~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2007-03-07 05:41:58 UTC
  • mto: (2321.1.1 integration)
  • mto: This revision was merged to the branch mainline in revision 2322.
  • Revision ID: robertc@robertcollins.net-20070307054158-j7yh62s59nz3rqdy
Update NEWS and clean out BRANCH.TODO.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2012 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
22
22
WorkingTree.open(dir).
23
23
"""
24
24
 
25
 
from __future__ import absolute_import
26
 
 
27
25
from cStringIO import StringIO
28
26
import os
29
27
import sys
30
28
 
31
29
from bzrlib.lazy_import import lazy_import
32
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
33
34
import errno
 
35
import itertools
 
36
import operator
34
37
import stat
 
38
from time import time
 
39
import warnings
35
40
 
 
41
import bzrlib
36
42
from bzrlib import (
37
43
    bzrdir,
38
44
    cache_utf8,
39
 
    config,
40
45
    conflicts as _mod_conflicts,
41
 
    controldir,
42
 
    debug,
 
46
    delta,
43
47
    dirstate,
44
48
    errors,
45
 
    filters as _mod_filters,
46
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
47
54
    osutils,
48
 
    revision as _mod_revision,
49
55
    revisiontree,
50
 
    trace,
 
56
    textui,
51
57
    transform,
52
 
    views,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
53
61
    )
 
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
 
64
import bzrlib.ui
54
65
""")
55
66
 
 
67
from bzrlib import symbol_versioning
56
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
57
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
58
 
from bzrlib.lock import LogicalLockResult
59
 
from bzrlib.lockable_files import LockableFiles
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
60
71
from bzrlib.lockdir import LockDir
61
 
from bzrlib.mutabletree import (
62
 
    MutableTree,
63
 
    needs_tree_write_lock,
64
 
    )
 
72
import bzrlib.mutabletree
 
73
from bzrlib.mutabletree import needs_tree_write_lock
65
74
from bzrlib.osutils import (
66
75
    file_kind,
67
76
    isdir,
 
77
    normpath,
68
78
    pathjoin,
 
79
    rand_chars,
69
80
    realpath,
70
81
    safe_unicode,
71
 
    )
72
 
from bzrlib.symbol_versioning import (
73
 
    deprecated_in,
74
 
    deprecated_method,
75
 
    )
 
82
    splitpath,
 
83
    )
 
84
from bzrlib.trace import mutter, note
76
85
from bzrlib.transport.local import LocalTransport
77
 
from bzrlib.tree import (
78
 
    InterTree,
79
 
    InventoryTree,
80
 
    )
81
 
from bzrlib.workingtree import (
82
 
    InventoryWorkingTree,
83
 
    WorkingTree,
84
 
    WorkingTreeFormatMetaDir,
85
 
    )
86
 
 
87
 
 
88
 
class DirStateWorkingTree(InventoryWorkingTree):
 
86
from bzrlib.tree import InterTree
 
87
from bzrlib.progress import DummyProgress, ProgressPhase
 
88
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
89
from bzrlib.rio import RioReader, rio_file, Stanza
 
90
from bzrlib.symbol_versioning import (deprecated_passed,
 
91
        deprecated_method,
 
92
        deprecated_function,
 
93
        DEPRECATED_PARAMETER,
 
94
        )
 
95
from bzrlib.tree import Tree
 
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
97
 
 
98
 
 
99
class WorkingTree4(WorkingTree3):
 
100
    """This is the Format 4 working tree.
 
101
 
 
102
    This differs from WorkingTree3 by:
 
103
     - Having a consolidated internal dirstate, stored in a
 
104
       randomly-accessible sorted file on disk.
 
105
     - Not having a regular inventory attribute.  One can be synthesized 
 
106
       on demand but this is expensive and should be avoided.
 
107
 
 
108
    This is new in bzr 0.15.
 
109
    """
89
110
 
90
111
    def __init__(self, basedir,
91
112
                 branch,
101
122
        """
102
123
        self._format = _format
103
124
        self.bzrdir = _bzrdir
 
125
        from bzrlib.trace import note, mutter
 
126
        assert isinstance(basedir, basestring), \
 
127
            "base directory %r is not a string" % basedir
104
128
        basedir = safe_unicode(basedir)
105
 
        trace.mutter("opening working tree %r", basedir)
 
129
        mutter("opening working tree %r", basedir)
106
130
        self._branch = branch
 
131
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
132
            "branch %r is not a Branch" % self.branch
107
133
        self.basedir = realpath(basedir)
108
134
        # if branch is at our basedir and is a format 6 or less
109
135
        # assume all other formats have their own control files.
 
136
        assert isinstance(_control_files, LockableFiles), \
 
137
            "_control_files must be a LockableFiles, not %r" % _control_files
110
138
        self._control_files = _control_files
111
 
        self._transport = self._control_files._transport
112
139
        self._dirty = None
113
140
        #-------------
114
141
        # during a read or write lock these objects are set, and are
116
143
        self._dirstate = None
117
144
        self._inventory = None
118
145
        #-------------
119
 
        self._setup_directory_is_tree_reference()
120
 
        self._detect_case_handling()
121
 
        self._rules_searcher = None
122
 
        self.views = self._make_views()
123
 
        #--- allow tests to select the dirstate iter_changes implementation
124
 
        self._iter_changes = dirstate._process_entry
125
146
 
126
147
    @needs_tree_write_lock
127
148
    def _add(self, files, ids, kinds):
129
150
        state = self.current_dirstate()
130
151
        for f, file_id, kind in zip(files, ids, kinds):
131
152
            f = f.strip('/')
 
153
            assert '//' not in f
 
154
            assert '..' not in f
132
155
            if self.path2id(f):
133
156
                # special case tree root handling.
134
157
                if f == '' and self.path2id(f) == ROOT_ID:
142
165
            state.add(f, file_id, kind, None, '')
143
166
        self._make_dirty(reset_inventory=True)
144
167
 
145
 
    def _get_check_refs(self):
146
 
        """Return the references needed to perform a check of this tree."""
147
 
        return [('trees', self.last_revision())]
148
 
 
149
168
    def _make_dirty(self, reset_inventory):
150
169
        """Make the tree state dirty.
151
170
 
159
178
    @needs_tree_write_lock
160
179
    def add_reference(self, sub_tree):
161
180
        # use standard implementation, which calls back to self._add
162
 
        #
 
181
        # 
163
182
        # So we don't store the reference_revision in the working dirstate,
164
 
        # it's just recorded at the moment of commit.
 
183
        # it's just recorded at the moment of commit. 
165
184
        self._add_reference(sub_tree)
166
185
 
167
186
    def break_lock(self):
203
222
 
204
223
    def _comparison_data(self, entry, path):
205
224
        kind, executable, stat_value = \
206
 
            WorkingTree._comparison_data(self, entry, path)
 
225
            WorkingTree3._comparison_data(self, entry, path)
207
226
        # it looks like a plain directory, but it's really a reference -- see
208
227
        # also kind()
209
 
        if (self._repo_supports_tree_reference and kind == 'directory'
210
 
            and entry is not None and entry.kind == 'tree-reference'):
 
228
        if (self._repo_supports_tree_reference and
 
229
            kind == 'directory' and
 
230
            self._directory_is_tree_reference(path)):
211
231
            kind = 'tree-reference'
212
232
        return kind, executable, stat_value
213
233
 
215
235
    def commit(self, message=None, revprops=None, *args, **kwargs):
216
236
        # mark the tree as dirty post commit - commit
217
237
        # can change the current versioned list by doing deletes.
218
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
238
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
219
239
        self._make_dirty(reset_inventory=True)
220
240
        return result
221
241
 
239
259
            return self._dirstate
240
260
        local_path = self.bzrdir.get_workingtree_transport(None
241
261
            ).local_abspath('dirstate')
242
 
        self._dirstate = dirstate.DirState.on_file(local_path,
243
 
            self._sha1_provider(), self._worth_saving_limit())
 
262
        self._dirstate = dirstate.DirState.on_file(local_path)
244
263
        return self._dirstate
245
264
 
246
 
    def _sha1_provider(self):
247
 
        """A function that returns a SHA1Provider suitable for this tree.
248
 
 
249
 
        :return: None if content filtering is not supported by this tree.
250
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
251
 
          form of files, i.e. after read filters are applied.
252
 
        """
253
 
        if self.supports_content_filtering():
254
 
            return ContentFilterAwareSHA1Provider(self)
255
 
        else:
256
 
            return None
257
 
 
258
 
    def _worth_saving_limit(self):
259
 
        """How many hash changes are ok before we must save the dirstate.
260
 
 
261
 
        :return: an integer. -1 means never save.
262
 
        """
263
 
        conf = self.get_config_stack()
264
 
        return conf.get('bzr.workingtree.worth_saving_limit')
 
265
    def _directory_is_tree_reference(self, relpath):
 
266
        # as a special case, if a directory contains control files then 
 
267
        # it's a tree reference, except that the root of the tree is not
 
268
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
269
        # TODO: We could ask all the control formats whether they
 
270
        # recognize this directory, but at the moment there's no cheap api
 
271
        # to do that.  Since we probably can only nest bzr checkouts and
 
272
        # they always use this name it's ok for now.  -- mbp 20060306
 
273
        #
 
274
        # FIXME: There is an unhandled case here of a subdirectory
 
275
        # containing .bzr but not a branch; that will probably blow up
 
276
        # when you try to commit it.  It might happen if there is a
 
277
        # checkout in a subdirectory.  This can be avoided by not adding
 
278
        # it.  mbp 20070306
265
279
 
266
280
    def filter_unversioned_files(self, paths):
267
281
        """Filter out paths that are versioned.
291
305
        self._inventory = None
292
306
        self._dirty = False
293
307
 
294
 
    @needs_tree_write_lock
295
 
    def _gather_kinds(self, files, kinds):
296
 
        """See MutableTree._gather_kinds."""
297
 
        for pos, f in enumerate(files):
298
 
            if kinds[pos] is None:
299
 
                kinds[pos] = self._kind(f)
300
 
 
301
308
    def _generate_inventory(self):
302
309
        """Create and set self.inventory from the dirstate object.
303
 
 
 
310
        
304
311
        This is relatively expensive: we have to walk the entire dirstate.
305
312
        Ideally we would not, and can deprecate this function.
306
313
        """
310
317
        state._read_dirblocks_if_needed()
311
318
        root_key, current_entry = self._get_entry(path='')
312
319
        current_id = root_key[2]
313
 
        if not (current_entry[0][0] == 'd'): # directory
314
 
            raise AssertionError(current_entry)
 
320
        assert current_entry[0][0] == 'd' # directory
315
321
        inv = Inventory(root_id=current_id)
316
322
        # Turn some things into local variables
317
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
340
346
                inv_entry = factory[kind](file_id, name_unicode,
341
347
                                          parent_ie.file_id)
342
348
                if kind == 'file':
343
 
                    # This is only needed on win32, where this is the only way
344
 
                    # we know the executable bit.
345
 
                    inv_entry.executable = executable
346
349
                    # not strictly needed: working tree
347
 
                    #inv_entry.text_size = size
348
 
                    #inv_entry.text_sha1 = sha1
 
350
                    #entry.executable = executable
 
351
                    #entry.text_size = size
 
352
                    #entry.text_sha1 = sha1
 
353
                    pass
349
354
                elif kind == 'directory':
350
355
                    # add this entry to the parent map.
351
356
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
352
357
                elif kind == 'tree-reference':
353
 
                    if not self._repo_supports_tree_reference:
354
 
                        raise errors.UnsupportedOperation(
355
 
                            self._generate_inventory,
356
 
                            self.branch.repository)
 
358
                    assert self._repo_supports_tree_reference
357
359
                    inv_entry.reference_revision = link_or_sha1 or None
358
 
                elif kind != 'symlink':
359
 
                    raise AssertionError("unknown kind %r" % kind)
 
360
                else:
 
361
                    assert 'unknown kind'
360
362
                # These checks cost us around 40ms on a 55k entry tree
361
 
                if file_id in inv_byid:
362
 
                    raise AssertionError('file_id %s already in'
363
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
364
 
                if name_unicode in parent_ie.children:
365
 
                    raise AssertionError('name %r already in parent'
366
 
                        % (name_unicode,))
 
363
                assert file_id not in inv_byid, ('file_id %s already in'
 
364
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
365
                assert name_unicode not in parent_ie.children
367
366
                inv_byid[file_id] = inv_entry
368
367
                parent_ie.children[name_unicode] = inv_entry
369
368
        self._inventory = inv
374
373
        If either file_id or path is supplied, it is used as the key to lookup.
375
374
        If both are supplied, the fastest lookup is used, and an error is
376
375
        raised if they do not both point at the same row.
377
 
 
 
376
        
378
377
        :param file_id: An optional unicode file_id to be looked up.
379
378
        :param path: An optional unicode path to be looked up.
380
379
        :return: The dirstate row tuple for path/file_id, or (None, None)
389
388
    def get_file_sha1(self, file_id, path=None, stat_value=None):
390
389
        # check file id is valid unconditionally.
391
390
        entry = self._get_entry(file_id=file_id, path=path)
392
 
        if entry[0] is None:
393
 
            raise errors.NoSuchId(self, file_id)
 
