
419
            try:
 
420
                stat_value = os.lstat(file_abspath)
 
421
            except OSError, e:
 
422
                if e.errno == errno.ENOENT:
 
423
                    return None
 
424
                else:
 
425
                    raise
 
426
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
427
                                          stat_value=stat_value)
 
428
        if entry[1][0][0] == 'f':
 
429
            return link_or_sha1
 
430
        return None
 
431
 
 
432
    def _get_inventory(self):
 
433
        """Get the inventory for the tree. This is only valid within a lock."""
 
434
        if self._inventory is not None:
 
435
            return self._inventory
 
436
        self._must_be_locked()
 
437
        self._generate_inventory()
 
438
        return self._inventory
 
439
 
 
440
    inventory = property(_get_inventory,
 
441
                         doc="Inventory of this Tree")
 
442
 
 
443
    @needs_read_lock
 
444
    def get_parent_ids(self):
 
445
        """See Tree.get_parent_ids.
 
446
        
 
447
        This implementation requests the ids list from the dirstate file.
 
448
        """
 
449
        return self.current_dirstate().get_parent_ids()
 
450
 
 
451
    def get_reference_revision(self, file_id, path=None):
 
452
        # referenced tree's revision is whatever's currently there
 
453
        return self.get_nested_tree(file_id, path).last_revision()
 
454
 
 
455
    def get_nested_tree(self, file_id, path=None):
 
456
        if path is None:
 
457
            path = self.id2path(file_id)
 
458
        # else: check file_id is at path?
 
459
        return WorkingTree.open(self.abspath(path))
 
460
 
 
461
    @needs_read_lock
 
462
    def get_root_id(self):
 
463
        """Return the id of this trees root"""
 
464
        return self._get_entry(path='')[0][2]
 
465
 
 
466
    def has_id(self, file_id):
 
467
        state = self.current_dirstate()
 
468
        file_id = osutils.safe_file_id(file_id)
 
469
        row, parents = self._get_entry(file_id=file_id)
 
470
        if row is None:
 
471
            return False
 
472
        return osutils.lexists(pathjoin(
 
473
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
474
 
 
475
    @needs_read_lock
 
476
    def id2path(self, file_id):
 
477
        "Convert a file-id to a path."
 
478
        file_id = osutils.safe_file_id(file_id)
 
479
        state = self.current_dirstate()
 
480
        entry = self._get_entry(file_id=file_id)
 
481
        if entry == (None, None):
 
482
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
483
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
484
        return path_utf8.decode('utf8')
 
485
 
 
486
    if not osutils.supports_executable():
 
487
        def is_executable(self, file_id, path=None):
 
488
            """Test if a file is executable or not.
 
489
 
 
490
            Note: The caller is expected to take a read-lock before calling this.
 
491
            """
 
492
            file_id = osutils.safe_file_id(file_id)
 
493
            entry = self._get_entry(file_id=file_id, path=path)
 
494
            if entry == (None, None):
 
495
                return False
 
496
            return entry[1][0][3]
 
497
    else:
 
498
        def is_executable(self, file_id, path=None):
 
499
            """Test if a file is executable or not.
 
500
 
 
501
            Note: The caller is expected to take a read-lock before calling this.
 
502
            """
 
503
            if not path:
 
504
                file_id = osutils.safe_file_id(file_id)
 
505
                path = self.id2path(file_id)
 
506
            mode = os.lstat(self.abspath(path)).st_mode
 
507
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
508
 
 
509
    @needs_read_lock
 
510
    def __iter__(self):
 
511
        """Iterate through file_ids for this tree.
 
512
 
 
513
        file_ids are in a WorkingTree if they are in the working inventory
 
514
        and the working file exists.
 
515
        """
 
516
        result = []
 
517
        for key, tree_details in self.current_dirstate()._iter_entries():
 
518
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
519
                # not relevant to the working tree
 
520
                continue
 
521
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
522
            if osutils.lexists(path):
 
523
                result.append(key[2])
 
524
        return iter(result)
 
525
 
 
526
    def iter_references(self):
 
527
        for key, tree_details in self.current_dirstate()._iter_entries():
 
528
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
529
                # not relevant to the working tree
 
530
                continue
 
531
            if not key[1]:
 
532
                # the root is not a reference.
 
533
                continue
 
534
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
535
            try:
 
536
                if self._kind(path) == 'tree-reference':
 
537
                    yield path, key[2]
 
538
            except errors.NoSuchFile:
 
539
                # path is missing on disk.
 
540
                continue
 
541
 
 
542
    def kind(self, file_id):
 
543
        """Return the kind of a file.
 
544
 
 
545
        This is always the actual kind that's on disk, regardless of what it
 
546
        was added as.
 
547
 
 
548
        Note: The caller is expected to take a read-lock before calling this.
 
549
        """
 
550
        relpath = self.id2path(file_id)
 
551
        assert relpath != None, \
 
552
            "path for id {%s} is None!" % file_id
 
553
        return self._kind(relpath)
 
554
 
 
555
    def _kind(self, relpath):
 
556
        abspath = self.abspath(relpath)
 
557
        kind = file_kind(abspath)
 
558
        if (self._repo_supports_tree_reference and
 
559
            kind == 'directory' and
 
560
            self._directory_is_tree_reference(relpath)):
 
561
            kind = 'tree-reference'
 
562
        return kind
 
563
 
 
564
    @needs_read_lock
 
565
    def _last_revision(self):
 
566
        """See Mutable.last_revision."""
 
567
        parent_ids = self.current_dirstate().get_parent_ids()
 
568
        if parent_ids:
 
569
            return parent_ids[0]
 
570
        else:
 
571
            return None
 
572
 
 
573
    def lock_read(self):
 
574
        """See Branch.lock_read, and WorkingTree.unlock."""
 
575
        self.branch.lock_read()
 
576
        try:
 
577
            self._control_files.lock_read()
 
578
            try:
 
579
                state = self.current_dirstate()
 
580
                if not state._lock_token:
 
581
                    state.lock_read()
 
582
                # set our support for tree references from the repository in
 
583
                # use.
 
584
                self._repo_supports_tree_reference = getattr(
 
585
                    self.branch.repository._format, "supports_tree_reference",
 
586
                    False)
 
587
            except:
 
588
                self._control_files.unlock()
 
589
                raise
 
590
        except:
 
591
            self.branch.unlock()
 
592
            raise
 
593
 
 
594
    def _lock_self_write(self):
 
595
        """This should be called after the branch is locked."""
 
596
        try:
 
597
            self._control_files.lock_write()
 
598
            try:
 
599
                state = self.current_dirstate()
 
600
                if not state._lock_token:
 
601
                    state.lock_write()
 
602
                # set our support for tree references from the repository in
 
603
                # use.
 
604
                self._repo_supports_tree_reference = getattr(
 
605
                    self.branch.repository._format, "supports_tree_reference",
 
606
                    False)
 
607
            except:
 
608
                self._control_files.unlock()
 
609
                raise
 
610
        except:
 
611
            self.branch.unlock()
 
612
            raise
 
613
 
 
614
    def lock_tree_write(self):
 
615
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
616
        self.branch.lock_read()
 
617
        self._lock_self_write()
 
618
 
 
619
    def lock_write(self):
 
620
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
621
        self.branch.lock_write()
 
622
        self._lock_self_write()
 
623
 
 
624
    @needs_tree_write_lock
 
625
    def move(self, from_paths, to_dir, after=False):
 
626
        """See WorkingTree.move()."""
 
627
        result = []
 
628
        if not from_paths:
 
629
            return result
 
630
 
 
631
        state = self.current_dirstate()
 
632
 
 
633
        assert not isinstance(from_paths, basestring)
 
634
        to_dir_utf8 = to_dir.encode('utf8')
 
635
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
636
        id_index = state._get_id_index()
 
637
        # check destination directory
 
638
        # get the details for it
 
639
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
640
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
641
        if not entry_present:
 
642
            raise errors.BzrMoveFailedError('', to_dir,
 
643
                errors.NotVersionedError(to_dir))
 
644
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
645
        # get a handle on the block itself.
 
646
        to_block_index = state._ensure_block(
 
647
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
648
        to_block = state._dirblocks[to_block_index]
 
649
        to_abs = self.abspath(to_dir)
 
650
        if not isdir(to_abs):
 
651
            raise errors.BzrMoveFailedError('',to_dir,
 
652
                errors.NotADirectory(to_abs))
 
653
 
 
654
        if to_entry[1][0][0] != 'd':
 
655
            raise errors.BzrMoveFailedError('',to_dir,
 
656
                errors.NotADirectory(to_abs))
 
657
 
 
658
        if self._inventory is not None:
 
659
            update_inventory = True
 
660
            inv = self.inventory
 
661
            to_dir_ie = inv[to_dir_id]
 
662
            to_dir_id = to_entry[0][2]
 
663
        else:
 
664
            update_inventory = False
 
665
 
 
666
        rollbacks = []
 
667
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
668
                     fingerprint, packed_stat, size,
 
669
                     to_block, to_key, to_path_utf8):
 
670
            state._make_absent(old_entry)
 
671
            from_key = old_entry[0]
 
672
            rollbacks.append(
 
673
                lambda:state.update_minimal(from_key,
 
674
                    minikind,
 
675
                    executable=executable,
 
676
                    fingerprint=fingerprint,
 
677
                    packed_stat=packed_stat,
 
678
                    size=size,
 
679
                    path_utf8=from_path_utf8))
 
680
            state.update_minimal(to_key,
 
681
                    minikind,
 
682
                    executable=executable,
 
683
                    fingerprint=fingerprint,
 
684
                    packed_stat=packed_stat,
 
685
                    size=size,
 
686
                    path_utf8=to_path_utf8)
 
687
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
688
            new_entry = to_block[1][added_entry_index]
 
689
            rollbacks.append(lambda:state._make_absent(new_entry))
 
690
 
 
691
        for from_rel in from_paths:
 
692
            # from_rel is 'pathinroot/foo/bar'
 
693
            from_rel_utf8 = from_rel.encode('utf8')
 
694
            from_dirname, from_tail = osutils.split(from_rel)
 
695
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
696
            from_entry = self._get_entry(path=from_rel)
 
697
            if from_entry == (None, None):
 
698
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
699
                    errors.NotVersionedError(path=str(from_rel)))
 
700
 
 
701
            from_id = from_entry[0][2]
 
702
            to_rel = pathjoin(to_dir, from_tail)
 
703
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
704
            item_to_entry = self._get_entry(path=to_rel)
 
705
            if item_to_entry != (None, None):
 
706
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
707
                    "Target is already versioned.")
 
