~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: John Arbash Meinel
  • Date: 2007-07-12 18:10:59 UTC
  • mto: This revision was merged to the branch mainline in revision 2643.
  • Revision ID: john@arbash-meinel.com-20070712181059-xnomv3tzzsb2hpx5
Add NEWS entries for performance improvements.

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
# This is the Windows equivalent of ENOTDIR
 
100
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
101
# just an error code.
 
102
ERROR_PATH_NOT_FOUND = 3
 
103
ERROR_DIRECTORY = 267
 
104
 
 
105
 
 
106
class WorkingTree4(WorkingTree3):
 
107
    """This is the Format 4 working tree.
 
108
 
 
109
    This differs from WorkingTree3 by:
 
110
     - Having a consolidated internal dirstate, stored in a
 
111
       randomly-accessible sorted file on disk.
 
112
     - Not having a regular inventory attribute.  One can be synthesized 
 
113
       on demand but this is expensive and should be avoided.
 
114
 
 
115
    This is new in bzr 0.15.
 
116
    """
89
117
 
90
118
    def __init__(self, basedir,
91
119
                 branch,
101
129
        """
102
130
        self._format = _format
103
131
        self.bzrdir = _bzrdir
 
132
        from bzrlib.trace import note, mutter
 
133
        assert isinstance(basedir, basestring), \
 
134
            "base directory %r is not a string" % basedir
104
135
        basedir = safe_unicode(basedir)
105
 
        trace.mutter("opening working tree %r", basedir)
 
136
        mutter("opening working tree %r", basedir)
106
137
        self._branch = branch
 
138
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
139
            "branch %r is not a Branch" % self.branch
107
140
        self.basedir = realpath(basedir)
108
141
        # if branch is at our basedir and is a format 6 or less
109
142
        # assume all other formats have their own control files.
 
143
        assert isinstance(_control_files, LockableFiles), \
 
144
            "_control_files must be a LockableFiles, not %r" % _control_files
110
145
        self._control_files = _control_files
111
 
        self._transport = self._control_files._transport
112
146
        self._dirty = None
113
147
        #-------------
114
148
        # during a read or write lock these objects are set, and are
116
150
        self._dirstate = None
117
151
        self._inventory = None
118
152
        #-------------
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
153
 
126
154
    @needs_tree_write_lock
127
155
    def _add(self, files, ids, kinds):
129
157
        state = self.current_dirstate()
130
158
        for f, file_id, kind in zip(files, ids, kinds):
131
159
            f = f.strip('/')
 
160
            assert '//' not in f
 
161
            assert '..' not in f
132
162
            if self.path2id(f):
133
163
                # special case tree root handling.
134
164
                if f == '' and self.path2id(f) == ROOT_ID:
142
172
            state.add(f, file_id, kind, None, '')
143
173
        self._make_dirty(reset_inventory=True)
144
174
 
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
175
    def _make_dirty(self, reset_inventory):
150
176
        """Make the tree state dirty.
151
177
 
159
185
    @needs_tree_write_lock
160
186
    def add_reference(self, sub_tree):
161
187
        # use standard implementation, which calls back to self._add
162
 
        #
 
188
        # 
163
189
        # So we don't store the reference_revision in the working dirstate,
164
 
        # it's just recorded at the moment of commit.
 
190
        # it's just recorded at the moment of commit. 
165
191
        self._add_reference(sub_tree)
166
192
 
167
193
    def break_lock(self):
203
229
 
204
230
    def _comparison_data(self, entry, path):
205
231
        kind, executable, stat_value = \
206
 
            WorkingTree._comparison_data(self, entry, path)
 
232
            WorkingTree3._comparison_data(self, entry, path)
207
233
        # it looks like a plain directory, but it's really a reference -- see
208
234
        # also kind()
209
 
        if (self._repo_supports_tree_reference and kind == 'directory'
210
 
            and entry is not None and entry.kind == 'tree-reference'):
 
235
        if (self._repo_supports_tree_reference and
 
236
            kind == 'directory' and
 
237
            self._directory_is_tree_reference(path)):
211
238
            kind = 'tree-reference'
212
239
        return kind, executable, stat_value
213
240
 
215
242
    def commit(self, message=None, revprops=None, *args, **kwargs):
216
243
        # mark the tree as dirty post commit - commit
217
244
        # can change the current versioned list by doing deletes.
218
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
245
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
219
246
        self._make_dirty(reset_inventory=True)
220
247
        return result
221
248
 