391
        assert entry[0] is not None, 'what error should this raise'
394
392
        if path is None:
395
393
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
396
394
 
397
395
        file_abspath = self.abspath(path)
398
396
        state = self.current_dirstate()
399
 
        if stat_value is None:
400
 
            try:
401
 
                stat_value = osutils.lstat(file_abspath)
402
 
            except OSError, e:
403
 
                if e.errno == errno.ENOENT:
404
 
                    return None
405
 
                else:
406
 
                    raise
407
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
408
 
            stat_value=stat_value)
 
397
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
398
                                          stat_value=stat_value)
409
399
        if entry[1][0][0] == 'f':
410
 
            if link_or_sha1 is None:
411
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
412
 
                try:
413
 
                    sha1 = osutils.sha_file(file_obj)
414
 
                finally:
415
 
                    file_obj.close()
416
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
417
 
                return sha1
418
 
            else:
419
 
                return link_or_sha1
 
400
            return link_or_sha1
420
401
        return None
421
402
 
422
 
    def _get_root_inventory(self):
 
403
    def _get_inventory(self):
423
404
        """Get the inventory for the tree. This is only valid within a lock."""
424
 
        if 'evil' in debug.debug_flags:
425
 
            trace.mutter_callsite(2,
426
 
                "accessing .inventory forces a size of tree translation.")
427
405
        if self._inventory is not None:
428
406
            return self._inventory
429
407
        self._must_be_locked()
430
408
        self._generate_inventory()
431
409
        return self._inventory
432
410
 
433
 
    @deprecated_method(deprecated_in((2, 5, 0)))
434
 
    def _get_inventory(self):
435
 
        return self.root_inventory
436
 
 
437
411
    inventory = property(_get_inventory,
438
412
                         doc="Inventory of this Tree")
439
413
 
440
 
    root_inventory = property(_get_root_inventory,
441
 
        "Root inventory of this tree")
442
 
 
443
414
    @needs_read_lock
444
415
    def get_parent_ids(self):
445
416
        """See Tree.get_parent_ids.
446
 
 
 
417
        
447
418
        This implementation requests the ids list from the dirstate file.
448
419
        """
449
420
        return self.current_dirstate().get_parent_ids()
465
436
 
466
437
    def has_id(self, file_id):
467
438
        state = self.current_dirstate()
 
439
        file_id = osutils.safe_file_id(file_id)
468
440
        row, parents = self._get_entry(file_id=file_id)
469
441
        if row is None:
470
442
            return False
471
443
        return osutils.lexists(pathjoin(
472
444
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
473
445
 
474
 
    def has_or_had_id(self, file_id):
475
 
        state = self.current_dirstate()
476
 
        row, parents = self._get_entry(file_id=file_id)
477
 
        return row is not None
478
 
 
479
446
    @needs_read_lock
480
447
    def id2path(self, file_id):
481
 
        "Convert a file-id to a path."
 
448
        file_id = osutils.safe_file_id(file_id)
482
449
        state = self.current_dirstate()
483
450
        entry = self._get_entry(file_id=file_id)
484
451
        if entry == (None, None):
486
453
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
487
454
        return path_utf8.decode('utf8')
488
455
 
489
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
490
 
        entry = self._get_entry(path=path)
491
 
        if entry == (None, None):
492
 
            return False # Missing entries are not executable
493
 
        return entry[1][0][3] # Executable?
494
 
 
495
 
    def is_executable(self, file_id, path=None):
496
 
        """Test if a file is executable or not.
497
 
 
498
 
        Note: The caller is expected to take a read-lock before calling this.
499
 
        """
500
 
        if not self._supports_executable():
501
 
            entry = self._get_entry(file_id=file_id, path=path)
502
 
            if entry == (None, None):
503
 
                return False
504
 
            return entry[1][0][3]
505
 
        else:
506
 
            self._must_be_locked()
507
 
            if not path:
508
 
                path = self.id2path(file_id)
509
 
            mode = osutils.lstat(self.abspath(path)).st_mode
510
 
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
511
 
 
512
 
    def all_file_ids(self):
513
 
        """See Tree.iter_all_file_ids"""
514
 
        self._must_be_locked()
515
 
        result = set()
516
 
        for key, tree_details in self.current_dirstate()._iter_entries():
517
 
            if tree_details[0][0] in ('a', 'r'): # relocated
518
 
                continue
519
 
            result.add(key[2])
520
 
        return result
521
 
 
522
456
    @needs_read_lock
523
457
    def __iter__(self):
524
458
        """Iterate through file_ids for this tree.
537
471
        return iter(result)
538
472
 
539
473
    def iter_references(self):
540
 
        if not self._repo_supports_tree_reference:
541
 
            # When the repo doesn't support references, we will have nothing to
542
 
            # return
543
 
            return
544
474
        for key, tree_details in self.current_dirstate()._iter_entries():
545
475
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
546
476
                # not relevant to the working tree
548
478
            if not key[1]:
549
479
                # the root is not a reference.
550
480
                continue
551
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
481
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
552
482
            try:
553
 
                if self._kind(relpath) == 'tree-reference':
554
 
                    yield relpath, key[2]
 
483
                if self._kind(path) == 'tree-reference':
 
484
                    yield path, key[2]
555
485
            except errors.NoSuchFile:
556
486
                # path is missing on disk.
557
487
                continue
558
488
 
559
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
560
 
        """See MutableTree._observed_sha1."""
561
 
        state = self.current_dirstate()
562
 
        entry = self._get_entry(file_id=file_id, path=path)
563
 
        state._observed_sha1(entry, sha1, statvalue)
564
 
 
 
489
    @needs_read_lock
565
490
    def kind(self, file_id):
566
491
        """Return the kind of a file.
567
492
 
568
493
        This is always the actual kind that's on disk, regardless of what it
569
494
        was added as.
570
 
 
571
 
        Note: The caller is expected to take a read-lock before calling this.
572
495
        """
573
496
        relpath = self.id2path(file_id)
574
 
        if relpath is None:
575
 
            raise AssertionError(
576
 
                "path for id {%s} is None!" % file_id)
 
497
        assert relpath != None, \
 
498
            "path for id {%s} is None!" % file_id
577
499
        return self._kind(relpath)
578
500
 
579
501
    def _kind(self, relpath):
580
502
        abspath = self.abspath(relpath)
581
503
        kind = file_kind(abspath)
582
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
583
 
            entry = self._get_entry(path=relpath)
584
 
            if entry[1] is not None:
585
 
                if entry[1][0][0] == 't':
586
 
                    kind = 'tree-reference'
 
504
        if (self._repo_supports_tree_reference and
 
505
            kind == 'directory' and
 
506
            self._directory_is_tree_reference(relpath)):
 
507
            kind = 'tree-reference'
587
508
        return kind
588
509
 
589
510
    @needs_read_lock
593
514
        if parent_ids:
594
515
            return parent_ids[0]
595
516
        else:
596
 
            return _mod_revision.NULL_REVISION
 
517
            return None
597
518
 
598
519
    def lock_read(self):
599
 
        """See Branch.lock_read, and WorkingTree.unlock.
600
 
 
601
 
        :return: A bzrlib.lock.LogicalLockResult.
602
 
        """
 
520
        """See Branch.lock_read, and WorkingTree.unlock."""
603
521
        self.branch.lock_read()
604
522
        try:
605
523
            self._control_files.lock_read()
610
528
                # set our support for tree references from the repository in
611
529
                # use.
612
530
                self._repo_supports_tree_reference = getattr(
613
 
                    self.branch.repository._format, "supports_tree_reference",
 
531
                    self.branch.repository._format, "support_tree_reference",
614
532
                    False)
615
533
            except:
616
534
                self._control_files.unlock()
618
536
        except:
619
537
            self.branch.unlock()
620
538
            raise
621
 
        return LogicalLockResult(self.unlock)
622
539
 
623
540
    def _lock_self_write(self):
624
541
        """This should be called after the branch is locked."""
631
548
                # set our support for tree references from the repository in
632
549
                # use.
633
550
                self._repo_supports_tree_reference = getattr(
634
 
                    self.branch.repository._format, "supports_tree_reference",
 
551
                    self.branch.repository._format, "support_tree_reference",
635
552
                    False)
636
553
            except:
637
554
                self._control_files.unlock()
639
556
        except:
640
557
            self.branch.unlock()
641
558
            raise
642
 
        return LogicalLockResult(self.unlock)
643
559
 
644
560
    def lock_tree_write(self):
645
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
561
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
649
562
        self.branch.lock_read()
650
 
        return self._lock_self_write()
 
563
        self._lock_self_write()
651
564
 
652
565
    def lock_write(self):
653
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
654
 
 
655
 
        :return: A bzrlib.lock.LogicalLockResult.
656
 
        """
 
566
        """See MutableTree.lock_write, and WorkingTree.unlock."""
657
567
        self.branch.lock_write()
658
 
        return self._lock_self_write()
 
568
        self._lock_self_write()
659
569
 
660
570
    @needs_tree_write_lock
661
571
    def move(self, from_paths, to_dir, after=False):
663
573
        result = []
664
574
        if not from_paths:
665
575
            return result
 
576
 
666
577
        state = self.current_dirstate()
667
 
        if isinstance(from_paths, basestring):
668
 
            raise ValueError()
 
578
 
 
579
        assert not isinstance(from_paths, basestring)
669
580
        to_dir_utf8 = to_dir.encode('utf8')
670
581
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
671
582
        id_index = state._get_id_index()
692
603
 
693
604
        if self._inventory is not None:
694
605
            update_inventory = True
695
 
            inv = self.root_inventory
 
606
            inv = self.inventory
 
607
            to_dir_ie = inv[to_dir_id]
696
608
            to_dir_id = to_entry[0][2]
697
 
            to_dir_ie = inv[to_dir_id]
698
609
        else:
699
610
            update_inventory = False
700
611
 
723
634
            new_entry = to_block[1][added_entry_index]
724
635
            rollbacks.append(lambda:state._make_absent(new_entry))
725
636
 
 
637
        # create rename entries and tuples
726
638
        for from_rel in from_paths:
727
639
            # from_rel is 'pathinroot/foo/bar'
728
640
            from_rel_utf8 = from_rel.encode('utf8')
731
643
            from_entry = self._get_entry(path=from_rel)
732
644
            if from_entry == (None, None):
733
645
                raise errors.BzrMoveFailedError(from_rel,to_dir,
734
 
                    errors.NotVersionedError(path=from_rel))
 
646
                    errors.NotVersionedError(path=str(from_rel)))
735
647
 
736
648
            from_id = from_entry[0][2]
737
649
            to_rel = pathjoin(to_dir, from_tail)
764
676
                if from_missing: # implicitly just update our path mapping
765
677
                    move_file = False
766
678
                elif not after:
767
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
679
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
680
                        extra="(Use --after to update the Bazaar id)")
768
681
 
769
682
            rollbacks = []
770
683
            def rollback_rename():
825
738
 
826
739
                if minikind == 'd':
827
740
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
828
 
                        """Recursively update all entries in this dirblock."""
829
 
                        if from_dir == '':
830
 
                            raise AssertionError("renaming root not supported")
 
741
                        """all entries in this block need updating.
 
742
 
 
743
                        TODO: This is pretty ugly, and doesn't support
 
744
                        reverting, but it works.
 
745
                        """
 
746
                        assert from_dir != '', "renaming root not supported"
831
747
                        from_key = (from_dir, '')
832
748
                        from_block_idx, present = \
833
749
                            state._find_block_index_from_key(from_key)
843
759
                        to_block_index = state._ensure_block(
844
760
                            to_block_index, to_entry_index, to_dir_utf8)
845
761
                        to_block = state._dirblocks[to_block_index]
846
 
 
847
 
                        # Grab a copy since move_one may update the list.
848
 
                        for entry in from_block[1][:]:
849
 
                            if not (entry[0][0] == from_dir):
850
 
                                raise AssertionError()
 
762
                        for entry in from_block[1]:
 
763
                            assert entry[0][0] == from_dir
851
764
                            cur_details = entry[1][0]
852
765
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
853
766
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
854
767
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
855
768
                            minikind = cur_details[0]
856
 
                            if minikind in 'ar':
857
 
                                # Deleted children of a renamed directory
858
 
                                # Do not need to be updated.
859
 
                                # Children that have been renamed out of this
860
 
                                # directory should also not be updated
861
 
                                continue
862
769
                            move_one(entry, from_path_utf8=from_path_utf8,
863
770
                                     minikind=minikind,
864
771
                                     executable=cur_details[3],
867
774
                                     size=cur_details[2],
868
775
                                     to_block=to_block,
869
776
                                     to_key=to_key,
870
 
                                     to_path_utf8=to_path_utf8)
 
777
                                     to_path_utf8=to_rel_utf8)
871
778
                            if minikind == 'd':
872
779
                                # We need to move all the children of this
873
780
                                # entry
878
785
                rollback_rename()
879
786
                raise
880
787
            result.append((from_rel, to_rel))
881
 
            state._mark_modified()
 
788
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
882
789
            self._make_dirty(reset_inventory=False)
883
790
 
884
791
        return result
894
801
    @needs_read_lock
895
802
    def path2id(self, path):
896
803
        """Return the id for path in this tree."""
897
 
        if isinstance(path, list):
898
 
            if path == []:
899
 
                path = [""]
900
 
            path = osutils.pathjoin(*path)
901
804
        path = path.strip('/')
902
805
        entry = self._get_entry(path=path)
903
806
        if entry == (None, None):
916
819
        for tree in trees:
917
820
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
918
821
                parents):
919
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
920
 
                    trees, require_versioned)
 
822
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
921
823
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
922
824
        # -- make all paths utf8 --
923
825
        paths_utf8 = set()
981
883
                    all_versioned = False
982
884
                    break
983
885
            if not all_versioned:
984
 
                raise errors.PathsNotVersionedError(
985
 
                    [p.decode('utf-8') for p in paths])
 
886
                raise errors.PathsNotVersionedError(paths)
986
887
        # -- remove redundancy in supplied paths to prevent over-scanning --
987
 
        search_paths = osutils.minimum_path_selection(paths)
988
 
        # sketch:
 
888
        search_paths = set()
 
889
        for path in paths:
 
890
            other_paths = paths.difference(set([path]))
 
891
            if not osutils.is_inside_any(other_paths, path):
 
892
                # this is a top level path, we must check it.
 