708
 
 
709
            if from_rel == to_rel:
 
710
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
711
                    "Source and target are identical.")
 
712
 
 
713
            from_missing = not self.has_filename(from_rel)
 
714
            to_missing = not self.has_filename(to_rel)
 
715
            if after:
 
716
                move_file = False
 
717
            else:
 
718
                move_file = True
 
719
            if to_missing:
 
720
                if not move_file:
 
721
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
722
                        errors.NoSuchFile(path=to_rel,
 
723
                        extra="New file has not been created yet"))
 
724
                elif from_missing:
 
725
                    # neither path exists
 
726
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
727
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
728
            else:
 
729
                if from_missing: # implicitly just update our path mapping
 
730
                    move_file = False
 
731
                elif not after:
 
732
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
733
                        extra="(Use --after to update the Bazaar id)")
 
734
 
 
735
            rollbacks = []
 
736
            def rollback_rename():
 
737
                """A single rename has failed, roll it back."""
 
738
                # roll back everything, even if we encounter trouble doing one
 
739
                # of them.
 
740
                #
 
741
                # TODO: at least log the other exceptions rather than just
 
742
                # losing them mbp 20070307
 
743
                exc_info = None
 
744
                for rollback in reversed(rollbacks):
 
745
                    try:
 
746
                        rollback()
 
747
                    except Exception, e:
 
748
                        exc_info = sys.exc_info()
 
749
                if exc_info:
 
750
                    raise exc_info[0], exc_info[1], exc_info[2]
 
751
 
 
752
            # perform the disk move first - its the most likely failure point.
 
753
            if move_file:
 
754
                from_rel_abs = self.abspath(from_rel)
 
755
                to_rel_abs = self.abspath(to_rel)
 
756
                try:
 
757
                    osutils.rename(from_rel_abs, to_rel_abs)
 
758
                except OSError, e:
 
759
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
760
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
761
            try:
 
762
                # perform the rename in the inventory next if needed: its easy
 
763
                # to rollback
 
764
                if update_inventory:
 
765
                    # rename the entry
 
766
                    from_entry = inv[from_id]
 
767
                    current_parent = from_entry.parent_id
 
768
                    inv.rename(from_id, to_dir_id, from_tail)
 
769
                    rollbacks.append(
 
770
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
771
                # finally do the rename in the dirstate, which is a little
 
772
                # tricky to rollback, but least likely to need it.
 
773
                old_block_index, old_entry_index, dir_present, file_present = \
 
774
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
775
                old_block = state._dirblocks[old_block_index][1]
 
776
                old_entry = old_block[old_entry_index]
 
777
                from_key, old_entry_details = old_entry
 
778
                cur_details = old_entry_details[0]
 
779
                # remove the old row
 
780
                to_key = ((to_block[0],) + from_key[1:3])
 
781
                minikind = cur_details[0]
 
782
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
783
                         minikind=minikind,
 
784
                         executable=cur_details[3],
 
785
                         fingerprint=cur_details[1],
 
786
                         packed_stat=cur_details[4],
 
787
                         size=cur_details[2],
 
788
                         to_block=to_block,
 
789
                         to_key=to_key,
 
790
                         to_path_utf8=to_rel_utf8)
 
791
 
 
792
                if minikind == 'd':
 
793
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
794
                        """Recursively update all entries in this dirblock."""
 
795
                        assert from_dir != '', "renaming root not supported"
 
796
                        from_key = (from_dir, '')
 
797
                        from_block_idx, present = \
 
798
                            state._find_block_index_from_key(from_key)
 
799
                        if not present:
 
800
                            # This is the old record, if it isn't present, then
 
801
                            # there is theoretically nothing to update.
 
802
                            # (Unless it isn't present because of lazy loading,
 
803
                            # but we don't do that yet)
 
804
                            return
 
805
                        from_block = state._dirblocks[from_block_idx]
 
806
                        to_block_index, to_entry_index, _, _ = \
 
807
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
808
                        to_block_index = state._ensure_block(
 
809
                            to_block_index, to_entry_index, to_dir_utf8)
 
810
                        to_block = state._dirblocks[to_block_index]
 
811
 
 
812
                        # Grab a copy since move_one may update the list.
 
813
                        for entry in from_block[1][:]:
 
814
                            assert entry[0][0] == from_dir
 
815
                            cur_details = entry[1][0]
 
816
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
817
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
818
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
819
                            minikind = cur_details[0]
 
820
                            if minikind in 'ar':
 
821
                                # Deleted children of a renamed directory
 
822
                                # Do not need to be updated.
 
823
                                # Children that have been renamed out of this
 
824
                                # directory should also not be updated
 
825
                                continue
 
826
                            move_one(entry, from_path_utf8=from_path_utf8,
 
827
                                     minikind=minikind,
 
828
                                     executable=cur_details[3],
 
829
                                     fingerprint=cur_details[1],
 
830
                                     packed_stat=cur_details[4],
 
831
                                     size=cur_details[2],
 
832
                                     to_block=to_block,
 
833
                                     to_key=to_key,
 
834
                                     to_path_utf8=to_path_utf8)
 
835
                            if minikind == 'd':
 
836
                                # We need to move all the children of this
 
837
                                # entry
 
838
                                update_dirblock(from_path_utf8, to_key,
 
839
                                                to_path_utf8)
 
840
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
841
            except:
 
842
                rollback_rename()
 
843
                raise
 
844
            result.append((from_rel, to_rel))
 
845
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
846
            self._make_dirty(reset_inventory=False)
 
847
 
 
848
        return result
 
849
 
 
850
    def _must_be_locked(self):
 
851
        if not self._control_files._lock_count:
 
852
            raise errors.ObjectNotLocked(self)
 
853
 
 
854
    def _new_tree(self):
 
855
        """Initialize the state in this tree to be a new tree."""
 
856
        self._dirty = True
 
857
 
 
858
    @needs_read_lock
 
859
    def path2id(self, path):
 
860
        """Return the id for path in this tree."""
 
861
        path = path.strip('/')
 
862
        entry = self._get_entry(path=path)
 
863
        if entry == (None, None):
 
864
            return None
 
865
        return entry[0][2]
 
866
 
 
867
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
868
        """See Tree.paths2ids().
 
869
 
 
870
        This specialisation fast-paths the case where all the trees are in the
 
871
        dirstate.
 
872
        """
 
873
        if paths is None:
 
874
            return None
 
875
        parents = self.get_parent_ids()
 
876
        for tree in trees:
 
877
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
878
                parents):
 
879
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
880
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
881
        # -- make all paths utf8 --
 
882
        paths_utf8 = set()
 
883
        for path in paths:
 
884
            paths_utf8.add(path.encode('utf8'))
 
885
        paths = paths_utf8
 
886
        # -- paths is now a utf8 path set --
 
887
        # -- get the state object and prepare it.
 
888
        state = self.current_dirstate()
 
889
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
890
            and '' not in paths):
 
891
            paths2ids = self._paths2ids_using_bisect
 
892
        else:
 
893
            paths2ids = self._paths2ids_in_memory
 
894
        return paths2ids(paths, search_indexes,
 
895
                         require_versioned=require_versioned)
 
896
 
 
897
    def _paths2ids_in_memory(self, paths, search_indexes,
 
898
                             require_versioned=True):
 
899
        state = self.current_dirstate()
 
900
        state._read_dirblocks_if_needed()
 
901
        def _entries_for_path(path):
 
902
            """Return a list with all the entries that match path for all ids.
 
903
            """
 
904
            dirname, basename = os.path.split(path)
 
905
            key = (dirname, basename, '')
 
906
            block_index, present = state._find_block_index_from_key(key)
 
907
            if not present:
 
908
                # the block which should contain path is absent.
 
909
                return []
 
910
            result = []
 
911
            block = state._dirblocks[block_index][1]
 
912
            entry_index, _ = state._find_entry_index(key, block)
 
913
            # we may need to look at multiple entries at this path: walk while the paths match.
 
914
            while (entry_index < len(block) and
 
915
                block[entry_index][0][0:2] == key[0:2]):
 
916
                result.append(block[entry_index])
 
917
                entry_index += 1
 
918
            return result
 
919
        if require_versioned:
 
920
            # -- check all supplied paths are versioned in a search tree. --
 
921
            all_versioned = True
 
922
            for path in paths:
 
923
                path_entries = _entries_for_path(path)
 
924
                if not path_entries:
 
925
                    # this specified path is not present at all: error
 
926
                    all_versioned = False
 
927
                    break
 
928
                found_versioned = False
 
929
                # for each id at this path
 
930
                for entry in path_entries:
 
931
                    # for each tree.
 
932
                    for index in search_indexes:
 
933
                        if entry[1][index][0] != 'a': # absent
 
934
                            found_versioned = True
 
935
                            # all good: found a versioned cell
 
936
                            break
 
937
                if not found_versioned:
 
938
                    # none of the indexes was not 'absent' at all ids for this
 
939
                    # path.
 
940
                    all_versioned = False
 
941
                    break
 
942
            if not all_versioned:
 
943
                raise errors.PathsNotVersionedError(paths)
 
944
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
945
        search_paths = set()
 
946
        for path in paths:
 
947
            other_paths = paths.difference(set([path]))
 
948
            if not osutils.is_inside_any(other_paths, path):
 
949
                # this is a top level path, we must check it.
 
950
                search_paths.add(path)
 
951
        # sketch: 
 
952
        # for all search_indexs in each path at or under each element of
 
953
        # search_paths, if the detail is relocated: add the id, and add the
 
954
        # relocated path as one to search if its not searched already. If the
 
955
        # detail is not relocated, add the id.
 
956
        searched_paths = set()
 
957
        found_ids = set()
 
958
        def _process_entry(entry):
 
959
            """Look at search_indexes within entry.
 