239
266
            return self._dirstate
240
267
        local_path = self.bzrdir.get_workingtree_transport(None
241
268
            ).local_abspath('dirstate')
242
 
        self._dirstate = dirstate.DirState.on_file(local_path,
243
 
            self._sha1_provider(), self._worth_saving_limit())
 
269
        self._dirstate = dirstate.DirState.on_file(local_path)
244
270
        return self._dirstate
245
271
 
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')
 
272
    def _directory_is_tree_reference(self, relpath):
 
273
        # as a special case, if a directory contains control files then 
 
274
        # it's a tree reference, except that the root of the tree is not
 
275
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
276
        # TODO: We could ask all the control formats whether they
 
277
        # recognize this directory, but at the moment there's no cheap api
 
278
        # to do that.  Since we probably can only nest bzr checkouts and
 
279
        # they always use this name it's ok for now.  -- mbp 20060306
 
280
        #
 
281
        # FIXME: There is an unhandled case here of a subdirectory
 
282
        # containing .bzr but not a branch; that will probably blow up
 
283
        # when you try to commit it.  It might happen if there is a
 
284
        # checkout in a subdirectory.  This can be avoided by not adding
 
285
        # it.  mbp 20070306
265
286
 
266
287
    def filter_unversioned_files(self, paths):
267
288
        """Filter out paths that are versioned.
300
321
 
301
322
    def _generate_inventory(self):
302
323
        """Create and set self.inventory from the dirstate object.
303
 
 
 
324
        
304
325
        This is relatively expensive: we have to walk the entire dirstate.
305
326
        Ideally we would not, and can deprecate this function.
306
327
        """
310
331
        state._read_dirblocks_if_needed()
311
332
        root_key, current_entry = self._get_entry(path='')
312
333
        current_id = root_key[2]
313
 
        if not (current_entry[0][0] == 'd'): # directory
314
 
            raise AssertionError(current_entry)
 
334
        assert current_entry[0][0] == 'd' # directory
315
335
        inv = Inventory(root_id=current_id)
316
336
        # Turn some things into local variables
317
337
        minikind_to_kind = dirstate.DirState._minikind_to_kind
350
370
                    # add this entry to the parent map.
351
371
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
352
372
                elif kind == 'tree-reference':
353
 
                    if not self._repo_supports_tree_reference:
354
 
                        raise errors.UnsupportedOperation(
355
 
                            self._generate_inventory,
356
 
                            self.branch.repository)
 
373
                    assert self._repo_supports_tree_reference, \
 
374
                        "repository of %r " \
 
375
                        "doesn't support tree references " \
 
376
                        "required by entry %r" \
 
377
                        % (self, name)
357
378
                    inv_entry.reference_revision = link_or_sha1 or None
358
379
                elif kind != 'symlink':
359
380
                    raise AssertionError("unknown kind %r" % kind)
360
381
                # 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,))
 
382
                assert file_id not in inv_byid, ('file_id %s already in'
 
383
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
384
                assert name_unicode not in parent_ie.children
367
385
                inv_byid[file_id] = inv_entry
368
386
                parent_ie.children[name_unicode] = inv_entry
369
387
        self._inventory = inv
374
392
        If either file_id or path is supplied, it is used as the key to lookup.
375
393
        If both are supplied, the fastest lookup is used, and an error is
376
394
        raised if they do not both point at the same row.
377
 
 
 
395
        
378
396
        :param file_id: An optional unicode file_id to be looked up.
379
397
        :param path: An optional unicode path to be looked up.
380
398
        :return: The dirstate row tuple for path/file_id, or (None, None)
389
407
    def get_file_sha1(self, file_id, path=None, stat_value=None):
390
408
        # check file id is valid unconditionally.
391
409
        entry = self._get_entry(file_id=file_id, path=path)
392
 
        if entry[0] is None:
393
 
            raise errors.NoSuchId(self, file_id)
 
410
        assert entry[0] is not None, 'what error should this raise'
394
411
        if path is None:
395
412
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
396
413
 
398
415
        state = self.current_dirstate()
399
416
        if stat_value is None:
400
417
            try:
401
 
                stat_value = osutils.lstat(file_abspath)
 
418
                stat_value = os.lstat(file_abspath)
402
419
            except OSError, e:
403
420
                if e.errno == errno.ENOENT:
404
421
                    return None
405
422
                else:
406
423
                    raise
407
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
408
 
            stat_value=stat_value)
 