893
                search_paths.add(path)
 
894
        # sketch: 
989
895
        # for all search_indexs in each path at or under each element of
990
896
        # search_paths, if the detail is relocated: add the id, and add the
991
897
        # relocated path as one to search if its not searched already. If the
1037
943
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1038
944
            for dir_name in split_paths:
1039
945
                if dir_name not in found_dir_names:
1040
 
                    raise errors.PathsNotVersionedError(
1041
 
                        [p.decode('utf-8') for p in paths])
 
946
                    raise errors.PathsNotVersionedError(paths)
1042
947
 
1043
948
        for dir_name_id, trees_info in found.iteritems():
1044
949
            for index in search_indexes:
1048
953
 
1049
954
    def read_working_inventory(self):
1050
955
        """Read the working inventory.
1051
 
 
 
956
        
1052
957
        This is a meaningless operation for dirstate, but we obey it anyhow.
1053
958
        """
1054
 
        return self.root_inventory
 
959
        return self.inventory
1055
960
 
1056
961
    @needs_read_lock
1057
962
    def revision_tree(self, revision_id):
1059
964
 
1060
965
        WorkingTree4 supplies revision_trees for any basis tree.
1061
966
        """
 
967
        revision_id = osutils.safe_revision_id(revision_id)
1062
968
        dirstate = self.current_dirstate()
1063
969
        parent_ids = dirstate.get_parent_ids()
1064
970
        if revision_id not in parent_ids:
1071
977
    @needs_tree_write_lock
1072
978
    def set_last_revision(self, new_revision):
1073
979
        """Change the last revision in the working tree."""
 
980
        new_revision = osutils.safe_revision_id(new_revision)
1074
981
        parents = self.get_parent_ids()
1075
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1076
 
            if len(parents) >= 2:
1077
 
                raise AssertionError(
1078
 
                    "setting the last parent to none with a pending merge is "
1079
 
                    "unsupported.")
 
982
        if new_revision in (NULL_REVISION, None):
 
983
            assert len(parents) < 2, (
 
984
                "setting the last parent to none with a pending merge is "
 
985
                "unsupported.")
1080
986
            self.set_parent_ids([])
1081
987
        else:
1082
988
            self.set_parent_ids([new_revision] + parents[1:],
1085
991
    @needs_tree_write_lock
1086
992
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1087
993
        """Set the parent ids to revision_ids.
1088
 
 
 
994
        
1089
995
        See also set_parent_trees. This api will try to retrieve the tree data
1090
996
        for each element of revision_ids from the trees repository. If you have
1091
997
        tree data already available, it is more efficient to use
1095
1001
        :param revision_ids: The revision_ids to set as the parent ids of this
1096
1002
            working tree. Any of these may be ghosts.
1097
1003
        """
 
1004
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1098
1005
        trees = []
1099
1006
        for revision_id in revision_ids:
1100
1007
            try:
1106
1013
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1107
1014
                revtree = None
1108
1015
            trees.append((revision_id, revtree))
 
1016
        self.current_dirstate()._validate()
1109
1017
        self.set_parent_trees(trees,
1110
1018
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1019
        self.current_dirstate()._validate()
1111
1020
 
1112
1021
    @needs_tree_write_lock
1113
1022
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1118
1027
            parent tree - i.e. a ghost.
1119
1028
        """
1120
1029
        dirstate = self.current_dirstate()
 
1030
        dirstate._validate()
1121
1031
        if len(parents_list) > 0:
1122
1032
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1123
1033
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1124
1034
        real_trees = []
1125
1035
        ghosts = []
1126
 
 
1127
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1128
 
        graph = self.branch.repository.get_graph()
1129
 
        heads = graph.heads(parent_ids)
1130
 
        accepted_revisions = set()
1131
 
 
1132
1036
        # convert absent trees to the null tree, which we convert back to
1133
1037
        # missing on access.
1134
1038
        for rev_id, tree in parents_list:
1135
 
            if len(accepted_revisions) > 0:
1136
 
                # we always accept the first tree
1137
 
                if rev_id in accepted_revisions or rev_id not in heads:
1138
 
                    # We have already included either this tree, or its
1139
 
                    # descendent, so we skip it.
1140
 
                    continue
1141
 
            _mod_revision.check_not_reserved_id(rev_id)
 
1039
            rev_id = osutils.safe_revision_id(rev_id)
1142
1040
            if tree is not None:
1143
1041
                real_trees.append((rev_id, tree))
1144
1042
            else:
1145
1043
                real_trees.append((rev_id,
1146
 
                    self.branch.repository.revision_tree(
1147
 
                        _mod_revision.NULL_REVISION)))
 
1044
                    self.branch.repository.revision_tree(None)))
1148
1045
                ghosts.append(rev_id)
1149
 
            accepted_revisions.add(rev_id)
1150
 
        updated = False
1151
 
        if (len(real_trees) == 1
1152
 
            and not ghosts
1153
 
            and self.branch.repository._format.fast_deltas
1154
 
            and isinstance(real_trees[0][1],
1155
 
                revisiontree.InventoryRevisionTree)
1156
 
            and self.get_parent_ids()):
1157
 
            rev_id, rev_tree = real_trees[0]
1158
 
            basis_id = self.get_parent_ids()[0]
1159
 
            # There are times when basis_tree won't be in
1160
 
            # self.branch.repository, (switch, for example)
1161
 
            try:
1162
 
                basis_tree = self.branch.repository.revision_tree(basis_id)
1163
 
            except errors.NoSuchRevision:
1164
 
                # Fall back to the set_parent_trees(), since we can't use
1165
 
                # _make_delta if we can't get the RevisionTree
1166
 
                pass
1167
 
            else:
1168
 
                delta = rev_tree.root_inventory._make_delta(
1169
 
                    basis_tree.root_inventory)
1170
 
                dirstate.update_basis_by_delta(delta, rev_id)
1171
 
                updated = True
1172
 
        if not updated:
1173
 
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1046
        dirstate._validate()
 
1047
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1048
        dirstate._validate()
1174
1049
        self._make_dirty(reset_inventory=False)
 
1050
        dirstate._validate()
1175
1051
 
1176
1052
    def _set_root_id(self, file_id):
1177
1053
        """See WorkingTree.set_root_id."""
1180
1056
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1181
1057
            self._make_dirty(reset_inventory=True)
1182
1058
 
1183
 
    def _sha_from_stat(self, path, stat_result):
1184
 
        """Get a sha digest from the tree's stat cache.
1185
 
 
1186
 
        The default implementation assumes no stat cache is present.
1187
 
 
1188
 
        :param path: The path.
1189
 
        :param stat_result: The stat result being looked up.
1190
 
        """
1191
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1192
 
 
1193
1059
    @needs_read_lock
1194
1060
    def supports_tree_reference(self):
1195
1061
        return self._repo_supports_tree_reference
1197
1063
    def unlock(self):
1198
1064
        """Unlock in format 4 trees needs to write the entire dirstate."""
1199
1065
        if self._control_files._lock_count == 1:
1200
 
            # do non-implementation specific cleanup
1201
 
            self._cleanup()
1202
 
 
1203
1066
            # eventually we should do signature checking during read locks for
1204
1067
            # dirstate updates.
1205
1068
            if self._control_files._lock_mode == 'w':
1235
1098
            return
1236
1099
        state = self.current_dirstate()
1237
1100
        state._read_dirblocks_if_needed()
1238
 
        ids_to_unversion = set(file_ids)
 
1101
        ids_to_unversion = set()
 
1102
        for file_id in file_ids:
 
1103
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1239
1104
        paths_to_unversion = set()
1240
1105
        # sketch:
1241
1106
        # check if the root is to be unversioned, if so, assert for now.
1268
1133
                # just forget the whole block.
1269
1134
                entry_index = 0
1270
1135
                while entry_index < len(block[1]):
1271
 
                    entry = block[1][entry_index]
1272
 
                    if entry[1][0][0] in 'ar':
1273
 
                        # don't remove absent or renamed entries
 
1136
                    # Mark this file id as having been removed
 
1137
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1138
                    if not state._make_absent(block[1][entry_index]):
1274
1139
                        entry_index += 1
1275
 
                    else:
1276
 
                        # Mark this file id as having been removed
1277
 
                        ids_to_unversion.discard(entry[0][2])
1278
 
                        if not state._make_absent(entry):
1279
 
                            # The block has not shrunk.
1280
 
                            entry_index += 1
1281
1140
                # go to the next block. (At the moment we dont delete empty
1282
1141
                # dirblocks)
1283
1142
                block_index += 1
1304
1163
        # have to change the legacy inventory too.
1305
1164
        if self._inventory is not None:
1306
1165
            for file_id in file_ids:
1307
 
                if self._inventory.has_id(file_id):
1308
 
                    self._inventory.remove_recursive_id(file_id)
1309
 
 
1310
 
    @needs_tree_write_lock
1311
 
    def rename_one(self, from_rel, to_rel, after=False):
1312
 
        """See WorkingTree.rename_one"""
1313
 
        self.flush()
1314
 
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
1315
 
 
1316
 
    @needs_tree_write_lock
1317
 
    def apply_inventory_delta(self, changes):
1318
 
        """See MutableTree.apply_inventory_delta"""
1319
 
        state = self.current_dirstate()
1320
 
        state.update_by_delta(changes)
1321
 
        self._make_dirty(reset_inventory=True)
1322
 
 
1323
 
    def update_basis_by_delta(self, new_revid, delta):
1324
 
        """See MutableTree.update_basis_by_delta."""
1325
 
        if self.last_revision() == new_revid:
1326
 
            raise AssertionError()
1327
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1328
 
 
1329
 
    @needs_read_lock
1330
 
    def _validate(self):
1331
 
        self._dirstate._validate()
 
1166
                self._inventory.remove_recursive_id(file_id)
1332
1167
 
1333
1168
    @needs_tree_write_lock
1334
1169
    def _write_inventory(self, inv):
1335
1170
        """Write inventory as the current inventory."""
1336
 
        if self._dirty:
1337
 
            raise AssertionError("attempting to write an inventory when the "
1338
 
                "dirstate is dirty will lose pending changes")
1339
 
        had_inventory = self._inventory is not None
1340
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1341
 
        # working inventory. We do this because self.inventory may be inv, or
1342
 
        # may have been modified, and either case would prevent a clean delta
1343
 
        # being created.
1344
 
        self._inventory = None
1345
 
        # generate a delta,
1346
 
        delta = inv._make_delta(self.root_inventory)
1347
 
        # and apply it.
1348
 
        self.apply_inventory_delta(delta)
1349
 
        if had_inventory:
 
1171
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1172
        self.current_dirstate().set_state_from_inventory(inv)
 
1173
        self._make_dirty(reset_inventory=False)
 
1174
        if self._inventory is not None:
1350
1175
            self._inventory = inv
1351
1176
        self.flush()
1352
1177
 
1353
 
    @needs_tree_write_lock
1354
 
    def reset_state(self, revision_ids=None):
1355
 
        """Reset the state of the working tree.
1356
 
 
1357
 
        This does a hard-reset to a last-known-good state. This is a way to
1358
 
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
1359
 
        """
1360
 
        if revision_ids is None:
1361
 
            revision_ids = self.get_parent_ids()
1362
 
        if not revision_ids:
1363
 
            base_tree = self.branch.repository.revision_tree(
1364
 
                _mod_revision.NULL_REVISION)
1365
 
            trees = []
1366
 
        else:
1367
 
            trees = zip(revision_ids,
1368
 
                        self.branch.repository.revision_trees(revision_ids))
1369
 
            base_tree = trees[0][1]
1370
 
        state = self.current_dirstate()
1371
 
        # We don't support ghosts yet
1372
 
        state.set_state_from_scratch(base_tree.root_inventory, trees, [])
1373
 
 
1374
 
 
1375
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1376
 
 
1377
 
    def __init__(self, tree):
1378
 
        self.tree = tree
1379
 
 
1380
 
    def sha1(self, abspath):
1381
 
        """See dirstate.SHA1Provider.sha1()."""
1382
 
        filters = self.tree._content_filter_stack(
1383
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1384
 