960
 
 
961
            If a specific tree's details are relocated, add the relocation
 
962
            target to search_paths if not searched already. If it is absent, do
 
963
            nothing. Otherwise add the id to found_ids.
 
964
            """
 
965
            for index in search_indexes:
 
966
                if entry[1][index][0] == 'r': # relocated
 
967
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
968
                        search_paths.add(entry[1][index][1])
 
969
                elif entry[1][index][0] != 'a': # absent
 
970
                    found_ids.add(entry[0][2])
 
971
        while search_paths:
 
972
            current_root = search_paths.pop()
 
973
            searched_paths.add(current_root)
 
974
            # process the entries for this containing directory: the rest will be
 
975
            # found by their parents recursively.
 
976
            root_entries = _entries_for_path(current_root)
 
977
            if not root_entries:
 
978
                # this specified path is not present at all, skip it.
 
979
                continue
 
980
            for entry in root_entries:
 
981
                _process_entry(entry)
 
982
            initial_key = (current_root, '', '')
 
983
            block_index, _ = state._find_block_index_from_key(initial_key)
 
984
            while (block_index < len(state._dirblocks) and
 
985
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
986
                for entry in state._dirblocks[block_index][1]:
 
987
                    _process_entry(entry)
 
988
                block_index += 1
 
989
        return found_ids
 
990
 
 
991
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
992
                                require_versioned=True):
 
993
        state = self.current_dirstate()
 
994
        found_ids = set()
 
995
 
 
996
        split_paths = sorted(osutils.split(p) for p in paths)
 
997
        found = state._bisect_recursive(split_paths)
 
998
 
 
999
        if require_versioned:
 
1000
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
1001
            for dir_name in split_paths:
 
1002
                if dir_name not in found_dir_names:
 
1003
                    raise errors.PathsNotVersionedError(paths)
 
1004
 
 
1005
        for dir_name_id, trees_info in found.iteritems():
 
1006
            for index in search_indexes:
 
1007
                if trees_info[index][0] not in ('r', 'a'):
 
1008
                    found_ids.add(dir_name_id[2])
 
1009
        return found_ids
 
1010
 
 
1011
    def read_working_inventory(self):
 
1012
        """Read the working inventory.
 
1013
        
 
1014
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1015
        """
 
1016
        return self.inventory
 
1017
 
 
1018
    @needs_read_lock
 
1019
    def revision_tree(self, revision_id):
 
1020
        """See Tree.revision_tree.
 
1021
 
 
1022
        WorkingTree4 supplies revision_trees for any basis tree.
 
1023
        """
 
1024
        revision_id = osutils.safe_revision_id(revision_id)
 
1025
        dirstate = self.current_dirstate()
 
1026
        parent_ids = dirstate.get_parent_ids()
 
1027
        if revision_id not in parent_ids:
 
1028
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1029
        if revision_id in dirstate.get_ghosts():
 
1030
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1031
        return DirStateRevisionTree(dirstate, revision_id,
 
1032
            self.branch.repository)
 
1033
 
 
1034
    @needs_tree_write_lock
 
1035
    def set_last_revision(self, new_revision):
 
1036
        """Change the last revision in the working tree."""
 
1037
        new_revision = osutils.safe_revision_id(new_revision)
 
1038
        parents = self.get_parent_ids()
 
1039
        if new_revision in (NULL_REVISION, None):
 
1040
            assert len(parents) < 2, (
 
1041
                "setting the last parent to none with a pending merge is "
 
1042
                "unsupported.")
 
1043
            self.set_parent_ids([])
 
1044
        else:
 
1045
            self.set_parent_ids([new_revision] + parents[1:],
 
1046
                allow_leftmost_as_ghost=True)
 
1047
 
 
1048
    @needs_tree_write_lock
 
1049
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1050
        """Set the parent ids to revision_ids.
 
1051
        
 
1052
        See also set_parent_trees. This api will try to retrieve the tree data
 
1053
        for each element of revision_ids from the trees repository. If you have
 
1054
        tree data already available, it is more efficient to use
 
1055
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1056
        an easier API to use.
 
1057
 
 
1058
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1059
            working tree. Any of these may be ghosts.
 
1060
        """
 
1061
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
 
1062
        trees = []
 
1063
        for revision_id in revision_ids:
 
1064
            try:
 
1065
                revtree = self.branch.repository.revision_tree(revision_id)
 
1066
                # TODO: jam 20070213 KnitVersionedFile raises
 
1067
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1068
                #       given revision_id is not present. Should Repository be
 
1069
                #       catching it and re-raising NoSuchRevision?
 
1070
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1071
                revtree = None
 
1072
            trees.append((revision_id, revtree))
 
1073
        self.set_parent_trees(trees,
 
1074
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1075
 
 
1076
    @needs_tree_write_lock
 
1077
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1078
        """Set the parents of the working tree.
 
1079
 
 
1080
        :param parents_list: A list of (revision_id, tree) tuples.
 
1081
            If tree is None, then that element is treated as an unreachable
 
1082
            parent tree - i.e. a ghost.
 
1083
        """
 
1084
        dirstate = self.current_dirstate()
 
1085
        if len(parents_list) > 0:
 
1086
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1087
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1088
        real_trees = []
 
1089
        ghosts = []
 
1090
        # convert absent trees to the null tree, which we convert back to
 
1091
        # missing on access.
 
1092
        for rev_id, tree in parents_list:
 
1093
            rev_id = osutils.safe_revision_id(rev_id)
 
1094
            _mod_revision.check_not_reserved_id(rev_id)
 
1095
            if tree is not None:
 
1096
                real_trees.append((rev_id, tree))
 
1097
            else:
 
1098
                real_trees.append((rev_id,
 
1099
                    self.branch.repository.revision_tree(None)))
 
1100
                ghosts.append(rev_id)
 
1101
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1102
        self._make_dirty(reset_inventory=False)
 
1103
 
 
1104
    def _set_root_id(self, file_id):
 
1105
        """See WorkingTree.set_root_id."""
 
1106
        state = self.current_dirstate()
 
1107
        state.set_path_id('', file_id)
 
1108
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1109
            self._make_dirty(reset_inventory=True)
 
1110
 
 
1111
    @needs_read_lock
 
1112
    def supports_tree_reference(self):
 
1113
        return self._repo_supports_tree_reference
 
1114
 
 
1115
    def unlock(self):
 
1116
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1117
        if self._control_files._lock_count == 1:
 
1118
            # eventually we should do signature checking during read locks for
 
1119
            # dirstate updates.
 
1120
            if self._control_files._lock_mode == 'w':
 
1121
                if self._dirty:
 
1122
                    self.flush()
 
1123
            if self._dirstate is not None:
 
1124
                # This is a no-op if there are no modifications.
 
1125
                self._dirstate.save()
 
1126
                self._dirstate.unlock()
 
1127
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1128
            #       point. Instead, it could check if the header has been
 
1129
            #       modified when it is locked, and if not, it can hang on to
 
1130
            #       the data it has in memory.
 
1131
            self._dirstate = None
 
1132
            self._inventory = None
 
1133
        # reverse order of locking.
 
1134
        try:
 
1135
            return self._control_files.unlock()
 
1136
        finally:
 
1137
            self.branch.unlock()
 
1138
 
 
1139
    @needs_tree_write_lock
 
1140
    def unversion(self, file_ids):
 
1141
        """Remove the file ids in file_ids from the current versioned set.
 
1142
 
 
1143
        When a file_id is unversioned, all of its children are automatically
 
1144
        unversioned.
 
1145
 
 
1146
        :param file_ids: The file ids to stop versioning.
 
1147
        :raises: NoSuchId if any fileid is not currently versioned.
 
1148
        """
 
1149
        if not file_ids:
 
1150
            return
 
1151
        state = self.current_dirstate()
 
1152
        state._read_dirblocks_if_needed()
 
1153
        ids_to_unversion = set()
 
1154
        for file_id in file_ids:
 
1155
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1156
        paths_to_unversion = set()
 
1157
        # sketch:
 
1158
        # check if the root is to be unversioned, if so, assert for now.
 
1159
        # walk the state marking unversioned things as absent.
 
1160
        # if there are any un-unversioned ids at the end, raise
 
1161
        for key, details in state._dirblocks[0][1]:
 
1162
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1163
                key[2] in ids_to_unversion):
 
1164
                # I haven't written the code to unversion / yet - it should be
 
1165
                # supported.
 
1166
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1167
        block_index = 0
 
1168
        while block_index < len(state._dirblocks):
 
1169
            # process one directory at a time.
 
1170
            block = state._dirblocks[block_index]
 
1171
            # first check: is the path one to remove - it or its children
 
1172
            delete_block = False
 
1173
            for path in paths_to_unversion:
 
1174
                if (block[0].startswith(path) and
 
1175
                    (len(block[0]) == len(path) or
 
1176
                     block[0][len(path)] == '/')):
 
1177
                    # this entire block should be deleted - its the block for a
 
1178
                    # path to unversion; or the child of one
 
1179
                    delete_block = True
 
1180
                    break
 
1181
            # TODO: trim paths_to_unversion as we pass by paths
 
1182
            if delete_block:
 
1183
                # this block is to be deleted: process it.
 
1184
                # TODO: we can special case the no-parents case and
 
1185
                # just forget the whole block.
 
1186
                entry_index = 0
 
1187
                while entry_index < len(block[1]):
 
1188
                    # Mark this file id as having been removed
 
1189
                    entry = block[1][entry_index]
 
1190
                    ids_to_unversion.discard(entry[0][2])
 
1191
                    if (entry[1][0][0] == 'a'
 
1192
                        or not state._make_absent(entry)):
 
1193
                        entry_index += 1
 
1194
                # go to the next block. (At the moment we dont delete empty
 
1195
                # dirblocks)
 
1196
                block_index += 1
 
1197
                continue
 
1198
            entry_index = 0
 
1199
            while entry_index < len(block[1]):
 
1200
                entry = block[1][entry_index]
 
1201
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1202
                    # ^ some parent row.
 
1203
                    entry[0][2] not in ids_to_unversion):
 
1204
                    # ^ not an id to unversion
 
1205
                    entry_index += 1
 
1206
                    continue
 
1207
                if entry[1][0][0] == 'd':
 
1208
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1209
                if not state._make_absent(entry):
 
1210
                    entry_index += 1
 
1211
                # we have unversioned this id
 
1212
                ids_to_unversion.remove(entry[0][2])
 
1213
            block_index += 1
 
1214
        if ids_to_unversion:
 
1215
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1216
        self._make_dirty(reset_inventory=False)
 
1217
        # have to change the legacy inventory too.
 