424
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
425
                                          stat_value=stat_value)
409
426
        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
 
427
            return link_or_sha1
420
428
        return None
421
429
 
422
 
    def _get_root_inventory(self):
 
430
    def _get_inventory(self):
423
431
        """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
432
        if self._inventory is not None:
428
433
            return self._inventory
429
434
        self._must_be_locked()
430
435
        self._generate_inventory()
431
436
        return self._inventory
432
437
 
433
 
    @deprecated_method(deprecated_in((2, 5, 0)))
434
 
    def _get_inventory(self):
435
 
        return self.root_inventory
436
 
 
437
438
    inventory = property(_get_inventory,
438
439
                         doc="Inventory of this Tree")
439
440
 
440
 
    root_inventory = property(_get_root_inventory,
441
 
        "Root inventory of this tree")
442
 
 
443
441
    @needs_read_lock
444
442
    def get_parent_ids(self):
445
443
        """See Tree.get_parent_ids.
446
 
 
 
444
        
447
445
        This implementation requests the ids list from the dirstate file.
448
446
        """
449
447
        return self.current_dirstate().get_parent_ids()
465
463
 
466
464
    def has_id(self, file_id):
467
465
        state = self.current_dirstate()
 
466
        file_id = osutils.safe_file_id(file_id)
468
467
        row, parents = self._get_entry(file_id=file_id)
469
468
        if row is None:
470
469
            return False
471
470
        return osutils.lexists(pathjoin(
472
471
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
473
472
 
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
473
    @needs_read_lock
480
474
    def id2path(self, file_id):
481
475
        "Convert a file-id to a path."
 
476
        file_id = osutils.safe_file_id(file_id)
482
477
        state = self.current_dirstate()
483
478
        entry = self._get_entry(file_id=file_id)
484
479
        if entry == (None, None):
486
481
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
487
482
        return path_utf8.decode('utf8')
488
483
 
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():
 
484
    if not osutils.supports_executable():
 
485
        def is_executable(self, file_id, path=None):
 
486
            """Test if a file is executable or not.
 
487
 
 
488
            Note: The caller is expected to take a read-lock before calling this.
 
489
            """
 
490
            file_id = osutils.safe_file_id(file_id)
501
491
            entry = self._get_entry(file_id=file_id, path=path)
502
492
            if entry == (None, None):
503
493
                return False
504
494
            return entry[1][0][3]
505
 
        else:
506
 
            self._must_be_locked()
 
495
    else:
 
496
        def is_executable(self, file_id, path=None):
 
497
            """Test if a file is executable or not.
 
498
 
 
499
            Note: The caller is expected to take a read-lock before calling this.
 
500
            """
507
501
            if not path:
 
502
                file_id = osutils.safe_file_id(file_id)
508
503
                path = self.id2path(file_id)
509
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
504
            mode = os.lstat(self.abspath(path)).st_mode
510
505
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
511
506
 
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
507
    @needs_read_lock
523
508
    def __iter__(self):
524
509
        """Iterate through file_ids for this tree.
537
522
        return iter(result)
538
523
 
539
524
    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
525
        for key, tree_details in self.current_dirstate()._iter_entries():
545
526
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
546
527
                # not relevant to the working tree
548
529
            if not key[1]:
549
530
                # the root is not a reference.
550
531
                continue
551
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
532
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
552
533
            try:
553
 
                if self._kind(relpath) == 'tree-reference':
554
 
                    yield relpath, key[2]
 
534
                if self._kind(path) == 'tree-reference':
 
535
                    yield path, key[2]
555
536
            except errors.NoSuchFile:
556
537
                # path is missing on disk.
557
538
                continue
558
539
 
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
 
 
565
540
    def kind(self, file_id):
566
541
        """Return the kind of a file.
567
542
 
571
546
        Note: The caller is expected to take a read-lock before calling this.
572
547
        """
573
548
        relpath = self.id2path(file_id)
574
 
        if relpath is None:
575
 
            raise AssertionError(
576
 
                "path for id {%s} is None!" % file_id)
 
549
        assert relpath != None, \
 
550
            "path for id {%s} is None!" % file_id
577
551
        return self._kind(relpath)
578
552
 
579
553
    def _kind(self, relpath):
580
554
        abspath = self.abspath(relpath)
581
555
        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'
 