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
1385
 
 
1386
 
    def stat_and_sha1(self, abspath):
1387
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1388
 
        filters = self.tree._content_filter_stack(
1389
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1390
 
        file_obj = file(abspath, 'rb', 65000)
1391
 
        try:
1392
 
            statvalue = os.fstat(file_obj.fileno())
1393
 
            if filters:
1394
 
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
1395
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1396
 
        finally:
1397
 
            file_obj.close()
1398
 
        return statvalue, sha1
1399
 
 
1400
 
 
1401
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1402
 
    """Dirstate working tree that supports content filtering.
1403
 
 
1404
 
    The dirstate holds the hash and size of the canonical form of the file, 
1405
 
    and most methods must return that.
1406
 
    """
1407
 
 
1408
 
    def _file_content_summary(self, path, stat_result):
1409
 
        # This is to support the somewhat obsolete path_content_summary method
1410
 
        # with content filtering: see
1411
 
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
1412
 
        #
1413
 
        # If the dirstate cache is up to date and knows the hash and size,
1414
 
        # return that.
1415
 
        # Otherwise if there are no content filters, return the on-disk size
1416
 
        # and leave the hash blank.
1417
 
        # Otherwise, read and filter the on-disk file and use its size and
1418
 
        # hash.
1419
 
        #
1420
 
        # The dirstate doesn't store the size of the canonical form so we
1421
 
        # can't trust it for content-filtered trees.  We just return None.
1422
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1423
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1424
 
        return ('file', None, executable, dirstate_sha1)
1425
 
 
1426
 
 
1427
 
class WorkingTree4(DirStateWorkingTree):
1428
 
    """This is the Format 4 working tree.
1429
 
 
1430
 
    This differs from WorkingTree by:
1431
 
     - Having a consolidated internal dirstate, stored in a
1432
 
       randomly-accessible sorted file on disk.
1433
 
     - Not having a regular inventory attribute.  One can be synthesized
1434
 
       on demand but this is expensive and should be avoided.
1435
 
 
1436
 
    This is new in bzr 0.15.
1437
 
    """
1438
 
 
1439
 
 
1440
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1441
 
    """This is the Format 5 working tree.
1442
 
 
1443
 
    This differs from WorkingTree4 by:
1444
 
     - Supporting content filtering.
1445
 
 
1446
 
    This is new in bzr 1.11.
1447
 
    """
1448
 
 
1449
 
 
1450
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1451
 
    """This is the Format 6 working tree.
1452
 
 
1453
 
    This differs from WorkingTree5 by:
1454
 
     - Supporting a current view that may mask the set of files in a tree
1455
 
       impacted by most user operations.
1456
 
 
1457
 
    This is new in bzr 1.14.
1458
 
    """
1459
 
 
1460
 
    def _make_views(self):
1461
 
        return views.PathBasedViews(self)
1462
 
 
1463
 
 
1464
 
class DirStateWorkingTreeFormat(WorkingTreeFormatMetaDir):
1465
 
 
1466
 
    missing_parent_conflicts = True
1467
 
 
1468
 
    supports_versioned_directories = True
1469
 
 
1470
 
    _lock_class = LockDir
1471
 
    _lock_file_name = 'lock'
1472
 
 
1473
 
    def _open_control_files(self, a_bzrdir):
1474
 
        transport = a_bzrdir.get_workingtree_transport(None)
1475
 
        return LockableFiles(transport, self._lock_file_name,
1476
 
                             self._lock_class)
1477
 
 
1478
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1479
 
                   accelerator_tree=None, hardlink=False):
 
1178
 
 
1179
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1180
    """The first consolidated dirstate working tree format.
 
1181
 
 
1182
    This format:
 
1183
        - exists within a metadir controlling .bzr
 
1184
        - includes an explicit version marker for the workingtree control
 
1185
          files, separate from the BzrDir format
 
1186
        - modifies the hash cache format
 
1187
        - is new in bzr 0.15
 
1188
        - uses a LockDir to guard access to it.
 
1189
    """
 
1190
 
 
1191
    def get_format_string(self):
 
1192
        """See WorkingTreeFormat.get_format_string()."""
 
1193
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1194
 
 
1195
    def get_format_description(self):
 
1196
        """See WorkingTreeFormat.get_format_description()."""
 
1197
        return "Working tree format 4"
 
1198
 
 
1199
    def initialize(self, a_bzrdir, revision_id=None):
1480
1200
        """See WorkingTreeFormat.initialize().
1481
1201
 
1482
1202
        :param revision_id: allows creating a working tree at a different
1483
 
            revision than the branch is at.
1484
 
        :param accelerator_tree: A tree which can be used for retrieving file
1485
 
            contents more quickly than the revision tree, i.e. a workingtree.
1486
 
            The revision tree will be used for cases where accelerator_tree's
1487
 
            content is different.
1488
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1489
 
            where possible.
 
1203
        revision than the branch is at.
1490
1204
 
1491
 
        These trees get an initial random root id, if their repository supports
1492
 
        rich root data, TREE_ROOT otherwise.
 
1205
        These trees get an initial random root id.
1493
1206
        """
 
1207
        revision_id = osutils.safe_revision_id(revision_id)
1494
1208
        if not isinstance(a_bzrdir.transport, LocalTransport):
1495
1209
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1496
1210
        transport = a_bzrdir.get_workingtree_transport(self)
1497
1211
        control_files = self._open_control_files(a_bzrdir)
1498
1212
        control_files.create_lock()
1499
1213
        control_files.lock_write()
1500
 
        transport.put_bytes('format', self.as_string(),
1501
 
            mode=a_bzrdir._get_file_mode())
1502
 
        if from_branch is not None:
1503
 
            branch = from_branch
1504
 
        else:
1505
 
            branch = a_bzrdir.open_branch()
 
1214
        control_files.put_utf8('format', self.get_format_string())
 
1215
        branch = a_bzrdir.open_branch()
1506
1216
        if revision_id is None:
1507
1217
            revision_id = branch.last_revision()
1508
1218
        local_path = transport.local_abspath('dirstate')
1509
1219
        # write out new dirstate (must exist when we create the tree)
1510
1220
        state = dirstate.DirState.initialize(local_path)
1511
1221
        state.unlock()
1512
 
        del state
1513
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1222
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1514
1223
                         branch,
1515
1224
                         _format=self,
1516
1225
                         _bzrdir=a_bzrdir,
1517
1226
                         _control_files=control_files)
1518
1227
        wt._new_tree()
1519
1228
        wt.lock_tree_write()
 
1229
        state._validate()
1520
1230
        try:
1521
 
            self._init_custom_control_files(wt)
1522
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
1523
 
                if branch.repository.supports_rich_root():
1524
 
                    wt._set_root_id(generate_ids.gen_root_id())
1525
 
                else:
1526
 
                    wt._set_root_id(ROOT_ID)
 
1231
            if revision_id in (None, NULL_REVISION):
 
1232
                wt._set_root_id(generate_ids.gen_root_id())
1527
1233
                wt.flush()
1528
 
            basis = None
1529
 
            # frequently, we will get here due to branching.  The accelerator
1530
 
            # tree will be the tree from the branch, so the desired basis
1531
 
            # tree will often be a parent of the accelerator tree.
1532
 
            if accelerator_tree is not None:
1533
 
                try:
1534
 
                    basis = accelerator_tree.revision_tree(revision_id)
1535
 
                except errors.NoSuchRevision:
1536
 
                    pass
1537
 
            if basis is None:
1538
 
                basis = branch.repository.revision_tree(revision_id)
1539
 
            if revision_id == _mod_revision.NULL_REVISION:
1540
 
                parents_list = []
1541
 
            else:
1542
 
                parents_list = [(revision_id, basis)]
 
1234
                wt.current_dirstate()._validate()
 
1235
            wt.set_last_revision(revision_id)
 
1236
            wt.flush()
 
1237
            basis = wt.basis_tree()
1543
1238
            basis.lock_read()
1544
 
            try:
1545
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1239
            # if the basis has a root id we have to use that; otherwise we use
 
1240
            # a new random one
 
1241
            basis_root_id = basis.get_root_id()
 
1242
            if basis_root_id is not None:
 
1243
                wt._set_root_id(basis_root_id)
1546
1244
                wt.flush()
1547
 
                # if the basis has a root id we have to use that; otherwise we
1548
 
                # use a new random one
1549
 
                basis_root_id = basis.get_root_id()
1550
 
                if basis_root_id is not None:
1551
 
                    wt._set_root_id(basis_root_id)
1552
 
                    wt.flush()
1553
 
                if wt.supports_content_filtering():
1554
 
                    # The original tree may not have the same content filters
1555
 
                    # applied so we can't safely build the inventory delta from
1556
 
                    # the source tree.
1557
 
                    delta_from_tree = False
1558
 
                else:
1559
 
                    delta_from_tree = True
1560
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1561
 
                # because wt4.apply_inventory_delta does not mutate the input
1562
 
                # inventory entries.
1563
 
                transform.build_tree(basis, wt, accelerator_tree,
1564
 
                                     hardlink=hardlink,
1565
 
                                     delta_from_tree=delta_from_tree)
1566
 
                for hook in MutableTree.hooks['post_build_tree']:
1567
 
                    hook(wt)
1568
 
            finally:
1569
 
                basis.unlock()
 
1245
            transform.build_tree(basis, wt)
 
1246
            basis.unlock()
1570
1247
        finally:
1571
1248
            control_files.unlock()
1572
1249
            wt.unlock()
1573
1250
        return wt
1574
1251
 
1575
 
    def _init_custom_control_files(self, wt):
1576
 
        """Subclasses with custom control files should override this method.
1577
 
 
1578
 
        The working tree and control files are locked for writing when this
1579
 
        method is called.
1580
 
 
1581
 
        :param wt: the WorkingTree object
1582
 
        """
1583
 
 
1584
 
    def open(self, a_bzrdir, _found=False):
1585
 
        """Return the WorkingTree object for a_bzrdir
1586
 
 
1587
 
        _found is a private parameter, do not use it. It is used to indicate
1588
 
               if format probing has already been done.
1589
 
        """
1590
 
        if not _found:
1591
 
            # we are being called directly and must probe.
1592
 
            raise NotImplementedError
1593
 
        if not isinstance(a_bzrdir.transport, LocalTransport):
1594
 
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1595
 
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
1596
 
        return wt
1597
 
 
1598
1252
    def _open(self, a_bzrdir, control_files):
1599
1253
        """Open the tree itself.
1600
1254
 
1601
1255
        :param a_bzrdir: the dir for the tree.
1602
1256
        :param control_files: the control files for the tree.
1603
1257
        """
1604
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1258
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1605
1259
                           branch=a_bzrdir.open_branch(),
1606
1260
                           _format=self,
1607
1261
                           _bzrdir=a_bzrdir,
1608
1262
                           _control_files=control_files)
1609
1263
 
1610
1264
    def __get_matchingbzrdir(self):
1611
 
        return self._get_matchingbzrdir()
1612
 
 
1613
 
    def _get_matchingbzrdir(self):
1614
 
        """Overrideable method to get a bzrdir for testing."""
1615
1265
        # please test against something that will let us do tree references
1616
 
        return controldir.format_registry.make_bzrdir(
1617
 
            'development-subtree')
 
1266
        return bzrdir.format_registry.make_bzrdir(
 
1267
            'dirstate-with-subtree')
1618
1268
 
1619
1269
    _matchingbzrdir = property(__get_matchingbzrdir)
1620
1270
 
1621
1271
 
1622
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1623
 
    """The first consolidated dirstate working tree format.
1624
 
 
1625
 
    This format:
1626
 
        - exists within a metadir controlling .bzr
1627
 
        - includes an explicit version marker for the workingtree control
1628
 
          files, separate from the ControlDir format
1629
 
        - modifies the hash cache format
1630
 
        - is new in bzr 0.15
1631
 
        - uses a LockDir to guard access to it.
1632
 
    """
1633
 
 
1634
 
    upgrade_recommended = False
1635
 
 
1636
 
    _tree_class = WorkingTree4
1637
 
 
1638
 
    @classmethod
1639
 
    def get_format_string(cls):
1640
 
        """See WorkingTreeFormat.get_format_string()."""
1641
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1642
 