1218
        if self._inventory is not None:
 
1219
            for file_id in file_ids:
 
1220
                self._inventory.remove_recursive_id(file_id)
 
1221
 
 
1222
    @needs_read_lock
 
1223
    def _validate(self):
 
1224
        self._dirstate._validate()
 
1225
 
 
1226
    @needs_tree_write_lock
 
1227
    def _write_inventory(self, inv):
 
1228
        """Write inventory as the current inventory."""
 
1229
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1230
        self.current_dirstate().set_state_from_inventory(inv)
 
1231
        self._make_dirty(reset_inventory=False)
 
1232
        if self._inventory is not None:
 
1233
            self._inventory = inv
 
1234
        self.flush()
 
1235
 
 
1236
 
 
1237
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1238
    """The first consolidated dirstate working tree format.
 
1239
 
 
1240
    This format:
 
1241
        - exists within a metadir controlling .bzr
 
1242
        - includes an explicit version marker for the workingtree control
 
1243
          files, separate from the BzrDir format
 
1244
        - modifies the hash cache format
 
1245
        - is new in bzr 0.15
 
1246
        - uses a LockDir to guard access to it.
 
1247
    """
 
1248
 
 
1249
    upgrade_recommended = False
 
1250
 
 
1251
    def get_format_string(self):
 
1252
        """See WorkingTreeFormat.get_format_string()."""
 
1253
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1254
 
 
1255
    def get_format_description(self):
 
1256
        """See WorkingTreeFormat.get_format_description()."""
 
1257
        return "Working tree format 4"
 
1258
 
 
1259
    def initialize(self, a_bzrdir, revision_id=None):
 
1260
        """See WorkingTreeFormat.initialize().
 
1261
 
 
1262
        :param revision_id: allows creating a working tree at a different
 
1263
        revision than the branch is at.
 
1264
 
 
1265
        These trees get an initial random root id, if their repository supports
 
1266
        rich root data, TREE_ROOT otherwise.
 
1267
        """
 
1268
        revision_id = osutils.safe_revision_id(revision_id)
 
1269
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1270
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1271
        transport = a_bzrdir.get_workingtree_transport(self)
 
1272
        control_files = self._open_control_files(a_bzrdir)
 
1273
        control_files.create_lock()
 
1274
        control_files.lock_write()
 
1275
        control_files.put_utf8('format', self.get_format_string())
 
1276
        branch = a_bzrdir.open_branch()
 
1277
        if revision_id is None:
 
1278
            revision_id = branch.last_revision()
 
1279
        local_path = transport.local_abspath('dirstate')
 
1280
        # write out new dirstate (must exist when we create the tree)
 
1281
        state = dirstate.DirState.initialize(local_path)
 
1282
        state.unlock()
 
1283
        del state
 
1284
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1285
                         branch,
 
1286
                         _format=self,
 
1287
                         _bzrdir=a_bzrdir,
 
1288
                         _control_files=control_files)
 
1289
        wt._new_tree()
 
1290
        wt.lock_tree_write()
 
1291
        try:
 
1292
            if revision_id in (None, NULL_REVISION):
 
1293
                if branch.repository.supports_rich_root():
 
1294
                    wt._set_root_id(generate_ids.gen_root_id())
 
1295
                else:
 
1296
                    wt._set_root_id(ROOT_ID)
 
1297
                wt.flush()
 
1298
            wt.set_last_revision(revision_id)
 
1299
            wt.flush()
 
1300
            basis = wt.basis_tree()
 
1301
            basis.lock_read()
 
1302
            # if the basis has a root id we have to use that; otherwise we use
 
1303
            # a new random one
 
1304
            basis_root_id = basis.get_root_id()
 
1305
            if basis_root_id is not None:
 
1306
                wt._set_root_id(basis_root_id)
 
1307
                wt.flush()
 
1308
            transform.build_tree(basis, wt)
 
1309
            basis.unlock()
 
1310
        finally:
 
1311
            control_files.unlock()
 
1312
            wt.unlock()
 
1313
        return wt
 
1314
 
 
1315
    def _open(self, a_bzrdir, control_files):
 
1316
        """Open the tree itself.
 
1317
 
 
1318
        :param a_bzrdir: the dir for the tree.
 
1319
        :param control_files: the control files for the tree.
 
1320
        """
 
1321
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1322
                           branch=a_bzrdir.open_branch(),
 
1323
                           _format=self,
 
1324
                           _bzrdir=a_bzrdir,
 
1325
                           _control_files=control_files)
 
1326
 
 
1327
    def __get_matchingbzrdir(self):
 
1328
        # please test against something that will let us do tree references
 
1329
        return bzrdir.format_registry.make_bzrdir(
 
1330
            'dirstate-with-subtree')
 
1331
 
 
1332
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1333
 
 
1334
 
 
1335
class DirStateRevisionTree(Tree):
 
1336
    """A revision tree pulling the inventory from a dirstate."""
 
1337
 
 
1338
    def __init__(self, dirstate, revision_id, repository):
 
1339
        self._dirstate = dirstate
 
1340
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1341
        self._repository = repository
 
1342
        self._inventory = None
 
1343
        self._locked = 0
 
1344
        self._dirstate_locked = False
 
1345
 
 
1346
    def __repr__(self):
 
1347
        return "<%s of %s in %s>" % \
 
1348
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1349
 
 
1350
    def annotate_iter(self, file_id,
 
1351
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1352
        """See Tree.annotate_iter"""
 
1353
        w = self._get_weave(file_id)
 
1354
        return w.annotate_iter(self.inventory[file_id].revision)
 
1355
 
 
1356
    def _get_ancestors(self, default_revision):
 
1357
        return set(self._repository.get_ancestry(self._revision_id,
 
1358
                                                 topo_sorted=False))
 
1359
    def _comparison_data(self, entry, path):
 
1360
        """See Tree._comparison_data."""
 
1361
        if entry is None:
 
1362
            return None, False, None
 
1363
        # trust the entry as RevisionTree does, but this may not be
 
1364
        # sensible: the entry might not have come from us?
 
1365
        return entry.kind, entry.executable, None
 
1366
 
 
1367
    def _file_size(self, entry, stat_value):
 
1368
        return entry.text_size
 
1369
 
 
1370
    def filter_unversioned_files(self, paths):
 
1371
        """Filter out paths that are not versioned.
 
1372
 
 
1373
        :return: set of paths.
 
1374
        """
 
1375
        pred = self.has_filename
 
1376
        return set((p for p in paths if not pred(p)))
 
1377
 
 
1378
    def get_root_id(self):
 
1379
        return self.path2id('')
 
1380
 
 
1381
    def _get_parent_index(self):
 
1382
        """Return the index in the dirstate referenced by this tree."""
 
1383
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1384
 
 
1385
    def _get_entry(self, file_id=None, path=None):
 
1386
        """Get the dirstate row for file_id or path.
 
1387
 
 
1388
        If either file_id or path is supplied, it is used as the key to lookup.
 
1389
        If both are supplied, the fastest lookup is used, and an error is
 
1390
        raised if they do not both point at the same row.
 
1391
        
 
1392
        :param file_id: An optional unicode file_id to be looked up.
 
1393
        :param path: An optional unicode path to be looked up.
 
1394
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1395
        """
 
1396
        if file_id is None and path is None:
 
1397
            raise errors.BzrError('must supply file_id or path')
 
1398
        file_id = osutils.safe_file_id(file_id)
 
1399
        if path is not None:
 
1400
            path = path.encode('utf8')
 
1401
        parent_index = self._get_parent_index()
 
1402
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1403
 
 
1404
    def _generate_inventory(self):
 
1405
        """Create and set self.inventory from the dirstate object.
 
1406
 
 
1407
        (So this is only called the first time the inventory is requested for
 
1408
        this tree; it then remains in memory until it's out of date.)
 
1409
 
 
1410
        This is relatively expensive: we have to walk the entire dirstate.
 
1411
        """
 
1412
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1413
            'dirstate revision tree'
 
1414
        # separate call for profiling - makes it clear where the costs are.
 
1415
        self._dirstate._read_dirblocks_if_needed()
 
1416
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1417
            'parent %s has disappeared from %s' % (
 
1418
            self._revision_id, self._dirstate.get_parent_ids())
 
1419
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1420
        # This is identical now to the WorkingTree _generate_inventory except
 
1421
        # for the tree index use.
 
1422
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1423
        current_id = root_key[2]
 
1424
        assert current_entry[parent_index][0] == 'd'
 
1425
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1426
        inv.root.revision = current_entry[parent_index][4]
 
1427
        # Turn some things into local variables
 
1428
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1429
        factory = entry_factory
 
1430
        utf8_decode = cache_utf8._utf8_decode
 
1431
        inv_byid = inv._byid
 
1432
        # we could do this straight out of the dirstate; it might be fast
 
1433
        # and should be profiled - RBC 20070216
 
1434
        parent_ies = {'' : inv.root}
 
1435
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1436
            dirname = block[0]
 
1437
            try:
 
1438
                parent_ie = parent_ies[dirname]
 
1439
            except KeyError:
 
1440
                # all the paths in this block are not versioned in this tree
 
1441
                continue
 
1442
            for key, entry in block[1]:
 
1443
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1444
                if minikind in ('a', 'r'): # absent, relocated
 
1445
                    # not this tree
 
1446
                    continue
 
1447
                name = key[1]
 
1448
                name_unicode = utf8_decode(name)[0]
 
1449
                file_id = key[2]
 
1450
                kind = minikind_to_kind[minikind]
 
1451
                inv_entry = factory[kind](file_id, name_unicode,
 
1452
                                          parent_ie.file_id)
 
1453
                inv_entry.revision = revid
 
1454
                if kind == 'file':
 
1455
                    inv_entry.executable = executable
 
1456
                    inv_entry.text_size = size
 
1457
                    inv_entry.text_sha1 = fingerprint
 
1458
                elif kind == 'directory':
 
1459
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1460
                elif kind == 'symlink':
 
1461
                    inv_entry.executable = False
 
1462
                    inv_entry.text_size = None
 
1463
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1464
                elif kind == 'tree-reference':
 
1465
                    inv_entry.reference_revision = fingerprint or None
 
1466
                else:
 
1467
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1468
                            % entry)
 
1469
                # These checks cost us around 40ms on a 55k entry tree
 
1470
                assert file_id not in inv_byid
 
1471
                assert name_unicode not in parent_ie.children
 
1472
                inv_byid[file_id] = inv_entry
 
1473
                parent_ie.children[name_unicode] = inv_entry
 
1474
        self._inventory = inv
 
1475
 
 
1476
    def get_file_mtime(self, file_id, path=None):
 
1477
        """Return the modification time for this record.
 