556
        if (self._repo_supports_tree_reference and
 
557
            kind == 'directory' and
 
558
            self._directory_is_tree_reference(relpath)):
 
559
            kind = 'tree-reference'
587
560
        return kind
588
561
 
589
562
    @needs_read_lock
593
566
        if parent_ids:
594
567
            return parent_ids[0]
595
568
        else:
596
 
            return _mod_revision.NULL_REVISION
 
569
            return None
597
570
 
598
571
    def lock_read(self):
599
 
        """See Branch.lock_read, and WorkingTree.unlock.
600
 
 
601
 
        :return: A bzrlib.lock.LogicalLockResult.
602
 
        """
 
572
        """See Branch.lock_read, and WorkingTree.unlock."""
603
573
        self.branch.lock_read()
604
574
        try:
605
575
            self._control_files.lock_read()
618
588
        except:
619
589
            self.branch.unlock()
620
590
            raise
621
 
        return LogicalLockResult(self.unlock)
622
591
 
623
592
    def _lock_self_write(self):
624
593
        """This should be called after the branch is locked."""
639
608
        except:
640
609
            self.branch.unlock()
641
610
            raise
642
 
        return LogicalLockResult(self.unlock)
643
611
 
644
612
    def lock_tree_write(self):
645
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
613
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
649
614
        self.branch.lock_read()
650
 
        return self._lock_self_write()
 
615
        self._lock_self_write()
651
616
 
652
617
    def lock_write(self):
653
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
654
 
 
655
 
        :return: A bzrlib.lock.LogicalLockResult.
656
 
        """
 
618
        """See MutableTree.lock_write, and WorkingTree.unlock."""
657
619
        self.branch.lock_write()
658
 
        return self._lock_self_write()
 
620
        self._lock_self_write()
659
621
 
660
622
    @needs_tree_write_lock
661
623
    def move(self, from_paths, to_dir, after=False):
663
625
        result = []
664
626
        if not from_paths:
665
627
            return result
 
628
 
666
629
        state = self.current_dirstate()
667
 
        if isinstance(from_paths, basestring):
668
 
            raise ValueError()
 
630
 
 
631
        assert not isinstance(from_paths, basestring)
669
632
        to_dir_utf8 = to_dir.encode('utf8')
670
633
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
671
634
        id_index = state._get_id_index()
692
655
 
693
656
        if self._inventory is not None:
694
657
            update_inventory = True
695
 
            inv = self.root_inventory
 
658
            inv = self.inventory
 
659
            to_dir_ie = inv[to_dir_id]
696
660
            to_dir_id = to_entry[0][2]
697
 
            to_dir_ie = inv[to_dir_id]
698
661
        else:
699
662
            update_inventory = False
700
663
 
731
694
            from_entry = self._get_entry(path=from_rel)
732
695
            if from_entry == (None, None):
733
696
                raise errors.BzrMoveFailedError(from_rel,to_dir,
734
 
                    errors.NotVersionedError(path=from_rel))
 
697
                    errors.NotVersionedError(path=str(from_rel)))
735
698
 
736
699
            from_id = from_entry[0][2]
737
700
            to_rel = pathjoin(to_dir, from_tail)
764
727
                if from_missing: # implicitly just update our path mapping
765
728
                    move_file = False
766
729
                elif not after:
767
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
730
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
731
                        extra="(Use --after to update the Bazaar id)")
768
732
 
769
733
            rollbacks = []
770
734
            def rollback_rename():
826
790
                if minikind == 'd':
827
791
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
828
792
                        """Recursively update all entries in this dirblock."""
829
 
                        if from_dir == '':
830
 
                            raise AssertionError("renaming root not supported")
 
793
                        assert from_dir != '', "renaming root not supported"
831
794
                        from_key = (from_dir, '')
832
795
                        from_block_idx, present = \
833
796
                            state._find_block_index_from_key(from_key)
846
809
 
847
810
                        # Grab a copy since move_one may update the list.
848
811
                        for entry in from_block[1][:]:
849
 
                            if not (entry[0][0] == from_dir):
850
 
                                raise AssertionError()
 
812
                            assert entry[0][0] == from_dir
851
813
                            cur_details = entry[1][0]
852
814
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
853
815
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
878
840
                rollback_rename()
879
841
                raise
880
842
            result.append((from_rel, to_rel))
881
 
            state._mark_modified()
 
843
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
882
844
            self._make_dirty(reset_inventory=False)
883
845
 