 
1643
 
    def get_format_description(self):
1644
 
        """See WorkingTreeFormat.get_format_description()."""
1645
 
        return "Working tree format 4"
1646
 
 
1647
 
 
1648
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1649
 
    """WorkingTree format supporting content filtering.
1650
 
    """
1651
 
 
1652
 
    upgrade_recommended = False
1653
 
 
1654
 
    _tree_class = WorkingTree5
1655
 
 
1656
 
    @classmethod
1657
 
    def get_format_string(cls):
1658
 
        """See WorkingTreeFormat.get_format_string()."""
1659
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1660
 
 
1661
 
    def get_format_description(self):
1662
 
        """See WorkingTreeFormat.get_format_description()."""
1663
 
        return "Working tree format 5"
1664
 
 
1665
 
    def supports_content_filtering(self):
1666
 
        return True
1667
 
 
1668
 
 
1669
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1670
 
    """WorkingTree format supporting views.
1671
 
    """
1672
 
 
1673
 
    upgrade_recommended = False
1674
 
 
1675
 
    _tree_class = WorkingTree6
1676
 
 
1677
 
    @classmethod
1678
 
    def get_format_string(cls):
1679
 
        """See WorkingTreeFormat.get_format_string()."""
1680
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1681
 
 
1682
 
    def get_format_description(self):
1683
 
        """See WorkingTreeFormat.get_format_description()."""
1684
 
        return "Working tree format 6"
1685
 
 
1686
 
    def _init_custom_control_files(self, wt):
1687
 
        """Subclasses with custom control files should override this method."""
1688
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1689
 
 
1690
 
    def supports_content_filtering(self):
1691
 
        return True
1692
 
 
1693
 
    def supports_views(self):
1694
 
        return True
1695
 
 
1696
 
    def _get_matchingbzrdir(self):
1697
 
        """Overrideable method to get a bzrdir for testing."""
1698
 
        # We use 'development-subtree' instead of '2a', because we have a
1699
 
        # few tests that want to test tree references
1700
 
        return bzrdir.format_registry.make_bzrdir('development-subtree')
1701
 
 
1702
 
 
1703
 
class DirStateRevisionTree(InventoryTree):
1704
 
    """A revision tree pulling the inventory from a dirstate.
1705
 
    
1706
 
    Note that this is one of the historical (ie revision) trees cached in the
1707
 
    dirstate for easy access, not the workingtree.
1708
 
    """
 
1272
class DirStateRevisionTree(Tree):
 
1273
    """A revision tree pulling the inventory from a dirstate."""
1709
1274
 
1710
1275
    def __init__(self, dirstate, revision_id, repository):
1711
1276
        self._dirstate = dirstate
1712
 
        self._revision_id = revision_id
 
1277
        self._revision_id = osutils.safe_revision_id(revision_id)
1713
1278
        self._repository = repository
1714
1279
        self._inventory = None
1715
1280
        self._locked = 0
1716
1281
        self._dirstate_locked = False
1717
 
        self._repo_supports_tree_reference = getattr(
1718
 
            repository._format, "supports_tree_reference",
1719
 
            False)
1720
1282
 
1721
1283
    def __repr__(self):
1722
1284
        return "<%s of %s in %s>" % \
1723
1285
            (self.__class__.__name__, self._revision_id, self._dirstate)
1724
1286
 
1725
 
    def annotate_iter(self, file_id,
1726
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1287
    def annotate_iter(self, file_id):
1727
1288
        """See Tree.annotate_iter"""
1728
 
        text_key = (file_id, self.get_file_revision(file_id))
1729
 
        annotations = self._repository.texts.annotate(text_key)
1730
 
        return [(key[-1], line) for (key, line) in annotations]
 
1289
        w = self._repository.weave_store.get_weave(file_id,
 
1290
                           self._repository.get_transaction())
 
1291
        return w.annotate_iter(self.inventory[file_id].revision)
1731
1292
 
1732
1293
    def _comparison_data(self, entry, path):
1733
1294
        """See Tree._comparison_data."""
1751
1312
    def get_root_id(self):
1752
1313
        return self.path2id('')
1753
1314
 
1754
 
    def id2path(self, file_id):
1755
 
        "Convert a file-id to a path."
1756
 
        entry = self._get_entry(file_id=file_id)
1757
 
        if entry == (None, None):
1758
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1759
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1760
 
        return path_utf8.decode('utf8')
1761
 
 
1762
 
    def iter_references(self):
1763
 
        if not self._repo_supports_tree_reference:
1764
 
            # When the repo doesn't support references, we will have nothing to
1765
 
            # return
1766
 
            return iter([])
1767
 
        # Otherwise, fall back to the default implementation
1768
 
        return super(DirStateRevisionTree, self).iter_references()
1769
 
 
1770
1315
    def _get_parent_index(self):
1771
1316
        """Return the index in the dirstate referenced by this tree."""
1772
1317
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1777
1322
        If either file_id or path is supplied, it is used as the key to lookup.
1778
1323
        If both are supplied, the fastest lookup is used, and an error is
1779
1324
        raised if they do not both point at the same row.
1780
 
 
 
1325
        
1781
1326
        :param file_id: An optional unicode file_id to be looked up.
1782
1327
        :param path: An optional unicode path to be looked up.
1783
1328
        :return: The dirstate row tuple for path/file_id, or (None, None)
1784
1329
        """
1785
1330
        if file_id is None and path is None:
1786
1331
            raise errors.BzrError('must supply file_id or path')
 
1332
        file_id = osutils.safe_file_id(file_id)
1787
1333
        if path is not None:
1788
1334
            path = path.encode('utf8')
1789
1335
        parent_index = self._get_parent_index()
1790
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
1791
 
            path_utf8=path)
 
1336
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1792
1337
 
1793
1338
    def _generate_inventory(self):
1794
1339
        """Create and set self.inventory from the dirstate object.
1798
1343
 
1799
1344
        This is relatively expensive: we have to walk the entire dirstate.
1800
1345
        """
1801
 
        if not self._locked:
1802
 
            raise AssertionError(
1803
 
                'cannot generate inventory of an unlocked '
1804
 
                'dirstate revision tree')
 
1346
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1347
            'dirstate revision tree'
1805
1348
        # separate call for profiling - makes it clear where the costs are.
1806
1349
        self._dirstate._read_dirblocks_if_needed()
1807
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1808
 
            raise AssertionError(
1809
 
                'parent %s has disappeared from %s' % (
1810
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1350
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1351
            'parent %s has disappeared from %s' % (
 
1352
            self._revision_id, self._dirstate.get_parent_ids())
1811
1353
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1812
1354
        # This is identical now to the WorkingTree _generate_inventory except
1813
1355
        # for the tree index use.
1814
1356
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1815
1357
        current_id = root_key[2]
1816
 
        if current_entry[parent_index][0] != 'd':
1817
 
            raise AssertionError()
 
1358
        assert current_entry[parent_index][0] == 'd'
1818
1359
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1819
1360
        inv.root.revision = current_entry[parent_index][4]
1820
1361
        # Turn some things into local variables
1851
1392
                elif kind == 'directory':
1852
1393
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1853
1394
                elif kind == 'symlink':
 
1395
                    inv_entry.executable = False
 
1396
                    inv_entry.text_size = size
1854
1397
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1855
1398
                elif kind == 'tree-reference':
1856
1399
                    inv_entry.reference_revision = fingerprint or None
1858
1401
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1859
1402
                            % entry)
1860
1403
                # These checks cost us around 40ms on a 55k entry tree
1861
 
                if file_id in inv_byid:
1862
 
                    raise AssertionError('file_id %s already in'
1863
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1864
 
                if name_unicode in parent_ie.children:
1865
 
                    raise AssertionError('name %r already in parent'
1866
 
                        % (name_unicode,))
 
1404
                assert file_id not in inv_byid
 
1405
                assert name_unicode not in parent_ie.children
1867
1406
                inv_byid[file_id] = inv_entry
1868
1407
                parent_ie.children[name_unicode] = inv_entry
1869
1408
        self._inventory = inv
1876
1415
        # Make sure the file exists
1877
1416
        entry = self._get_entry(file_id, path=path)
1878
1417
        if entry == (None, None): # do we raise?
1879
 
            raise errors.NoSuchId(self, file_id)
 
1418
            return None
1880
1419
        parent_index = self._get_parent_index()
1881
1420
        last_changed_revision = entry[1][parent_index][4]
1882
 
        try:
1883
 
            rev = self._repository.get_revision(last_changed_revision)
1884
 
        except errors.NoSuchRevision:
1885
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1886
 
        return rev.timestamp
 
1421
        return self._repository.get_revision(last_changed_revision).timestamp
1887
1422
 
1888
1423
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1889
1424
        entry = self._get_entry(file_id=file_id, path=path)
1893
1428
            return parent_details[1]
1894
1429
        return None
1895
1430
 
1896
 
    @needs_read_lock
1897
 
    def get_file_revision(self, file_id):
1898
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1899
 
        return inv[inv_file_id].revision
1900
 
 
1901
 
    def get_file(self, file_id, path=None):
 
1431
    def get_file(self, file_id):
1902
1432
        return StringIO(self.get_file_text(file_id))
1903
1433
 
 
1434
    def get_file_lines(self, file_id):
 
1435
        ie = self.inventory[file_id]
 
1436
        return self._repository.weave_store.get_weave(file_id,
 
1437
                self._repository.get_transaction()).get_lines(ie.revision)
 
1438
 
1904
1439
    def get_file_size(self, file_id):
1905
 
        """See Tree.get_file_size"""
1906
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1907
 
        return inv[inv_file_id].text_size
 
1440
        return self.inventory[file_id].text_size
1908
1441
 
1909
 
    def get_file_text(self, file_id, path=None):
1910
 
        content = None
1911
 
        for _, content_iter in self.iter_files_bytes([(file_id, None)]):
1912
 
            if content is not None:
1913
 
                raise AssertionError('iter_files_bytes returned'
1914
 
                    ' too many entries')
1915
 
            # For each entry returned by iter_files_bytes, we must consume the
1916
 
            # content_iter before we step the files iterator.
1917
 
            content = ''.join(content_iter)
1918
 
        if content is None:
1919
 
            raise AssertionError('iter_files_bytes did not return'
1920
 
                ' the requested data')
1921
 
        return content
 
1442
    def get_file_text(self, file_id):
 
1443
        return ''.join(self.get_file_lines(file_id))
1922
1444
 
1923
1445
    def get_reference_revision(self, file_id, path=None):
1924
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1925
 
        return inv[inv_file_id].reference_revision
1926
 
 
1927
 
    def iter_files_bytes(self, desired_files):
1928
 
        """See Tree.iter_files_bytes.
1929
 
 
1930
 
        This version is implemented on top of Repository.iter_files_bytes"""
1931
 
        parent_index = self._get_parent_index()
1932
 
        repo_desired_files = []
1933
 
        for file_id, identifier in desired_files:
1934
 
            entry = self._get_entry(file_id)
1935
 
            if entry == (None, None):
1936
 
                raise errors.NoSuchId(self, file_id)
1937
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1938
 
                                       identifier))
1939
 
        return self._repository.iter_files_bytes(repo_desired_files)
1940
 
 
1941
 
    def get_symlink_target(self, file_id, path=None):
 
1446
        return self.inventory[file_id].reference_revision
 
1447
 
 
1448
    def get_symlink_target(self, file_id):
1942
1449
        entry = self._get_entry(file_id=file_id)
1943
1450
        parent_index = self._get_parent_index()
1944
1451
        if entry[1][parent_index][0] != 'l':
1945
1452
            return None
1946
1453
        else:
1947
 
            target = entry[1][parent_index][1]
1948
 
            target = target.decode('utf8')
1949
 
            return target
 
1454
            # At present, none of the tree implementations supports non-ascii
 
1455
            # symlink targets. So we will just assume that the dirstate path is
 
1456
            # correct.
 