1478
 
 
1479
        We return the timestamp of the last-changed revision.
 
1480
        """
 
1481
        # Make sure the file exists
 
1482
        entry = self._get_entry(file_id, path=path)
 
1483
        if entry == (None, None): # do we raise?
 
1484
            return None
 
1485
        parent_index = self._get_parent_index()
 
1486
        last_changed_revision = entry[1][parent_index][4]
 
1487
        return self._repository.get_revision(last_changed_revision).timestamp
 
1488
 
 
1489
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1490
        entry = self._get_entry(file_id=file_id, path=path)
 
1491
        parent_index = self._get_parent_index()
 
1492
        parent_details = entry[1][parent_index]
 
1493
        if parent_details[0] == 'f':
 
1494
            return parent_details[1]
 
1495
        return None
 
1496
 
 
1497
    @symbol_versioning.deprecated_method(symbol_versioning.zero_nineteen)
 
1498
    def get_weave(self, file_id):
 
1499
        return self._get_weave(file_id)
 
1500
 
 
1501
    def _get_weave(self, file_id):
 
1502
        return self._repository.weave_store.get_weave(file_id,
 
1503
                self._repository.get_transaction())
 
1504
 
 
1505
    def get_file(self, file_id):
 
1506
        return StringIO(self.get_file_text(file_id))
 
1507
 
 
1508
    def get_file_lines(self, file_id):
 
1509
        ie = self.inventory[file_id]
 
1510
        return self._get_weave(file_id).get_lines(ie.revision)
 
1511
 
 
1512
    def get_file_size(self, file_id):
 
1513
        return self.inventory[file_id].text_size
 
1514
 
 
1515
    def get_file_text(self, file_id):
 
1516
        return ''.join(self.get_file_lines(file_id))
 
1517
 
 
1518
    def get_reference_revision(self, file_id, path=None):
 
1519
        return self.inventory[file_id].reference_revision
 
1520
 
 
1521
    def get_symlink_target(self, file_id):
 
1522
        entry = self._get_entry(file_id=file_id)
 
1523
        parent_index = self._get_parent_index()
 
1524
        if entry[1][parent_index][0] != 'l':
 
1525
            return None
 
1526
        else:
 
1527
            # At present, none of the tree implementations supports non-ascii
 
1528
            # symlink targets. So we will just assume that the dirstate path is
 
1529
            # correct.
 
1530
            return entry[1][parent_index][1]
 
1531
 
 
1532
    def get_revision_id(self):
 
1533
        """Return the revision id for this tree."""
 
1534
        return self._revision_id
 
1535
 
 
1536
    def _get_inventory(self):
 
1537
        if self._inventory is not None:
 
1538
            return self._inventory
 
1539
        self._must_be_locked()
 
1540
        self._generate_inventory()
 
1541
        return self._inventory
 
1542
 
 
1543
    inventory = property(_get_inventory,
 
1544
                         doc="Inventory of this Tree")
 
1545
 
 
1546
    def get_parent_ids(self):
 
1547
        """The parents of a tree in the dirstate are not cached."""
 
1548
        return self._repository.get_revision(self._revision_id).parent_ids
 
1549
 
 
1550
    def has_filename(self, filename):
 
1551
        return bool(self.path2id(filename))
 
1552
 
 
1553
    def kind(self, file_id):
 
1554
        return self.inventory[file_id].kind
 
1555
 
 
1556
    def is_executable(self, file_id, path=None):
 
1557
        ie = self.inventory[file_id]
 
1558
        if ie.kind != "file":
 
1559
            return None
 
1560
        return ie.executable
 
1561
 
 
1562
    def list_files(self, include_root=False):
 
1563
        # We use a standard implementation, because DirStateRevisionTree is
 
1564
        # dealing with one of the parents of the current state
 
1565
        inv = self._get_inventory()
 
1566
        entries = inv.iter_entries()
 
1567
        if self.inventory.root is not None and not include_root:
 
1568
            entries.next()
 
1569
        for path, entry in entries:
 
1570
            yield path, 'V', entry.kind, entry.file_id, entry
 
1571
 
 
1572
    def lock_read(self):
 
1573
        """Lock the tree for a set of operations."""
 
1574
        if not self._locked:
 
1575
            self._repository.lock_read()
 
1576
            if self._dirstate._lock_token is None:
 
1577
                self._dirstate.lock_read()
 
1578
                self._dirstate_locked = True
 
1579
        self._locked += 1
 
1580
 
 
1581
    def _must_be_locked(self):
 
1582
        if not self._locked:
 
1583
            raise errors.ObjectNotLocked(self)
 
1584
 
 
1585
    @needs_read_lock
 
1586
    def path2id(self, path):
 
1587
        """Return the id for path in this tree."""
 
1588
        # lookup by path: faster than splitting and walking the ivnentory.
 
1589
        entry = self._get_entry(path=path)
 
1590
        if entry == (None, None):
 
1591
            return None
 
1592
        return entry[0][2]
 
1593
 
 
1594
    def unlock(self):
 
1595
        """Unlock, freeing any cache memory used during the lock."""
 
1596
        # outside of a lock, the inventory is suspect: release it.
 
1597
        self._locked -=1
 
1598
        if not self._locked:
 
1599
            self._inventory = None
 
1600
            self._locked = 0
 
1601
            if self._dirstate_locked:
 
1602
                self._dirstate.unlock()
 
1603
                self._dirstate_locked = False
 
1604
            self._repository.unlock()
 
1605
 
 
1606
    def walkdirs(self, prefix=""):
 
1607
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1608
        # implementation based on an inventory.  
 
1609
        # This should be cleaned up to use the much faster Dirstate code
 
1610
        # So for now, we just build up the parent inventory, and extract
 
1611
        # it the same way RevisionTree does.
 
1612
        _directory = 'directory'
 
1613
        inv = self._get_inventory()
 
1614
        top_id = inv.path2id(prefix)
 
1615
        if top_id is None:
 
1616
            pending = []
 
1617
        else:
 
1618
            pending = [(prefix, top_id)]
 
1619
        while pending:
 
1620
            dirblock = []
 
1621
            relpath, file_id = pending.pop()
 
1622
            # 0 - relpath, 1- file-id
 
1623
            if relpath:
 
1624
                relroot = relpath + '/'
 
1625
            else:
 
1626
                relroot = ""
 
1627
            # FIXME: stash the node in pending
 
1628
            entry = inv[file_id]
 
1629
            for name, child in entry.sorted_children():
 
1630
                toppath = relroot + name
 
1631
                dirblock.append((toppath, name, child.kind, None,
 
1632
                    child.file_id, child.kind
 
1633
                    ))
 
1634
            yield (relpath, entry.file_id), dirblock
 
1635
            # push the user specified dirs from dirblock
 
1636
            for dir in reversed(dirblock):
 
1637
                if dir[2] == _directory:
 
1638
                    pending.append((dir[0], dir[4]))
 
1639
 
 
1640
 
 
1641
class InterDirStateTree(InterTree):
 
1642
    """Fast path optimiser for changes_from with dirstate trees.
 
1643
    
 
1644
    This is used only when both trees are in the dirstate working file, and 
 
1645
    the source is any parent within the dirstate, and the destination is 
 
1646
    the current working tree of the same dirstate.
 
1647
    """
 
1648
    # this could be generalized to allow comparisons between any trees in the
 
1649
    # dirstate, and possibly between trees stored in different dirstates.
 
1650
 
 
1651
    def __init__(self, source, target):
 
1652
        super(InterDirStateTree, self).__init__(source, target)
 
1653
        if not InterDirStateTree.is_compatible(source, target):
 
1654
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1655
 
 
1656
    @staticmethod
 
1657
    def make_source_parent_tree(source, target):
 
1658
        """Change the source tree into a parent of the target."""
 
1659
        revid = source.commit('record tree')
 
1660
        target.branch.repository.fetch(source.branch.repository, revid)
 
1661
        target.set_parent_ids([revid])
 
1662
        return target.basis_tree(), target
 
1663
 
 
1664
    _matching_from_tree_format = WorkingTreeFormat4()
 
1665
    _matching_to_tree_format = WorkingTreeFormat4()
 
1666
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1667
 
 
1668
    def _iter_changes(self, include_unchanged=False,
 
1669
                      specific_files=None, pb=None, extra_trees=[],
 
1670
                      require_versioned=True, want_unversioned=False):
 
1671
        """Return the changes from source to target.
 
1672
 
 
1673
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
1674
            for details.
 
1675
        :param specific_files: An optional list of file paths to restrict the
 
1676
            comparison to. When mapping filenames to ids, all matches in all
 
1677
            trees (including optional extra_trees) are used, and all children of
 
1678
            matched directories are included.
 
1679
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1680
            unchanged entries in the result.
 
1681
        :param extra_trees: An optional list of additional trees to use when
 
1682
            mapping the contents of specific_files (paths) to file_ids.
 
1683
        :param require_versioned: If True, all files in specific_files must be
 
1684
            versioned in one of source, target, extra_trees or
 
1685
            PathsNotVersionedError is raised.
 
1686
        :param want_unversioned: Should unversioned files be returned in the
 
1687
            output. An unversioned file is defined as one with (False, False)
 
1688
            for the versioned pair.
 
1689
        """
 
1690
        utf8_decode = cache_utf8._utf8_decode
 
1691
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1692
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1693
        # NB: show_status depends on being able to pass in non-versioned files
 
1694
        # and report them as unknown
 
1695
        # TODO: handle extra trees in the dirstate.
 