884
846
        return result
894
856
    @needs_read_lock
895
857
    def path2id(self, path):
896
858
        """Return the id for path in this tree."""
897
 
        if isinstance(path, list):
898
 
            if path == []:
899
 
                path = [""]
900
 
            path = osutils.pathjoin(*path)
901
859
        path = path.strip('/')
902
860
        entry = self._get_entry(path=path)
903
861
        if entry == (None, None):
916
874
        for tree in trees:
917
875
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
918
876
                parents):
919
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
920
 
                    trees, require_versioned)
 
877
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
921
878
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
922
879
        # -- make all paths utf8 --
923
880
        paths_utf8 = set()
981
938
                    all_versioned = False
982
939
                    break
983
940
            if not all_versioned:
984
 
                raise errors.PathsNotVersionedError(
985
 
                    [p.decode('utf-8') for p in paths])
 
941
                raise errors.PathsNotVersionedError(paths)
986
942
        # -- remove redundancy in supplied paths to prevent over-scanning --
987
 
        search_paths = osutils.minimum_path_selection(paths)
988
 
        # sketch:
 
943
        search_paths = set()
 
944
        for path in paths:
 
945
            other_paths = paths.difference(set([path]))
 
946
            if not osutils.is_inside_any(other_paths, path):
 
947
                # this is a top level path, we must check it.
 
948
                search_paths.add(path)
 
949
        # sketch: 
989
950
        # for all search_indexs in each path at or under each element of
990
951
        # search_paths, if the detail is relocated: add the id, and add the
991
952
        # relocated path as one to search if its not searched already. If the
1037
998
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1038
999
            for dir_name in split_paths:
1039
1000
                if dir_name not in found_dir_names:
1040
 
                    raise errors.PathsNotVersionedError(
1041
 
                        [p.decode('utf-8') for p in paths])
 
1001
                    raise errors.PathsNotVersionedError(paths)
1042
1002
 
1043
1003
        for dir_name_id, trees_info in found.iteritems():
1044
1004
            for index in search_indexes:
1048
1008
 
1049
1009
    def read_working_inventory(self):
1050
1010
        """Read the working inventory.
1051
 
 
 
1011
        
1052
1012
        This is a meaningless operation for dirstate, but we obey it anyhow.
1053
1013
        """
1054
 
        return self.root_inventory
 
1014
        return self.inventory
1055
1015
 
1056
1016
    @needs_read_lock
1057
1017
    def revision_tree(self, revision_id):
1059
1019
 
1060
1020
        WorkingTree4 supplies revision_trees for any basis tree.
1061
1021
        """
 
1022
        revision_id = osutils.safe_revision_id(revision_id)
1062
1023
        dirstate = self.current_dirstate()
1063
1024
        parent_ids = dirstate.get_parent_ids()
1064
1025
        if revision_id not in parent_ids:
1071
1032
    @needs_tree_write_lock
1072
1033
    def set_last_revision(self, new_revision):
1073
1034
        """Change the last revision in the working tree."""
 
1035
        new_revision = osutils.safe_revision_id(new_revision)
1074
1036
        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.")
 
1037
        if new_revision in (NULL_REVISION, None):
 
1038
            assert len(parents) < 2, (
 
1039
                "setting the last parent to none with a pending merge is "
 
1040
                "unsupported.")
1080
1041
            self.set_parent_ids([])
1081
1042
        else:
1082
1043
            self.set_parent_ids([new_revision] + parents[1:],
1085
1046
    @needs_tree_write_lock
1086
1047
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1087
1048
        """Set the parent ids to revision_ids.
1088
 
 
 
1049
        
1089
1050
        See also set_parent_trees. This api will try to retrieve the tree data
1090
1051
        for each element of revision_ids from the trees repository. If you have
1091
1052
        tree data already available, it is more efficient to use
1095
1056
        :param revision_ids: The revision_ids to set as the parent ids of this
1096
1057
            working tree. Any of these may be ghosts.
1097
1058
        """
 
1059
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1098
1060
        trees = []
1099
1061
        for revision_id in revision_ids:
1100
1062
            try:
1123
1085
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1124
1086
        real_trees = []
1125
1087
        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
1088
        # convert absent trees to the null tree, which we convert back to
1133
1089
        # missing on access.
1134
1090
        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)
 