1457
            return entry[1][parent_index][1]
1950
1458
 
1951
1459
    def get_revision_id(self):
1952
1460
        """Return the revision id for this tree."""
1953
1461
        return self._revision_id
1954
1462
 
1955
 
    def _get_root_inventory(self):
 
1463
    def _get_inventory(self):
1956
1464
        if self._inventory is not None:
1957
1465
            return self._inventory
1958
1466
        self._must_be_locked()
1959
1467
        self._generate_inventory()
1960
1468
        return self._inventory
1961
1469
 
1962
 
    root_inventory = property(_get_root_inventory,
1963
 
                         doc="Inventory of this Tree")
1964
 
 
1965
 
    @deprecated_method(deprecated_in((2, 5, 0)))
1966
 
    def _get_inventory(self):
1967
 
        return self.root_inventory
1968
 
 
1969
1470
    inventory = property(_get_inventory,
1970
1471
                         doc="Inventory of this Tree")
1971
1472
 
1977
1478
        return bool(self.path2id(filename))
1978
1479
 
1979
1480
    def kind(self, file_id):
1980
 
        entry = self._get_entry(file_id=file_id)[1]
1981
 
        if entry is None:
1982
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1983
 
        parent_index = self._get_parent_index()
1984
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1985
 
 
1986
 
    def stored_kind(self, file_id):
1987
 
        """See Tree.stored_kind"""
1988
 
        return self.kind(file_id)
1989
 
 
1990
 
    def path_content_summary(self, path):
1991
 
        """See Tree.path_content_summary."""
1992
 
        inv, inv_file_id = self._path2inv_file_id(path)
1993
 
        if inv_file_id is None:
1994
 
            return ('missing', None, None, None)
1995
 
        entry = inv[inv_file_id]
1996
 
        kind = entry.kind
1997
 
        if kind == 'file':
1998
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1999
 
        elif kind == 'symlink':
2000
 
            return (kind, None, None, entry.symlink_target)
2001
 
        else:
2002
 
            return (kind, None, None, None)
 
1481
        return self.inventory[file_id].kind
2003
1482
 
2004
1483
    def is_executable(self, file_id, path=None):
2005
 
        inv, inv_file_id = self._unpack_file_id(file_id)
2006
 
        ie = inv[inv_file_id]
 
1484
        ie = self.inventory[file_id]
2007
1485
        if ie.kind != "file":
2008
 
            return False
 
1486
            return None
2009
1487
        return ie.executable
2010
1488
 
2011
 
    def is_locked(self):
2012
 
        return self._locked
2013
 
 
2014
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1489
    def list_files(self, include_root=False):
2015
1490
        # We use a standard implementation, because DirStateRevisionTree is
2016
1491
        # dealing with one of the parents of the current state
2017
 
        if from_dir is None:
2018
 
            inv = self.root_inventory
2019
 
            from_dir_id = None
2020
 
        else:
2021
 
            inv, from_dir_id = self._path2inv_file_id(from_dir)
2022
 
            if from_dir_id is None:
2023
 
                # Directory not versioned
2024
 
                return
2025
 
        # FIXME: Support nested trees
2026
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
2027
 
        if inv.root is not None and not include_root and from_dir is None:
 
1492
        inv = self._get_inventory()
 
1493
        entries = inv.iter_entries()
 
1494
        if self.inventory.root is not None and not include_root:
2028
1495
            entries.next()
2029
1496
        for path, entry in entries:
2030
1497
            yield path, 'V', entry.kind, entry.file_id, entry
2031
1498
 
2032
1499
    def lock_read(self):
2033
 
        """Lock the tree for a set of operations.
2034
 
 
2035
 
        :return: A bzrlib.lock.LogicalLockResult.
2036
 
        """
 
1500
        """Lock the tree for a set of operations."""
2037
1501
        if not self._locked:
2038
1502
            self._repository.lock_read()
2039
1503
            if self._dirstate._lock_token is None:
2040
1504
                self._dirstate.lock_read()
2041
1505
                self._dirstate_locked = True
2042
1506
        self._locked += 1
2043
 
        return LogicalLockResult(self.unlock)
2044
1507
 
2045
1508
    def _must_be_locked(self):
2046
1509
        if not self._locked:
2050
1513
    def path2id(self, path):
2051
1514
        """Return the id for path in this tree."""
2052
1515
        # lookup by path: faster than splitting and walking the ivnentory.
2053
 
        if isinstance(path, list):
2054
 
            if path == []:
2055
 
                path = [""]
2056
 
            path = osutils.pathjoin(*path)
2057
1516
        entry = self._get_entry(path=path)
2058
1517
        if entry == (None, None):
2059
1518
            return None
2071
1530
                self._dirstate_locked = False
2072
1531
            self._repository.unlock()
2073
1532
 
2074
 
    @needs_read_lock
2075
 
    def supports_tree_reference(self):
2076
 
        return self._repo_supports_tree_reference
2077
 
 
2078
1533
    def walkdirs(self, prefix=""):
2079
1534
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2080
 
        # implementation based on an inventory.
 
1535
        # implementation based on an inventory.  
2081
1536
        # This should be cleaned up to use the much faster Dirstate code
2082
1537
        # So for now, we just build up the parent inventory, and extract
2083
1538
        # it the same way RevisionTree does.
2084
1539
        _directory = 'directory'
2085
 
        inv = self._get_root_inventory()
 
1540
        inv = self._get_inventory()
2086
1541
        top_id = inv.path2id(prefix)
2087
1542
        if top_id is None:
2088
1543
            pending = []
2112
1567
 
2113
1568
class InterDirStateTree(InterTree):
2114
1569
    """Fast path optimiser for changes_from with dirstate trees.
2115
 
 
2116
 
    This is used only when both trees are in the dirstate working file, and
2117
 
    the source is any parent within the dirstate, and the destination is
 
1570
    
 
1571
    This is used only when both trees are in the dirstate working file, and 
 
1572
    the source is any parent within the dirstate, and the destination is 
2118
1573
    the current working tree of the same dirstate.
2119
1574
    """
2120
1575
    # this could be generalized to allow comparisons between any trees in the
2129
1584
    def make_source_parent_tree(source, target):
2130
1585
        """Change the source tree into a parent of the target."""
2131
1586
        revid = source.commit('record tree')
2132
 
        target.branch.fetch(source.branch, revid)
 
1587
        target.branch.repository.fetch(source.branch.repository, revid)
2133
1588
        target.set_parent_ids([revid])
2134
1589
        return target.basis_tree(), target
2135
1590
 
2136
 
    @classmethod
2137
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
2138
 
        result = klass.make_source_parent_tree(source, target)
2139
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2140
 
        return result
2141
 
 
2142
 
    @classmethod
2143
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2144
 
                                                  target):
2145
 
        from bzrlib.tests.test__dirstate_helpers import \
2146
 
            compiled_dirstate_helpers_feature
2147
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2148
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2149
 
        result = klass.make_source_parent_tree(source, target)
2150
 
        result[1]._iter_changes = ProcessEntryC
2151
 
        return result
2152
 
 
2153
1591
    _matching_from_tree_format = WorkingTreeFormat4()
2154
1592
    _matching_to_tree_format = WorkingTreeFormat4()
2155
 
 
2156
 
    @classmethod
2157
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2158
 
        # This method shouldn't be called, because we have python and C
2159
 
        # specific flavours.
2160
 
        raise NotImplementedError
2161
 
 
2162
 
    def iter_changes(self, include_unchanged=False,
 
1593
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1594
 
 
1595
    def _iter_changes(self, include_unchanged=False,
2163
1596
                      specific_files=None, pb=None, extra_trees=[],
2164
1597
                      require_versioned=True, want_unversioned=False):
2165
1598
        """Return the changes from source to target.
2166
1599
 
2167
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1600
        :return: An iterator that yields tuples. See InterTree._iter_changes
2168
1601
            for details.
2169
1602
        :param specific_files: An optional list of file paths to restrict the
2170
1603
            comparison to. When mapping filenames to ids, all matches in all
2181
1614
            output. An unversioned file is defined as one with (False, False)
2182
1615
            for the versioned pair.
2183
1616
        """
 
1617
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1618
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1619
        # NB: show_status depends on being able to pass in non-versioned files
 
1620
        # and report them as unknown
2184
1621
        # TODO: handle extra trees in the dirstate.
2185
 
        if (extra_trees or specific_files == []):
 
1622
        # TODO: handle comparisons as an empty tree as a different special
 
1623
        # case? mbp 20070226
 
1624
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2186
1625
            # we can't fast-path these cases (yet)
2187
 
            return super(InterDirStateTree, self).iter_changes(
 
1626
            for f in super(InterDirStateTree, self)._iter_changes(
2188
1627
                include_unchanged, specific_files, pb, extra_trees,
2189
 
                require_versioned, want_unversioned=want_unversioned)
 
1628
                require_versioned, want_unversioned=want_unversioned):
 
1629
                yield f
 
1630
            return
2190
1631
        parent_ids = self.target.get_parent_ids()
2191
 
        if not (self.source._revision_id in parent_ids
2192
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2193
 
            raise AssertionError(
2194
 
                "revision {%s} is not stored in {%s}, but %s "
2195
 
                "can only be used for trees stored in the dirstate"
2196
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1632
        assert (self.source._revision_id in parent_ids), \
 
1633
                "revision {%s} is not stored in {%s}, but %s " \
 
1634
                "can only be used for trees stored in the dirstate" \
 
1635
                % (self.source._revision_id, self.target, self._iter_changes)
2197
1636
        target_index = 0
2198
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1637
        if self.source._revision_id == NULL_REVISION:
2199
1638
            source_index = None
2200
1639
            indices = (target_index,)
2201
1640
        else:
2202
 
            if not (self.source._revision_id in parent_ids):
2203
 
                raise AssertionError(
2204
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2205
 
                    self.source._revision_id, parent_ids))
 
1641
            assert (self.source._revision_id in parent_ids), \
 
1642
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1643
                self.source._revision_id, parent_ids)
2206
1644
            source_index = 1 + parent_ids.index(self.source._revision_id)
2207
 
            indices = (source_index, target_index)
 
1645
            indices = (source_index,target_index)
2208
1646
        # -- make all specific_files utf8 --
2209
1647
        if specific_files:
2210
1648
            specific_files_utf8 = set()
2211
1649
            for path in specific_files:
2212
 
                # Note, if there are many specific files, using cache_utf8
2213
 
                # would be good here.
2214
1650
                specific_files_utf8.add(path.encode('utf8'))
2215
1651
            specific_files = specific_files_utf8
2216
1652
        else:
2217
1653
            specific_files = set([''])
2218
1654
        # -- specific_files is now a utf8 path set --
2219
 
 
2220
1655
        # -- get the state object and prepare it.
2221
1656
        state = self.target.current_dirstate()
2222
1657
        state._read_dirblocks_if_needed()
 
1658
        def _entries_for_path(path):
 
1659
            """Return a list with all the entries that match path for all ids.
 
1660
            """
 
1661
            dirname, basename = os.path.split(path)
 
1662
            key = (dirname, basename, '')
 
1663
            block_index, present = state._find_block_index_from_key(key)
 
1664
            if not present:
 
1665
                # the block which should contain path is absent.
 
1666
                return []
 
1667
            result = []
 
1668
            block = state._dirblocks[block_index][1]
 
1669
            entry_index, _ = state._find_entry_index(key, block)
 
1670
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1671
            while (entry_index < len(block) and
 
1672
                block[entry_index][0][0:2] == key[0:2]):
 
1673
                result.append(block[entry_index])
 
1674
                entry_index += 1
 
1675
            return result
2223
1676
        if require_versioned:
2224
1677
            # -- check all supplied paths are versioned in a search tree. --
2225
 
            not_versioned = []
 
1678
            all_versioned = True
2226
1679
            for path in specific_files:
2227
 
                path_entries = state._entries_for_path(path)
 
1680
                path_entries = _entries_for_path(path)
2228
1681
                if not path_entries:
2229
1682
                    # this specified path is not present at all: error
2230
 
                    not_versioned.append(path.decode('utf-8'))
2231
 
                    continue
 
1683
                    all_versioned = False
 
1684
                    break
2232
1685
                found_versioned = False
2233
1686
                # for each id at this path
2234
1687
                for entry in path_entries:
2241
1694
                if not found_versioned:
2242
1695
                    # none of the indexes was not 'absent' at all ids for this
2243
1696
                    # path.
2244
 
                    not_versioned.append(path.decode('utf-8'))
2245
 
            if len(not_versioned) > 0:
2246
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1697
                    all_versioned = False
 
1698
                    break
 
1699
            if not all_versioned:
 
1700
                raise errors.PathsNotVersionedError(specific_files)
2247
1701
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2248
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1702
        search_specific_files = set()
 
1703
        for path in specific_files:
 
1704
            other_specific_files = specific_files.difference(set([path]))
 
1705
            if not osutils.is_inside_any(other_specific_files, path):
 
1706
                # this is a top level path, we must check it.
 
1707
                search_specific_files.add(path)
 
1708
        # sketch: 
 
1709
        # compare source_index and target_index at or under each element of search_specific_files.
 
1710
        # follow the following comparison table. Note that we only want to do diff operations when
 
1711
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1712
        # for the target.
 
1713
        # cases:
 
1714
        # 
 
1715
        # Source | Target | disk | action
 
1716
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1717
        #        |        |      | diff check on source-target
 
1718
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1719
        #        |        |      | ???
 
1720
        #   r    |  a     |      | add source to search
 
1721
        #   r    |  a     |  a   | 
 
1722
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1723
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1724
        #   a    | fdlt   |      | add new id
 
1725
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1726
        #   a    |  a     |      | not present in either tree, skip
 
1727
        #   a    |  a     |  a   | not present in any tree, skip
 
1728
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1729
        #        |        |      | may not be selected by the users list of paths.
 
1730
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1731
        #        |        |      | may not be selected by the users list of paths.
 
1732
        #  fdlt  | fdlt   |      | content in both: diff them
 
1733
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1734
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1735
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1736
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1737
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1738
        #        |        |      | this id at the other path.
 
1739
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1740
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1741
        #        |        |      | this id at the other path.
 
1742
 
 
1743
        # for all search_indexs in each path at or under each element of
 
1744
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1745
        # relocated path as one to search if its not searched already. If the
 
1746
        # detail is not relocated, add the id.
 