1696
        # TODO: handle comparisons as an empty tree as a different special
 
1697
        # case? mbp 20070226
 
1698
        if extra_trees or (self.source._revision_id == NULL_REVISION):
 
1699
            # we can't fast-path these cases (yet)
 
1700
            for f in super(InterDirStateTree, self)._iter_changes(
 
1701
                include_unchanged, specific_files, pb, extra_trees,
 
1702
                require_versioned, want_unversioned=want_unversioned):
 
1703
                yield f
 
1704
            return
 
1705
        parent_ids = self.target.get_parent_ids()
 
1706
        assert (self.source._revision_id in parent_ids), \
 
1707
                "revision {%s} is not stored in {%s}, but %s " \
 
1708
                "can only be used for trees stored in the dirstate" \
 
1709
                % (self.source._revision_id, self.target, self._iter_changes)
 
1710
        target_index = 0
 
1711
        if self.source._revision_id == NULL_REVISION:
 
1712
            source_index = None
 
1713
            indices = (target_index,)
 
1714
        else:
 
1715
            assert (self.source._revision_id in parent_ids), \
 
1716
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1717
                self.source._revision_id, parent_ids)
 
1718
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1719
            indices = (source_index,target_index)
 
1720
        # -- make all specific_files utf8 --
 
1721
        if specific_files:
 
1722
            specific_files_utf8 = set()
 
1723
            for path in specific_files:
 
1724
                specific_files_utf8.add(path.encode('utf8'))
 
1725
            specific_files = specific_files_utf8
 
1726
        else:
 
1727
            specific_files = set([''])
 
1728
        # -- specific_files is now a utf8 path set --
 
1729
        # -- get the state object and prepare it.
 
1730
        state = self.target.current_dirstate()
 
1731
        state._read_dirblocks_if_needed()
 
1732
        def _entries_for_path(path):
 
1733
            """Return a list with all the entries that match path for all ids.
 
1734
            """
 
1735
            dirname, basename = os.path.split(path)
 
1736
            key = (dirname, basename, '')
 
1737
            block_index, present = state._find_block_index_from_key(key)
 
1738
            if not present:
 
1739
                # the block which should contain path is absent.
 
1740
                return []
 
1741
            result = []
 
1742
            block = state._dirblocks[block_index][1]
 
1743
            entry_index, _ = state._find_entry_index(key, block)
 
1744
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1745
            while (entry_index < len(block) and
 
1746
                block[entry_index][0][0:2] == key[0:2]):
 
1747
                result.append(block[entry_index])
 
1748
                entry_index += 1
 
1749
            return result
 
1750
        if require_versioned:
 
1751
            # -- check all supplied paths are versioned in a search tree. --
 
1752
            all_versioned = True
 
1753
            for path in specific_files:
 
1754
                path_entries = _entries_for_path(path)
 
1755
                if not path_entries:
 
1756
                    # this specified path is not present at all: error
 
1757
                    all_versioned = False
 
1758
                    break
 
1759
                found_versioned = False
 
1760
                # for each id at this path
 
1761
                for entry in path_entries:
 
1762
                    # for each tree.
 
1763
                    for index in indices:
 
1764
                        if entry[1][index][0] != 'a': # absent
 
1765
                            found_versioned = True
 
1766
                            # all good: found a versioned cell
 
1767
                            break
 
1768
                if not found_versioned:
 
1769
                    # none of the indexes was not 'absent' at all ids for this
 
1770
                    # path.
 
1771
                    all_versioned = False
 
1772
                    break
 
1773
            if not all_versioned:
 
1774
                raise errors.PathsNotVersionedError(specific_files)
 
1775
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1776
        search_specific_files = set()
 
1777
        for path in specific_files:
 
1778
            other_specific_files = specific_files.difference(set([path]))
 
1779
            if not osutils.is_inside_any(other_specific_files, path):
 
1780
                # this is a top level path, we must check it.
 
1781
                search_specific_files.add(path)
 
1782
        # sketch: 
 
1783
        # compare source_index and target_index at or under each element of search_specific_files.
 
1784
        # follow the following comparison table. Note that we only want to do diff operations when
 
1785
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1786
        # for the target.
 
1787
        # cases:
 
1788
        # 
 
1789
        # Source | Target | disk | action
 
1790
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1791
        #        |        |      | diff check on source-target
 
1792
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1793
        #        |        |      | ???
 
1794
        #   r    |  a     |      | add source to search
 
1795
        #   r    |  a     |  a   | 
 
1796
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1797
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1798
        #   a    | fdlt   |      | add new id
 
1799
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1800
        #   a    |  a     |      | not present in either tree, skip
 
1801
        #   a    |  a     |  a   | not present in any tree, skip
 
1802
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1803
        #        |        |      | may not be selected by the users list of paths.
 
1804
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1805
        #        |        |      | may not be selected by the users list of paths.
 
1806
        #  fdlt  | fdlt   |      | content in both: diff them
 
1807
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1808
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1809
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1810
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1811
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1812
        #        |        |      | this id at the other path.
 
1813
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1814
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1815
        #        |        |      | this id at the other path.
 
1816
 
 
1817
        # for all search_indexs in each path at or under each element of
 
1818
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1819
        # relocated path as one to search if its not searched already. If the
 
1820
        # detail is not relocated, add the id.
 
1821
        searched_specific_files = set()
 
1822
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1823
        # Using a list so that we can access the values and change them in
 
1824
        # nested scope. Each one is [path, file_id, entry]
 
1825
        last_source_parent = [None, None]
 
1826
        last_target_parent = [None, None]
 
1827
 
 
1828
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1829
 
 
1830
        # Just a sentry, so that _process_entry can say that this
 
1831
        # record is handled, but isn't interesting to process (unchanged)
 
1832
        uninteresting = object()
 
1833
 
 
1834
 
 
1835
        old_dirname_to_file_id = {}
 
1836
        new_dirname_to_file_id = {}
 
1837
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1838
        #       keeping a cache of directories that we have seen.
 
1839
 
 
1840
        def _process_entry(entry, path_info):
 
1841
            """Compare an entry and real disk to generate delta information.
 
1842
 
 
1843
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1844
                the path of entry. If None, then the path is considered absent.
 
1845
                (Perhaps we should pass in a concrete entry for this ?)
 
1846
                Basename is returned as a utf8 string because we expect this
 
1847
                tuple will be ignored, and don't want to take the time to
 
1848
                decode.
 
1849
            :return: None if these don't match
 
1850
                     A tuple of information about the change, or
 
1851
                     the object 'uninteresting' if these match, but are
 
1852
                     basically identical.
 
1853
            """
 
1854
            if source_index is None:
 
1855
                source_details = NULL_PARENT_DETAILS
 
1856
            else:
 
1857
                source_details = entry[1][source_index]
 
1858
            target_details = entry[1][target_index]
 
1859
            target_minikind = target_details[0]
 
1860
            if path_info is not None and target_minikind in 'fdlt':
 
1861
                assert target_index == 0
 
1862
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1863
                                                  stat_value=path_info[3])
 
1864
                # The entry may have been modified by update_entry
 
1865
                target_details = entry[1][target_index]
 
1866
                target_minikind = target_details[0]
 
1867
            else:
 
1868
                link_or_sha1 = None
 
1869
            file_id = entry[0][2]
 
1870
            source_minikind = source_details[0]
 
1871
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1872
                # claimed content in both: diff
 
1873
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1874
                #        |        |      | diff check on source-target
 
1875
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1876
                #        |        |      | ???
 
1877
                if source_minikind in 'r':
 
1878
                    # add the source to the search path to find any children it
 
1879
                    # has.  TODO ? : only add if it is a container ?
 
1880
                    if not osutils.is_inside_any(searched_specific_files,
 
1881
                                                 source_details[1]):
 
1882
                        search_specific_files.add(source_details[1])
 
1883
                    # generate the old path; this is needed for stating later
 
1884
                    # as well.
 
1885
                    old_path = source_details[1]
 
1886
                    old_dirname, old_basename = os.path.split(old_path)
 
1887
                    path = pathjoin(entry[0][0], entry[0][1])
 
1888
                    old_entry = state._get_entry(source_index,
 
1889
                                                 path_utf8=old_path)
 
1890
                    # update the source details variable to be the real
 
1891
                    # location.
 
1892
                    source_details = old_entry[1][source_index]
 
1893
                    source_minikind = source_details[0]
 
1894
                else:
 
1895
                    old_dirname = entry[0][0]
 
1896
                    old_basename = entry[0][1]
 
1897
                    old_path = path = None
 
1898
                if path_info is None:
 
1899
                    # the file is missing on disk, show as removed.
 
1900
                    content_change = True
 
1901
                    target_kind = None
 
1902
                    target_exec = False
 
1903
                else:
 
1904
                    # source and target are both versioned and disk file is present.
 
1905
                    target_kind = path_info[2]
 
1906
                    if target_kind == 'directory':
 
1907
                        if path is None:
 
1908
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1909
                        new_dirname_to_file_id[path] = file_id
 
1910
                        if source_minikind != 'd':
 
1911
                            content_change = True
 
1912
                        else:
 
1913
                            # directories have no fingerprint
 
1914
                            content_change = False
 
1915
                        target_exec = False
 
1916
                    elif target_kind == 'file':
 
1917
                        if source_minikind != 'f':
 
1918
                            content_change = True
 
1919
                        else:
 
1920
                            # We could check the size, but we already have the
 
1921
                            # sha1 hash.
 
1922
                            content_change = (link_or_sha1 != source_details[1])
 
1923
                        # Target details is updated at update_entry time
 
1924
                        if use_filesystem_for_exec:
 
1925
                            # We don't need S_ISREG here, because we are sure
 
1926
                            # we are dealing with a file.
 