1091
            rev_id = osutils.safe_revision_id(rev_id)
1142
1092
            if tree is not None:
1143
1093
                real_trees.append((rev_id, tree))
1144
1094
            else:
1145
1095
                real_trees.append((rev_id,
1146
 
                    self.branch.repository.revision_tree(
1147
 
                        _mod_revision.NULL_REVISION)))
 
1096
                    self.branch.repository.revision_tree(None)))
1148
1097
                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)
 
1098
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1174
1099
        self._make_dirty(reset_inventory=False)
1175
1100
 
1176
1101
    def _set_root_id(self, file_id):
1180
1105
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1181
1106
            self._make_dirty(reset_inventory=True)
1182
1107
 
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
1108
    @needs_read_lock
1194
1109
    def supports_tree_reference(self):
1195
1110
        return self._repo_supports_tree_reference
1197
1112
    def unlock(self):
1198
1113
        """Unlock in format 4 trees needs to write the entire dirstate."""
1199
1114
        if self._control_files._lock_count == 1:
1200
 
            # do non-implementation specific cleanup
1201
 
            self._cleanup()
1202
 
 
1203
1115
            # eventually we should do signature checking during read locks for
1204
1116
            # dirstate updates.
1205
1117
            if self._control_files._lock_mode == 'w':
1235
1147
            return
1236
1148
        state = self.current_dirstate()
1237
1149
        state._read_dirblocks_if_needed()
1238
 
        ids_to_unversion = set(file_ids)
 
1150
        ids_to_unversion = set()
 
1151
        for file_id in file_ids:
 
1152
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1239
1153
        paths_to_unversion = set()
1240
1154
        # sketch:
1241
1155
        # check if the root is to be unversioned, if so, assert for now.
1268
1182
                # just forget the whole block.
1269
1183
                entry_index = 0
1270
1184
                while entry_index < len(block[1]):
 
1185
                    # Mark this file id as having been removed
1271
1186
                    entry = block[1][entry_index]
1272
 
                    if entry[1][0][0] in 'ar':
1273
 
                        # don't remove absent or renamed entries
 
1187
                    ids_to_unversion.discard(entry[0][2])
 
1188
                    if (entry[1][0][0] == 'a'
 
1189
                        or not state._make_absent(entry)):
1274
1190
                        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
1191
                # go to the next block. (At the moment we dont delete empty
1282
1192
                # dirblocks)
1283
1193
                block_index += 1
1304
1214
        # have to change the legacy inventory too.
1305
1215
        if self._inventory is not None:
1306
1216
            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)
 
1217
                self._inventory.remove_recursive_id(file_id)
1328
1218
 
1329
1219
    @needs_read_lock
1330
1220
    def _validate(self):
1333
1223
    @needs_tree_write_lock
1334
1224
    def _write_inventory(self, inv):
1335
1225
        """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:
 
1226
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1227
        self.current_dirstate().set_state_from_inventory(inv)
 
1228
        self._make_dirty(reset_inventory=False)
 
1229
        if self._inventory is not None:
1350
1230
            self._inventory = inv
1351
1231
        self.flush()
1352
1232
 
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):
 
1233
 
 
1234
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1235
    """The first consolidated dirstate working tree format.
 
1236
 
 
1237
    This format:
 
1238
        - exists within a metadir controlling .bzr
 
1239
        - includes an explicit version marker for the workingtree control
 
1240
          files, separate from the BzrDir format
 
1241
        - modifies the hash cache format
 
1242
        - is new in bzr 0.15
 
1243
        - uses a LockDir to guard access to it.
 
1244
    """
 
1245
 
 
1246
    upgrade_recommended = False
 
1247
 
 
1248
    def get_format_string(self):
 
1249
        """See WorkingTreeFormat.get_format_string()."""
 
1250
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1251
 
 
1252
    def get_format_description(self):
 
1253
        """See WorkingTreeFormat.get_format_description()."""
 
1254
        return "Working tree format 4"
 
1255
 
 
1256
    def initialize(self, a_bzrdir, revision_id=None):
1480
1257
        """See WorkingTreeFormat.initialize().
1481
1258
 
1482
1259
        :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.
 