1747
        searched_specific_files = set()
 
1748
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1749
        # Using a list so that we can access the values and change them in
 
1750
        # nested scope. Each one is [path, file_id, entry]
 
1751
        last_source_parent = [None, None, None]
 
1752
        last_target_parent = [None, None, None]
2249
1753
 
2250
1754
        use_filesystem_for_exec = (sys.platform != 'win32')
2251
 
        iter_changes = self.target._iter_changes(include_unchanged,
2252
 
            use_filesystem_for_exec, search_specific_files, state,
2253
 
            source_index, target_index, want_unversioned, self.target)
2254
 
        return iter_changes.iter_changes()
 
1755
 
 
1756
        def _process_entry(entry, path_info):
 
1757
            """Compare an entry and real disk to generate delta information.
 
1758
 
 
1759
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1760
                the path of entry. If None, then the path is considered absent.
 
1761
                (Perhaps we should pass in a concrete entry for this ?)
 
1762
                Basename is returned as a utf8 string because we expect this
 
1763
                tuple will be ignored, and don't want to take the time to
 
1764
                decode.
 
1765
            """
 
1766
            if source_index is None:
 
1767
                source_details = NULL_PARENT_DETAILS
 
1768
            else:
 
1769
                source_details = entry[1][source_index]
 
1770
            target_details = entry[1][target_index]
 
1771
            target_minikind = target_details[0]
 
1772
            if path_info is not None and target_minikind in 'fdlt':
 
1773
                assert target_index == 0
 
1774
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1775
                                                  stat_value=path_info[3])
 
1776
                # The entry may have been modified by update_entry
 
1777
                target_details = entry[1][target_index]
 
1778
                target_minikind = target_details[0]
 
1779
            else:
 
1780
                link_or_sha1 = None
 
1781
            source_minikind = source_details[0]
 
1782
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1783
                # claimed content in both: diff
 
1784
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1785
                #        |        |      | diff check on source-target
 
1786
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1787
                #        |        |      | ???
 
1788
                if source_minikind in 'r':
 
1789
                    # add the source to the search path to find any children it
 
1790
                    # has.  TODO ? : only add if it is a container ?
 
1791
                    if not osutils.is_inside_any(searched_specific_files,
 
1792
                                                 source_details[1]):
 
1793
                        search_specific_files.add(source_details[1])
 
1794
                    # generate the old path; this is needed for stating later
 
1795
                    # as well.
 
1796
                    old_path = source_details[1]
 
1797
                    old_dirname, old_basename = os.path.split(old_path)
 
1798
                    path = pathjoin(entry[0][0], entry[0][1])
 
1799
                    old_entry = state._get_entry(source_index,
 
1800
                                                 path_utf8=old_path)
 
1801
                    # update the source details variable to be the real
 
1802
                    # location.
 
1803
                    source_details = old_entry[1][source_index]
 
1804
                    source_minikind = source_details[0]
 
1805
                else:
 
1806
                    old_dirname = entry[0][0]
 
1807
                    old_basename = entry[0][1]
 
1808
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1809
                if path_info is None:
 
1810
                    # the file is missing on disk, show as removed.
 
1811
                    content_change = True
 
1812
                    target_kind = None
 
1813
                    target_exec = False
 
1814
                else:
 
1815
                    # source and target are both versioned and disk file is present.
 
1816
                    target_kind = path_info[2]
 
1817
                    if target_kind == 'directory':
 
1818
                        if source_minikind != 'd':
 
1819
                            content_change = True
 
1820
                        else:
 
1821
                            # directories have no fingerprint
 
1822
                            content_change = False
 
1823
                        target_exec = False
 
1824
                    elif target_kind == 'file':
 
1825
                        if source_minikind != 'f':
 
1826
                            content_change = True
 
1827
                        else:
 
1828
                            # We could check the size, but we already have the
 
1829
                            # sha1 hash.
 
1830
                            content_change = (link_or_sha1 != source_details[1])
 
1831
                        # Target details is updated at update_entry time
 
1832
                        if use_filesystem_for_exec:
 
1833
                            # We don't need S_ISREG here, because we are sure
 
1834
                            # we are dealing with a file.
 
1835
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1836
                        else:
 
1837
                            target_exec = target_details[3]
 
1838
                    elif target_kind == 'symlink':
 
1839
                        if source_minikind != 'l':
 
1840
                            content_change = True
 
1841
                        else:
 
1842
                            content_change = (link_or_sha1 != source_details[1])
 
1843
                        target_exec = False
 
1844
                    elif target_kind == 'tree-reference':
 
1845
                        if source_minikind != 't':
 
1846
                            content_change = True
 
1847
                        else:
 
1848
                            content_change = False
 
1849
                        target_exec = False
 
1850
                    else:
 
1851
                        raise Exception, "unknown kind %s" % path_info[2]
 
1852
                # parent id is the entry for the path in the target tree
 
1853
                if old_dirname == last_source_parent[0]:
 
1854
                    source_parent_id = last_source_parent[1]
 
1855
                else:
 
1856
                    source_parent_entry = state._get_entry(source_index,
 
1857
                                                           path_utf8=old_dirname)
 
1858
                    source_parent_id = source_parent_entry[0][2]
 
1859
                    if source_parent_id == entry[0][2]:
 
1860
                        # This is the root, so the parent is None
 
1861
                        source_parent_id = None
 
1862
                    else:
 
1863
                        last_source_parent[0] = old_dirname
 
1864
                        last_source_parent[1] = source_parent_id
 
1865
                        last_source_parent[2] = source_parent_entry
 
1866
 
 
1867
                new_dirname = entry[0][0]
 
1868
                if new_dirname == last_target_parent[0]:
 
1869
                    target_parent_id = last_target_parent[1]
 
1870
                else:
 
1871
                    # TODO: We don't always need to do the lookup, because the
 
1872
                    #       parent entry will be the same as the source entry.
 
1873
                    target_parent_entry = state._get_entry(target_index,
 
1874
                                                           path_utf8=new_dirname)
 
1875
                    target_parent_id = target_parent_entry[0][2]
 
1876
                    if target_parent_id == entry[0][2]:
 
1877
                        # This is the root, so the parent is None
 
1878
                        target_parent_id = None
 
1879
                    else:
 
1880
                        last_target_parent[0] = new_dirname
 
1881
                        last_target_parent[1] = target_parent_id
 
1882
                        last_target_parent[2] = target_parent_entry
 
1883
 
 
1884
                source_exec = source_details[3]
 
1885
                return ((entry[0][2], (old_path, path), content_change,
 
1886
                        (True, True),
 
1887
                        (source_parent_id, target_parent_id),
 
1888
                        (old_basename, entry[0][1]),
 
1889
                        (_minikind_to_kind[source_minikind], target_kind),
 
1890
                        (source_exec, target_exec)),)
 
1891
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1892
                # looks like a new file
 
1893
                if path_info is not None:
 
1894
                    path = pathjoin(entry[0][0], entry[0][1])
 
1895
                    # parent id is the entry for the path in the target tree
 
1896
                    # TODO: these are the same for an entire directory: cache em.
 
1897
                    parent_id = state._get_entry(target_index,
 
1898
                                                 path_utf8=entry[0][0])[0][2]
 
1899
                    if parent_id == entry[0][2]:
 
1900
                        parent_id = None
 
1901
                    if use_filesystem_for_exec:
 
1902
                        # We need S_ISREG here, because we aren't sure if this
 
1903
                        # is a file or not.
 
1904
                        target_exec = bool(
 
1905
                            stat.S_ISREG(path_info[3].st_mode)
 
1906
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1907
                    else:
 
1908
                        target_exec = target_details[3]
 
1909
                    return ((entry[0][2], (None, path), True,
 
1910
                            (False, True),
 
1911
                            (None, parent_id),
 
1912
                            (None, entry[0][1]),
 
1913
                            (None, path_info[2]),
 
1914
                            (None, target_exec)),)
 
1915
                else:
 
1916
                    # but its not on disk: we deliberately treat this as just
 
1917
                    # never-present. (Why ?! - RBC 20070224)
 
1918
                    pass
 
1919
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1920
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1921
                # if its still on disk, *and* theres no other entry at this
 
1922
                # path [we dont know this in this routine at the moment -
 
1923
                # perhaps we should change this - then it would be an unknown.
 
1924
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1925
                # parent id is the entry for the path in the target tree
 
1926
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1927
                if parent_id == entry[0][2]:
 
1928
                    parent_id = None
 
1929
                return ((entry[0][2], (old_path, None), True,
 
1930
                        (True, False),
 
1931
                        (parent_id, None),
 
1932
                        (entry[0][1], None),
 
1933
                        (_minikind_to_kind[source_minikind], None),
 
1934
                        (source_details[3], None)),)
 
1935
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1936
                # a rename; could be a true rename, or a rename inherited from
 
1937
                # a renamed parent. TODO: handle this efficiently. Its not
 
1938
                # common case to rename dirs though, so a correct but slow
 
1939
                # implementation will do.
 
1940
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1941
                    search_specific_files.add(target_details[1])
 
1942
            elif source_minikind in 'r' and target_minikind in 'r':
 
1943
                # neither of the selected trees contain this file,
 
1944
                # so skip over it. This is not currently directly tested, but
 
1945
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1946
                pass
 
1947
            else:
 
1948
                raise AssertionError("don't know how to compare "
 
1949
                    "source_minikind=%r, target_minikind=%r"
 
1950
                    % (source_minikind, target_minikind))
 
1951
                ## import pdb;pdb.set_trace()
 
1952
            return ()
 
1953
        while search_specific_files:
 
1954
            # TODO: the pending list should be lexically sorted?  the
 
1955
            # interface doesn't require it.
 
1956
            current_root = search_specific_files.pop()
 
1957
            searched_specific_files.add(current_root)
 
1958
            # process the entries for this containing directory: the rest will be
 
1959
            # found by their parents recursively.
 
1960
            root_entries = _entries_for_path(current_root)
 
1961
            root_abspath = self.target.abspath(current_root)
 
1962
            try:
 
1963
                root_stat = os.lstat(root_abspath)
 
1964
            except OSError, e:
 
1965
                if e.errno == errno.ENOENT:
 
1966
                    # the path does not exist: let _process_entry know that.
 
1967
                    root_dir_info = None
 
1968
                else:
 
1969
                    # some other random error: hand it up.
 
1970
                    raise
 
1971
            else:
 
1972
                root_dir_info = ('', current_root,
 
1973
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1974
                    root_abspath)
 
1975
                if root_dir_info[2] == 'directory':
 
1976
                    if self.target._directory_is_tree_reference(
 
1977
                        current_root.decode('utf8')):
 
1978
                        root_dir_info = root_dir_info[:2] + \
 
1979
                            ('tree-reference',) + root_dir_info[3:]
 
1980
 
 
1981
            if not root_entries and not root_dir_info:
 
1982
                # this specified path is not present at all, skip it.
 
1983
                continue
 
1984
            path_handled = False
 
1985
            for entry in root_entries:
 
1986
                for result in _process_entry(entry, root_dir_info):
 
1987
                    # this check should probably be outside the loop: one
 
1988
                    # 'iterate two trees' api, and then _iter_changes filters
 
1989
                    # unchanged pairs. - RBC 20070226
 
1990
                    path_handled = True
 
1991
                    if (include_unchanged
 
1992
                        or result[2]                    # content change
 
1993
                        or result[3][0] != result[3][1] # versioned status
 
1994
                        or result[4][0] != result[4][1] # parent id
 
1995
                        or result[5][0] != result[5][1] # name
 
1996
                        or result[6][0] != result[6][1] # kind
 
1997
                        or result[7][0] != result[7][1] # executable
 
1998
                        ):
 
1999
                        result = (result[0],
 
2000
                            ((utf8_decode(result[1][0])[0]),
 
2001
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
2002
                        yield result
 
2003
            if want_unversioned and not path_handled:
 
2004
                new_executable = bool(
 
2005
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2006
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2007
                yield (None, (None, current_root), True, (False, False),
 
2008
                    (None, None),
 
2009
                    (None, splitpath(current_root)[-1]),
 
2010
                    (None, root_dir_info[2]), (None, new_executable))
 
2011
            dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2012
            initial_key = (current_root, '', '')
 
2013
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2014
            if block_index == 0:
 
2015
                # we have processed the total root already, but because the
 
2016
                # initial key matched it we should skip it here.
 
2017
                block_index +=1
 
2018
            try:
 
2019
                current_dir_info = dir_iterator.next()
 
2020
            except OSError, e:
 
2021
                if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
2022
                    # there may be directories in the inventory even though
 
2023
                    # this path is not a file on disk: so mark it as end of
 
2024
                    # iterator
 
2025
                    current_dir_info = None
 
2026
                else:
 
2027
                    raise
 
2028
            else:
 
2029
                if current_dir_info[0][0] == '':
 
2030
                    # remove .bzr from iteration
 
2031
                    bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2032
                    assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2033
                    del current_dir_info[1][bzr_index]
 
2034
            # walk until both the directory listing and the versioned metadata
 
2035
            # are exhausted. 
 