1927
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1928
                        else:
 
1929
                            target_exec = target_details[3]
 
1930
                    elif target_kind == 'symlink':
 
1931
                        if source_minikind != 'l':
 
1932
                            content_change = True
 
1933
                        else:
 
1934
                            content_change = (link_or_sha1 != source_details[1])
 
1935
                        target_exec = False
 
1936
                    elif target_kind == 'tree-reference':
 
1937
                        if source_minikind != 't':
 
1938
                            content_change = True
 
1939
                        else:
 
1940
                            content_change = False
 
1941
                        target_exec = False
 
1942
                    else:
 
1943
                        raise Exception, "unknown kind %s" % path_info[2]
 
1944
                if source_minikind == 'd':
 
1945
                    if path is None:
 
1946
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1947
                    old_dirname_to_file_id[old_path] = file_id
 
1948
                # parent id is the entry for the path in the target tree
 
1949
                if old_dirname == last_source_parent[0]:
 
1950
                    source_parent_id = last_source_parent[1]
 
1951
                else:
 
1952
                    try:
 
1953
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1954
                    except KeyError:
 
1955
                        source_parent_entry = state._get_entry(source_index,
 
1956
                                                               path_utf8=old_dirname)
 
1957
                        source_parent_id = source_parent_entry[0][2]
 
1958
                    if source_parent_id == entry[0][2]:
 
1959
                        # This is the root, so the parent is None
 
1960
                        source_parent_id = None
 
1961
                    else:
 
1962
                        last_source_parent[0] = old_dirname
 
1963
                        last_source_parent[1] = source_parent_id
 
1964
                new_dirname = entry[0][0]
 
1965
                if new_dirname == last_target_parent[0]:
 
1966
                    target_parent_id = last_target_parent[1]
 
1967
                else:
 
1968
                    try:
 
1969
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1970
                    except KeyError:
 
1971
                        # TODO: We don't always need to do the lookup, because the
 
1972
                        #       parent entry will be the same as the source entry.
 
1973
                        target_parent_entry = state._get_entry(target_index,
 
1974
                                                               path_utf8=new_dirname)
 
1975
                        assert target_parent_entry != (None, None), (
 
1976
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1977
                            % (new_dirname, entry))
 
1978
                        target_parent_id = target_parent_entry[0][2]
 
1979
                    if target_parent_id == entry[0][2]:
 
1980
                        # This is the root, so the parent is None
 
1981
                        target_parent_id = None
 
1982
                    else:
 
1983
                        last_target_parent[0] = new_dirname
 
1984
                        last_target_parent[1] = target_parent_id
 
1985
 
 
1986
                source_exec = source_details[3]
 
1987
                if (include_unchanged
 
1988
                    or content_change
 
1989
                    or source_parent_id != target_parent_id
 
1990
                    or old_basename != entry[0][1]
 
1991
                    or source_exec != target_exec
 
1992
                    ):
 
1993
                    if old_path is None:
 
1994
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1995
                        old_path_u = utf8_decode(old_path)[0]
 
1996
                        path_u = old_path_u
 
1997
                    else:
 
1998
                        old_path_u = utf8_decode(old_path)[0]
 
1999
                        if old_path == path:
 
2000
                            path_u = old_path_u
 
2001
                        else:
 
2002
                            path_u = utf8_decode(path)[0]
 
2003
                    source_kind = _minikind_to_kind[source_minikind]
 
2004
                    return (entry[0][2],
 
2005
                           (old_path_u, path_u),
 
2006
                           content_change,
 
2007
                           (True, True),
 
2008
                           (source_parent_id, target_parent_id),
 
2009
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2010
                           (source_kind, target_kind),
 
2011
                           (source_exec, target_exec))
 
2012
                else:
 
2013
                    return uninteresting
 
2014
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2015
                # looks like a new file
 
2016
                if path_info is not None:
 
2017
                    path = pathjoin(entry[0][0], entry[0][1])
 
2018
                    # parent id is the entry for the path in the target tree
 
2019
                    # TODO: these are the same for an entire directory: cache em.
 
2020
                    parent_id = state._get_entry(target_index,
 
2021
                                                 path_utf8=entry[0][0])[0][2]
 
2022
                    if parent_id == entry[0][2]:
 
2023
                        parent_id = None
 
2024
                    if use_filesystem_for_exec:
 
2025
                        # We need S_ISREG here, because we aren't sure if this
 
2026
                        # is a file or not.
 
2027
                        target_exec = bool(
 
2028
                            stat.S_ISREG(path_info[3].st_mode)
 
2029
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2030
                    else:
 
2031
                        target_exec = target_details[3]
 
2032
                    return (entry[0][2],
 
2033
                           (None, utf8_decode(path)[0]),
 
2034
                           True,
 
2035
                           (False, True),
 
2036
                           (None, parent_id),
 
2037
                           (None, utf8_decode(entry[0][1])[0]),
 
2038
                           (None, path_info[2]),
 
2039
                           (None, target_exec))
 
2040
                else:
 
2041
                    # but its not on disk: we deliberately treat this as just
 
2042
                    # never-present. (Why ?! - RBC 20070224)
 
2043
                    pass
 
2044
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2045
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2046
                # if its still on disk, *and* theres no other entry at this
 
2047
                # path [we dont know this in this routine at the moment -
 
2048
                # perhaps we should change this - then it would be an unknown.
 
2049
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2050
                # parent id is the entry for the path in the target tree
 
2051
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2052
                if parent_id == entry[0][2]:
 
2053
                    parent_id = None
 
2054
                return (entry[0][2],
 
2055
                       (utf8_decode(old_path)[0], None),
 
2056
                       True,
 
2057
                       (True, False),
 
2058
                       (parent_id, None),
 
2059
                       (utf8_decode(entry[0][1])[0], None),
 
2060
                       (_minikind_to_kind[source_minikind], None),
 
2061
                       (source_details[3], None))
 
2062
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2063
                # a rename; could be a true rename, or a rename inherited from
 
2064
                # a renamed parent. TODO: handle this efficiently. Its not
 
2065
                # common case to rename dirs though, so a correct but slow
 
2066
                # implementation will do.
 
2067
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2068
                    search_specific_files.add(target_details[1])
 
2069
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2070
                # neither of the selected trees contain this file,
 
2071
                # so skip over it. This is not currently directly tested, but
 
2072
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2073
                pass
 
2074
            else:
 
2075
                raise AssertionError("don't know how to compare "
 
2076
                    "source_minikind=%r, target_minikind=%r"
 
2077
                    % (source_minikind, target_minikind))
 
2078
                ## import pdb;pdb.set_trace()
 
2079
            return None
 
2080
 
 
2081
        while search_specific_files:
 
2082
            # TODO: the pending list should be lexically sorted?  the
 
2083
            # interface doesn't require it.
 
2084
            current_root = search_specific_files.pop()
 
2085
            current_root_unicode = current_root.decode('utf8')
 
2086
            searched_specific_files.add(current_root)
 
2087
            # process the entries for this containing directory: the rest will be
 
2088
            # found by their parents recursively.
 
2089
            root_entries = _entries_for_path(current_root)
 
2090
            root_abspath = self.target.abspath(current_root_unicode)
 
2091
            try:
 
2092
                root_stat = os.lstat(root_abspath)
 
2093
            except OSError, e:
 
2094
                if e.errno == errno.ENOENT:
 
2095
                    # the path does not exist: let _process_entry know that.
 
2096
                    root_dir_info = None
 
2097
                else:
 
2098
                    # some other random error: hand it up.
 
2099
                    raise
 
2100
            else:
 
2101
                root_dir_info = ('', current_root,
 
2102
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2103
                    root_abspath)
 
2104
                if root_dir_info[2] == 'directory':
 
2105
                    if self.target._directory_is_tree_reference(
 
2106
                        current_root.decode('utf8')):
 
2107
                        root_dir_info = root_dir_info[:2] + \
 
2108
                            ('tree-reference',) + root_dir_info[3:]
 
2109
 
 
2110
            if not root_entries and not root_dir_info:
 
2111
                # this specified path is not present at all, skip it.
 
2112
                continue
 
2113
            path_handled = False
 
2114
            for entry in root_entries:
 
2115
                result = _process_entry(entry, root_dir_info)
 
2116
                if result is not None:
 
2117
                    path_handled = True
 
2118
                    if result is not uninteresting:
 
2119
                        yield result
 
2120
            if want_unversioned and not path_handled and root_dir_info:
 
2121
                new_executable = bool(
 
2122
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2123
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2124
                yield (None,
 
2125
                       (None, current_root_unicode),
 
2126
                       True,
 
2127
                       (False, False),
 
2128
                       (None, None),
 
2129
                       (None, splitpath(current_root_unicode)[-1]),
 
2130
                       (None, root_dir_info[2]),
 
2131
                       (None, new_executable)
 
2132
                      )
 
2133
            initial_key = (current_root, '', '')
 
2134
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2135
            if block_index == 0:
 
2136
                # we have processed the total root already, but because the
 
2137
                # initial key matched it we should skip it here.
 
2138
                block_index +=1
 
2139
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2140
                current_dir_info = None
 
2141
            else:
 
2142
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2143
                try:
 
2144
                    current_dir_info = dir_iterator.next()
 
2145
                except OSError, e:
 
2146
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2147
                    # python 2.5 has e.errno == EINVAL,
 
2148
                    #            and e.winerror == ERROR_DIRECTORY
 
2149
                    e_winerror = getattr(e, 'winerror', None)
 
2150
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2151
                    # there may be directories in the inventory even though
 
2152
                    # this path is not a file on disk: so mark it as end of
 
2153
                    # iterator
 
2154
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2155
                        current_dir_info = None
 
2156
                    elif (sys.platform == 'win32'
 
2157
                          and (e.errno in win_errors
 
2158
                               or e_winerror in win_errors)):
 
2159
                        current_dir_info = None
 
2160
                    else:
 
2161
                        raise
 
2162
                else:
 
2163
                    if current_dir_info[0][0] == '':
 
2164
                        # remove .bzr from iteration
 
2165
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2166
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2167
                        del current_dir_info[1][bzr_index]
 
2168
            # walk until both the directory listing and the versioned metadata
 
2169
            # are exhausted. 
 
2170
            if (block_index < len(state._dirblocks) and
 
2171
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2172
                current_block = state._dirblocks[block_index]
 
2173
            else:
 
2174
                current_block = None
 
2175
            while (current_dir_info is not None or
 
2176
                   current_block is not None):
 
2177
                if (current_dir_info and current_block
 
2178
                    and current_dir_info[0][0] != current_block[0]):
 
2179
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2180
                        # filesystem data refers to paths not covered by the dirblock.
 