1260
        revision than the branch is at.
1490
1261
 
1491
1262
        These trees get an initial random root id, if their repository supports
1492
1263
        rich root data, TREE_ROOT otherwise.
1493
1264
        """
 
1265
        revision_id = osutils.safe_revision_id(revision_id)
1494
1266
        if not isinstance(a_bzrdir.transport, LocalTransport):
1495
1267
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1496
1268
        transport = a_bzrdir.get_workingtree_transport(self)
1497
1269
        control_files = self._open_control_files(a_bzrdir)
1498
1270
        control_files.create_lock()
1499
1271
        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()
 
1272
        control_files.put_utf8('format', self.get_format_string())
 
1273
        branch = a_bzrdir.open_branch()
1506
1274
        if revision_id is None:
1507
1275
            revision_id = branch.last_revision()
1508
1276
        local_path = transport.local_abspath('dirstate')
1510
1278
        state = dirstate.DirState.initialize(local_path)
1511
1279
        state.unlock()
1512
1280
        del state
1513
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1281
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1514
1282
                         branch,
1515
1283
                         _format=self,
1516
1284
                         _bzrdir=a_bzrdir,
1518
1286
        wt._new_tree()
1519
1287
        wt.lock_tree_write()
1520
1288
        try:
1521
 
            self._init_custom_control_files(wt)
1522
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1289
            if revision_id in (None, NULL_REVISION):
1523
1290
                if branch.repository.supports_rich_root():
1524
1291
                    wt._set_root_id(generate_ids.gen_root_id())
1525
1292
                else:
1526
1293
                    wt._set_root_id(ROOT_ID)
1527
1294
                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)]
 
1295
            wt.set_last_revision(revision_id)
 
1296
            wt.flush()
 
1297
            basis = wt.basis_tree()
1543
1298
            basis.lock_read()
1544
 
            try:
1545
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1299
            # if the basis has a root id we have to use that; otherwise we use
 
1300
            # a new random one
 
1301
            basis_root_id = basis.get_root_id()
 
1302
            if basis_root_id is not None:
 
1303
                wt._set_root_id(basis_root_id)
1546
1304
                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()
 
1305
            transform.build_tree(basis, wt)
 
1306
            basis.unlock()
1570
1307
        finally:
1571
1308
            control_files.unlock()
1572
1309
            wt.unlock()
1573
1310
        return wt
1574
1311
 
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
1312
    def _open(self, a_bzrdir, control_files):
1599
1313
        """Open the tree itself.
1600
1314
 
1601
1315
        :param a_bzrdir: the dir for the tree.
1602
1316
        :param control_files: the control files for the tree.
1603
1317
        """
1604
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1318
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1605
1319
                           branch=a_bzrdir.open_branch(),
1606
1320
                           _format=self,
1607
1321
                           _bzrdir=a_bzrdir,
1608
1322
                           _control_files=control_files)
1609
1323
 
1610
1324
    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
1325
        # please test against something that will let us do tree references
1616
 
        return controldir.format_registry.make_bzrdir(
1617
 
            'development-subtree')
 
1326
        return bzrdir.format_registry.make_bzrdir(
 
1327
            'dirstate-with-subtree')
1618
1328
 
1619
1329
    _matchingbzrdir = property(__get_matchingbzrdir)
1620
1330
 
1621
1331
 
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
 
    """
 
1332
class DirStateRevisionTree(Tree):
 
1333
    """A revision tree pulling the inventory from a dirstate."""
1709
1334
 
1710
1335
    def __init__(self, dirstate, revision_id, repository):
1711
1336
        self._dirstate = dirstate
1712
 
        self._revision_id = revision_id
 
1337
        self._revision_id = osutils.safe_revision_id(revision_id)
1713
1338
        self._repository = repository
1714
1339
        self._inventory = None
1715
1340
        self._locked = 0
1716
1341
        self._dirstate_locked = False
1717
 
        self._repo_supports_tree_reference = getattr(
1718
 
            repository._format, "supports_tree_reference",
1719
 
            False)
1720
1342
 
1721
1343
    def __repr__(self):
1722
1344
        return "<%s of %s in %s>" % \
1723
1345
            (self.__class__.__name__, self._revision_id, self._dirstate)
1724
1346
 
1725
 
    def annotate_iter(self, file_id,
1726
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1347
    def annotate_iter(self, file_id):
1727
1348
        """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]
 
1349
        w = self._repository.weave_store.get_weave(file_id,
 
1350
                           self._repository.get_transaction())
 
1351
        return w.annotate_iter(self.inventory[file_id].revision)
1731
1352
 
1732
1353
    def _comparison_data(self, entry, path):
1733
1354
        """See Tree._comparison_data."""
1751
1372
    def get_root_id(self):
1752
1373
        return self.path2id('')
1753
1374
 