2036
            if (block_index < len(state._dirblocks) and
 
2037
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2038
                current_block = state._dirblocks[block_index]
 
2039
            else:
 
2040
                current_block = None
 
2041
            while (current_dir_info is not None or
 
2042
                   current_block is not None):
 
2043
                if (current_dir_info and current_block
 
2044
                    and current_dir_info[0][0] != current_block[0]):
 
2045
                    if current_dir_info[0][0] < current_block[0] :
 
2046
                        # filesystem data refers to paths not covered by the dirblock.
 
2047
                        # this has two possibilities:
 
2048
                        # A) it is versioned but empty, so there is no block for it
 
2049
                        # B) it is not versioned.
 
2050
                        # in either case it was processed by the containing directories walk:
 
2051
                        # if it is root/foo, when we walked root we emitted it,
 
2052
                        # or if we ere given root/foo to walk specifically, we
 
2053
                        # emitted it when checking the walk-root entries
 
2054
                        # advance the iterator and loop - we dont need to emit it.
 
2055
                        try:
 
2056
                            current_dir_info = dir_iterator.next()
 
2057
                        except StopIteration:
 
2058
                            current_dir_info = None
 
2059
                    else:
 
2060
                        # We have a dirblock entry for this location, but there
 
2061
                        # is no filesystem path for this. This is most likely
 
2062
                        # because a directory was removed from the disk.
 
2063
                        # We don't have to report the missing directory,
 
2064
                        # because that should have already been handled, but we
 
2065
                        # need to handle all of the files that are contained
 
2066
                        # within.
 
2067
                        for current_entry in current_block[1]:
 
2068
                            # entry referring to file not present on disk.
 
2069
                            # advance the entry only, after processing.
 
2070
                            for result in _process_entry(current_entry, None):
 
2071
                                # this check should probably be outside the loop: one
 
2072
                                # 'iterate two trees' api, and then _iter_changes filters
 
2073
                                # unchanged pairs. - RBC 20070226
 
2074
                                if (include_unchanged
 
2075
                                    or result[2]                    # content change
 
2076
                                    or result[3][0] != result[3][1] # versioned status
 
2077
                                    or result[4][0] != result[4][1] # parent id
 
2078
                                    or result[5][0] != result[5][1] # name
 
2079
                                    or result[6][0] != result[6][1] # kind
 
2080
                                    or result[7][0] != result[7][1] # executable
 
2081
                                    ):
 
2082
                                    result = (result[0],
 
2083
                                        ((utf8_decode(result[1][0])[0]),
 
2084
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2085
                                    yield result
 
2086
                        block_index +=1
 
2087
                        if (block_index < len(state._dirblocks) and
 
2088
                            osutils.is_inside(current_root,
 
2089
                                              state._dirblocks[block_index][0])):
 
2090
                            current_block = state._dirblocks[block_index]
 
2091
                        else:
 
2092
                            current_block = None
 
2093
                    continue
 
2094
                entry_index = 0
 
2095
                if current_block and entry_index < len(current_block[1]):
 
2096
                    current_entry = current_block[1][entry_index]
 
2097
                else:
 
2098
                    current_entry = None
 
2099
                advance_entry = True
 
2100
                path_index = 0
 
2101
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2102
                    current_path_info = current_dir_info[1][path_index]
 
2103
                    if current_path_info[2] == 'directory':
 
2104
                        if self.target._directory_is_tree_reference(
 
2105
                            current_path_info[0].decode('utf8')):
 
2106
                            current_path_info = current_path_info[:2] + \
 
2107
                                ('tree-reference',) + current_path_info[3:]
 
2108
                else:
 
2109
                    current_path_info = None
 
2110
                advance_path = True
 
2111
                path_handled = False
 
2112
                while (current_entry is not None or
 
2113
                    current_path_info is not None):
 
2114
                    if current_entry is None:
 
2115
                        # the check for path_handled when the path is adnvaced
 
2116
                        # will yield this path if needed.
 
2117
                        pass
 
2118
                    elif current_path_info is None:
 
2119
                        # no path is fine: the per entry code will handle it.
 
2120
                        for result in _process_entry(current_entry, current_path_info):
 
2121
                            # this check should probably be outside the loop: one
 
2122
                            # 'iterate two trees' api, and then _iter_changes filters
 
2123
                            # unchanged pairs. - RBC 20070226
 
2124
                            if (include_unchanged
 
2125
                                or result[2]                    # content change
 
2126
                                or result[3][0] != result[3][1] # versioned status
 
2127
                                or result[4][0] != result[4][1] # parent id
 
2128
                                or result[5][0] != result[5][1] # name
 
2129
                                or result[6][0] != result[6][1] # kind
 
2130
                                or result[7][0] != result[7][1] # executable
 
2131
                                ):
 
2132
                                result = (result[0],
 
2133
                                    ((utf8_decode(result[1][0])[0]),
 
2134
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2135
                                yield result
 
2136
                    elif current_entry[0][1] != current_path_info[1]:
 
2137
                        if current_path_info[1] < current_entry[0][1]:
 
2138
                            # extra file on disk: pass for now, but only
 
2139
                            # increment the path, not the entry
 
2140
                            advance_entry = False
 
2141
                        else:
 
2142
                            # entry referring to file not present on disk.
 
2143
                            # advance the entry only, after processing.
 
2144
                            for result in _process_entry(current_entry, None):
 
2145
                                # this check should probably be outside the loop: one
 
2146
                                # 'iterate two trees' api, and then _iter_changes filters
 
2147
                                # unchanged pairs. - RBC 20070226
 
2148
                                path_handled = True
 
2149
                                if (include_unchanged
 
2150
                                    or result[2]                    # content change
 
2151
                                    or result[3][0] != result[3][1] # versioned status
 
2152
                                    or result[4][0] != result[4][1] # parent id
 
2153
                                    or result[5][0] != result[5][1] # name
 
2154
                                    or result[6][0] != result[6][1] # kind
 
2155
                                    or result[7][0] != result[7][1] # executable
 
2156
                                    ):
 
2157
                                    result = (result[0],
 
2158
                                        ((utf8_decode(result[1][0])[0]),
 
2159
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2160
                                    yield result
 
2161
                            advance_path = False
 
2162
                    else:
 
2163
                        for result in _process_entry(current_entry, current_path_info):
 
2164
                            # this check should probably be outside the loop: one
 
2165
                            # 'iterate two trees' api, and then _iter_changes filters
 
2166
                            # unchanged pairs. - RBC 20070226
 
2167
                            path_handled = True
 
2168
                            if (include_unchanged
 
2169
                                or result[2]                    # content change
 
2170
                                or result[3][0] != result[3][1] # versioned status
 
2171
                                or result[4][0] != result[4][1] # parent id
 
2172
                                or result[5][0] != result[5][1] # name
 
2173
                                or result[6][0] != result[6][1] # kind
 
2174
                                or result[7][0] != result[7][1] # executable
 
2175
                                ):
 
2176
                                result = (result[0],
 
2177
                                    ((utf8_decode(result[1][0])[0]),
 
2178
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2179
                                yield result
 
2180
                    if advance_entry and current_entry is not None:
 
2181
                        entry_index += 1
 
2182
                        if entry_index < len(current_block[1]):
 
2183
                            current_entry = current_block[1][entry_index]
 
2184
                        else:
 
2185
                            current_entry = None
 
2186
                    else:
 
2187
                        advance_entry = True # reset the advance flaga
 
2188
                    if advance_path and current_path_info is not None:
 
2189
                        if not path_handled:
 
2190
                            # unversioned in all regards
 
2191
                            if want_unversioned:
 
2192
                                new_executable = bool(
 
2193
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2194
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2195
                                if want_unversioned:
 
2196
                                    yield (None, (None, current_path_info[0]),
 
2197
                                        True,
 
2198
                                        (False, False),
 
2199
                                        (None, None),
 
2200
                                        (None, current_path_info[1]),
 
2201
                                        (None, current_path_info[2]),
 
2202
                                        (None, new_executable))
 
2203
                            # dont descend into this unversioned path if it is
 
2204
                            # a dir
 
2205
                            if current_path_info[2] in (
 
2206
                                'directory', 'tree-referene'):
 
2207
                                del current_dir_info[1][path_index]
 
2208
                                path_index -= 1
 
2209
                        path_index += 1
 
2210
                        if path_index < len(current_dir_info[1]):
 
2211
                            current_path_info = current_dir_info[1][path_index]
 
2212
                            if current_path_info[2] == 'directory':
 
2213
                                if self.target._directory_is_tree_reference(
 
2214
                                    current_path_info[0].decode('utf8')):
 
2215
                                    current_path_info = current_path_info[:2] + \
 
2216
                                        ('tree-reference',) + current_path_info[3:]
 
2217
                        else:
 
2218
                            current_path_info = None
 
2219
                        path_handled = False
 
2220
                    else:
 
2221
                        advance_path = True # reset the advance flagg.
 
2222
                if current_block is not None:
 
2223
                    block_index += 1
 
2224
                    if (block_index < len(state._dirblocks) and
 
2225
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2226
                        current_block = state._dirblocks[block_index]
 
2227
                    else:
 
2228
                        current_block = None
 
2229
                if current_dir_info is not None:
 
2230
                    try:
 
2231
                        current_dir_info = dir_iterator.next()
 
2232
                    except StopIteration:
 
2233
                        current_dir_info = None
 
2234
 
2255
2235
 
2256
2236
    @staticmethod
2257
2237
    def is_compatible(source, target):
2258
2238
        # the target must be a dirstate working tree
2259
 
        if not isinstance(target, DirStateWorkingTree):
 
2239
        if not isinstance(target, WorkingTree4):
2260
2240
            return False
2261
 
        # the source must be a revtree or dirstate rev tree.
 
2241
        # the source must be a revtreee or dirstate rev tree.
2262
2242
        if not isinstance(source,
2263
2243
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2264
2244
            return False
2265
2245
        # the source revid must be in the target dirstate
2266
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2246
        if not (source._revision_id == NULL_REVISION or
2267
2247
            source._revision_id in target.get_parent_ids()):
2268
 
            # TODO: what about ghosts? it may well need to
 
2248
            # TODO: what about ghosts? it may well need to 
2269
2249
            # check for them explicitly.
2270
2250
            return False
2271
2251
        return True
2281
2261
 
2282
2262
    def convert(self, tree):
2283
2263
        # lock the control files not the tree, so that we dont get tree
2284
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2264
        # on-unlock behaviours, and so that noone else diddles with the 
2285
2265
        # tree during upgrade.
2286
2266
        tree._control_files.lock_write()
2287
2267
        try:
2288
 
            tree.read_working_inventory()
2289
2268
            self.create_dirstate_data(tree)
2290
2269
            self.update_format(tree)
2291
2270
            self.remove_xml_files(tree)
2313
2292
 
2314
2293
    def update_format(self, tree):
2315
2294
        """Change the format marker."""
2316
 
        tree._transport.put_bytes('format',
2317
 
            self.target_format.as_string(),
2318
 
            mode=tree.bzrdir._get_file_mode())
2319
 
 
2320
 
 
2321
 
class Converter4to5(object):
2322
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2323
 
 
2324
 
    def __init__(self):
2325
 
        self.target_format = WorkingTreeFormat5()
2326
 
 
2327
 
    def convert(self, tree):
2328
 
        # lock the control files not the tree, so that we don't get tree
2329
 
        # on-unlock behaviours, and so that no-one else diddles with the
2330
 
        # tree during upgrade.
2331
 
        tree._control_files.lock_write()
2332
 
        try:
2333
 
            self.update_format(tree)
2334
 
        finally:
2335
 
            tree._control_files.unlock()
2336
 
 
2337
 
    def update_format(self, tree):
2338
 
        """Change the format marker."""
2339
 
        tree._transport.put_bytes('format',
2340
 
            self.target_format.as_string(),
2341
 
            mode=tree.bzrdir._get_file_mode())
2342
 
 
2343
 
 
2344
 
class Converter4or5to6(object):
2345
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2346
 
 
2347
 
    def __init__(self):
2348
 
        self.target_format = WorkingTreeFormat6()
2349
 
 
2350
 
    def convert(self, tree):
2351
 
        # lock the control files not the tree, so that we don't get tree
2352
 
        # on-unlock behaviours, and so that no-one else diddles with the
2353
 
        # tree during upgrade.
2354
 
        tree._control_files.lock_write()
2355
 
        try:
2356
 
            self.init_custom_control_files(tree)
2357
 
            self.update_format(tree)
2358
 
        finally:
2359
 
            tree._control_files.unlock()
2360
 
 
2361
 
    def init_custom_control_files(self, tree):
2362
 
        """Initialize custom control files."""
2363
 
        tree._transport.put_bytes('views', '',
2364
 
            mode=tree.bzrdir._get_file_mode())
2365
 
 
2366
 
    def update_format(self, tree):
2367
 
        """Change the format marker."""
2368
 
        tree._transport.put_bytes('format',
2369
 
            self.target_format.as_string(),
2370
 
            mode=tree.bzrdir._get_file_mode())
 
2295
        tree._control_files.put_utf8('format',
 
2296
            self.target_format.get_format_string())