2181
                        # this has two possibilities:
 
2182
                        # A) it is versioned but empty, so there is no block for it
 
2183
                        # B) it is not versioned.
 
2184
 
 
2185
                        # if (A) then we need to recurse into it to check for
 
2186
                        # new unknown files or directories.
 
2187
                        # if (B) then we should ignore it, because we don't
 
2188
                        # recurse into unknown directories.
 
2189
                        path_index = 0
 
2190
                        while path_index < len(current_dir_info[1]):
 
2191
                                current_path_info = current_dir_info[1][path_index]
 
2192
                                if want_unversioned:
 
2193
                                    if current_path_info[2] == 'directory':
 
2194
                                        if self.target._directory_is_tree_reference(
 
2195
                                            current_path_info[0].decode('utf8')):
 
2196
                                            current_path_info = current_path_info[:2] + \
 
2197
                                                ('tree-reference',) + current_path_info[3:]
 
2198
                                    new_executable = bool(
 
2199
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2200
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2201
                                    yield (None,
 
2202
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2203
                                        True,
 
2204
                                        (False, False),
 
2205
                                        (None, None),
 
2206
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2207
                                        (None, current_path_info[2]),
 
2208
                                        (None, new_executable))
 
2209
                                # dont descend into this unversioned path if it is
 
2210
                                # a dir
 
2211
                                if current_path_info[2] in ('directory',
 
2212
                                                            'tree-reference'):
 
2213
                                    del current_dir_info[1][path_index]
 
2214
                                    path_index -= 1
 
2215
                                path_index += 1
 
2216
 
 
2217
                        # This dir info has been handled, go to the next
 
2218
                        try:
 
2219
                            current_dir_info = dir_iterator.next()
 
2220
                        except StopIteration:
 
2221
                            current_dir_info = None
 
2222
                    else:
 
2223
                        # We have a dirblock entry for this location, but there
 
2224
                        # is no filesystem path for this. This is most likely
 
2225
                        # because a directory was removed from the disk.
 
2226
                        # We don't have to report the missing directory,
 
2227
                        # because that should have already been handled, but we
 
2228
                        # need to handle all of the files that are contained
 
2229
                        # within.
 
2230
                        for current_entry in current_block[1]:
 
2231
                            # entry referring to file not present on disk.
 
2232
                            # advance the entry only, after processing.
 
2233
                            result = _process_entry(current_entry, None)
 
2234
                            if result is not None:
 
2235
                                if result is not uninteresting:
 
2236
                                    yield result
 
2237
                        block_index +=1
 
2238
                        if (block_index < len(state._dirblocks) and
 
2239
                            osutils.is_inside(current_root,
 
2240
                                              state._dirblocks[block_index][0])):
 
2241
                            current_block = state._dirblocks[block_index]
 
2242
                        else:
 
2243
                            current_block = None
 
2244
                    continue
 
2245
                entry_index = 0
 
2246
                if current_block and entry_index < len(current_block[1]):
 
2247
                    current_entry = current_block[1][entry_index]
 
2248
                else:
 
2249
                    current_entry = None
 
2250
                advance_entry = True
 
2251
                path_index = 0
 
2252
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2253
                    current_path_info = current_dir_info[1][path_index]
 
2254
                    if current_path_info[2] == 'directory':
 
2255
                        if self.target._directory_is_tree_reference(
 
2256
                            current_path_info[0].decode('utf8')):
 
2257
                            current_path_info = current_path_info[:2] + \
 
2258
                                ('tree-reference',) + current_path_info[3:]
 
2259
                else:
 
2260
                    current_path_info = None
 
2261
                advance_path = True
 
2262
                path_handled = False
 
2263
                while (current_entry is not None or
 
2264
                    current_path_info is not None):
 
2265
                    if current_entry is None:
 
2266
                        # the check for path_handled when the path is adnvaced
 
2267
                        # will yield this path if needed.
 
2268
                        pass
 
2269
                    elif current_path_info is None:
 
2270
                        # no path is fine: the per entry code will handle it.
 
2271
                        result = _process_entry(current_entry, current_path_info)
 
2272
                        if result is not None:
 
2273
                            if result is not uninteresting:
 
2274
                                yield result
 
2275
                    elif (current_entry[0][1] != current_path_info[1]
 
2276
                          or current_entry[1][target_index][0] in 'ar'):
 
2277
                        # The current path on disk doesn't match the dirblock
 
2278
                        # record. Either the dirblock is marked as absent, or
 
2279
                        # the file on disk is not present at all in the
 
2280
                        # dirblock. Either way, report about the dirblock
 
2281
                        # entry, and let other code handle the filesystem one.
 
2282
 
 
2283
                        # Compare the basename for these files to determine
 
2284
                        # which comes first
 
2285
                        if current_path_info[1] < current_entry[0][1]:
 
2286
                            # extra file on disk: pass for now, but only
 
2287
                            # increment the path, not the entry
 
2288
                            advance_entry = False
 
2289
                        else:
 
2290
                            # entry referring to file not present on disk.
 
2291
                            # advance the entry only, after processing.
 
2292
                            result = _process_entry(current_entry, None)
 
2293
                            if result is not None:
 
2294
                                if result is not uninteresting:
 
2295
                                    yield result
 
2296
                            advance_path = False
 
2297
                    else:
 
2298
                        result = _process_entry(current_entry, current_path_info)
 
2299
                        if result is not None:
 
2300
                            path_handled = True
 
2301
                            if result is not uninteresting:
 
2302
                                yield result
 
2303
                    if advance_entry and current_entry is not None:
 
2304
                        entry_index += 1
 
2305
                        if entry_index < len(current_block[1]):
 
2306
                            current_entry = current_block[1][entry_index]
 
2307
                        else:
 
2308
                            current_entry = None
 
2309
                    else:
 
2310
                        advance_entry = True # reset the advance flaga
 
2311
                    if advance_path and current_path_info is not None:
 
2312
                        if not path_handled:
 
2313
                            # unversioned in all regards
 
2314
                            if want_unversioned:
 
2315
                                new_executable = bool(
 
2316
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2317
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2318
                                yield (None,
 
2319
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2320
                                    True,
 
2321
                                    (False, False),
 
2322
                                    (None, None),
 
2323
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2324
                                    (None, current_path_info[2]),
 
2325
                                    (None, new_executable))
 
2326
                            # dont descend into this unversioned path if it is
 
2327
                            # a dir
 
2328
                            if current_path_info[2] in ('directory'):
 
2329
                                del current_dir_info[1][path_index]
 
2330
                                path_index -= 1
 
2331
                        # dont descend the disk iterator into any tree 
 
2332
                        # paths.
 
2333
                        if current_path_info[2] == 'tree-reference':
 
2334
                            del current_dir_info[1][path_index]
 
2335
                            path_index -= 1
 
2336
                        path_index += 1
 
2337
                        if path_index < len(current_dir_info[1]):
 
2338
                            current_path_info = current_dir_info[1][path_index]
 
2339
                            if current_path_info[2] == 'directory':
 
2340
                                if self.target._directory_is_tree_reference(
 
2341
                                    current_path_info[0].decode('utf8')):
 
2342
                                    current_path_info = current_path_info[:2] + \
 
2343
                                        ('tree-reference',) + current_path_info[3:]
 
2344
                        else:
 
2345
                            current_path_info = None
 
2346
                        path_handled = False
 
2347
                    else:
 
2348
                        advance_path = True # reset the advance flagg.
 
2349
                if current_block is not None:
 
2350
                    block_index += 1
 
2351
                    if (block_index < len(state._dirblocks) and
 
2352
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2353
                        current_block = state._dirblocks[block_index]
 
2354
                    else:
 
2355
                        current_block = None
 
2356
                if current_dir_info is not None:
 
2357
                    try:
 
2358
                        current_dir_info = dir_iterator.next()
 
2359
                    except StopIteration:
 
2360
                        current_dir_info = None
 
2361
 
 
2362
 
 
2363
    @staticmethod
 
2364
    def is_compatible(source, target):
 
2365
        # the target must be a dirstate working tree
 
2366
        if not isinstance(target, WorkingTree4):
 
2367
            return False
 
2368
        # the source must be a revtreee or dirstate rev tree.
 
2369
        if not isinstance(source,
 
2370
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2371
            return False
 
2372
        # the source revid must be in the target dirstate
 
2373
        if not (source._revision_id == NULL_REVISION or
 
2374
            source._revision_id in target.get_parent_ids()):
 
2375
            # TODO: what about ghosts? it may well need to 
 
2376
            # check for them explicitly.
 
2377
            return False
 
2378
        return True
 
2379
 
 
2380
InterTree.register_optimiser(InterDirStateTree)
 
2381
 
 
2382
 
 
2383
class Converter3to4(object):
 
2384
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2385
 
 
2386
    def __init__(self):
 
2387
        self.target_format = WorkingTreeFormat4()
 
2388
 
 
2389
    def convert(self, tree):
 
2390
        # lock the control files not the tree, so that we dont get tree
 
2391
        # on-unlock behaviours, and so that noone else diddles with the 
 
2392
        # tree during upgrade.
 
2393
        tree._control_files.lock_write()
 
2394
        try:
 
2395
            tree.read_working_inventory()
 
2396
            self.create_dirstate_data(tree)
 
2397
            self.update_format(tree)
 
2398
            self.remove_xml_files(tree)
 
2399
        finally:
 
2400
            tree._control_files.unlock()
 
2401
 
 
2402
    def create_dirstate_data(self, tree):
 
2403
        """Create the dirstate based data for tree."""
 
2404
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2405
            ).local_abspath('dirstate')
 
2406
        state = dirstate.DirState.from_tree(tree, local_path)
 
2407
        state.save()
 
2408
        state.unlock()
 
2409
 
 
2410
    def remove_xml_files(self, tree):
 
2411
        """Remove the oldformat 3 data."""
 
2412
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2413
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2414
            'pending-merges', 'stat-cache']:
 
2415
            try:
 
2416
                transport.delete(path)
 
2417
            except errors.NoSuchFile:
 
2418
                # some files are optional - just deal.
 
2419
                pass
 
2420
 
 
2421
    def update_format(self, tree):
 
2422
        """Change the format marker."""
 
2423
        tree._control_files.put_utf8('format',
 
2424
            self.target_format.get_format_string())