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
1375
    def _get_parent_index(self):
1771
1376
        """Return the index in the dirstate referenced by this tree."""
1772
1377
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1777
1382
        If either file_id or path is supplied, it is used as the key to lookup.
1778
1383
        If both are supplied, the fastest lookup is used, and an error is
1779
1384
        raised if they do not both point at the same row.
1780
 
 
 
1385
        
1781
1386
        :param file_id: An optional unicode file_id to be looked up.
1782
1387
        :param path: An optional unicode path to be looked up.
1783
1388
        :return: The dirstate row tuple for path/file_id, or (None, None)
1784
1389
        """
1785
1390
        if file_id is None and path is None:
1786
1391
            raise errors.BzrError('must supply file_id or path')
 
1392
        file_id = osutils.safe_file_id(file_id)
1787
1393
        if path is not None:
1788
1394
            path = path.encode('utf8')
1789
1395
        parent_index = self._get_parent_index()
1790
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
1791
 
            path_utf8=path)
 
1396
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1792
1397
 
1793
1398
    def _generate_inventory(self):
1794
1399
        """Create and set self.inventory from the dirstate object.
1798
1403
 
1799
1404
        This is relatively expensive: we have to walk the entire dirstate.
1800
1405
        """
1801
 
        if not self._locked:
1802
 
            raise AssertionError(
1803
 
                'cannot generate inventory of an unlocked '
1804
 
                'dirstate revision tree')
 
1406
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1407
            'dirstate revision tree'
1805
1408
        # separate call for profiling - makes it clear where the costs are.
1806
1409
        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()))
 
1410
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1411
            'parent %s has disappeared from %s' % (
 
1412
            self._revision_id, self._dirstate.get_parent_ids())
1811
1413
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1812
1414
        # This is identical now to the WorkingTree _generate_inventory except
1813
1415
        # for the tree index use.
1814
1416
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1815
1417
        current_id = root_key[2]
1816
 
        if current_entry[parent_index][0] != 'd':
1817
 
            raise AssertionError()
 
1418
        assert current_entry[parent_index][0] == 'd'
1818
1419
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1819
1420
        inv.root.revision = current_entry[parent_index][4]
1820
1421
        # Turn some things into local variables
1851
1452
                elif kind == 'directory':
1852
1453
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1853
1454
                elif kind == 'symlink':
 
1455
                    inv_entry.executable = False
 
1456
                    inv_entry.text_size = None
1854
1457
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1855
1458
                elif kind == 'tree-reference':
1856
1459
                    inv_entry.reference_revision = fingerprint or None
1858
1461
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1859
1462
                            % entry)
1860
1463
                # 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,))
 
1464
                assert file_id not in inv_byid
 
1465
                assert name_unicode not in parent_ie.children
1867
1466
                inv_byid[file_id] = inv_entry
1868
1467
                parent_ie.children[name_unicode] = inv_entry
1869
1468
        self._inventory = inv
1876
1475
        # Make sure the file exists
1877
1476
        entry = self._get_entry(file_id, path=path)
1878
1477
        if entry == (None, None): # do we raise?
1879
 
            raise errors.NoSuchId(self, file_id)
 
1478
            return None
1880
1479
        parent_index = self._get_parent_index()
1881
1480
        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
 
1481
        return self._repository.get_revision(last_changed_revision).timestamp
1887
1482
 
1888
1483
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1889
1484
        entry = self._get_entry(file_id=file_id, path=path)
1893
1488
            return parent_details[1]
1894
1489
        return None
1895
1490
 
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):
 
1491
    def get_file(self, file_id):
1902
1492
        return StringIO(self.get_file_text(file_id))
1903
1493
 
 
1494
    def get_file_lines(self, file_id):
 
1495
        ie = self.inventory[file_id]
 
1496
        return self._repository.weave_store.get_weave(file_id,
 
1497
                self._repository.get_transaction()).get_lines(ie.revision)
 
1498
 
1904
1499
    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
 
1500
        return self.inventory[file_id].text_size
1908
1501
 
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
 
1502
    def get_file_text(self, file_id):
 
1503
        return ''.join(self.get_file_lines(file_id))
1922
1504
 
1923
1505
    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):
 
1506
        return self.inventory[file_id].reference_revision
 
1507
 
 
1508
    def get_symlink_target(self, file_id):
1942
1509
        entry = self._get_entry(file_id=file_id)
1943
1510
        parent_index = self._get_parent_index()
1944
1511
        if entry[1][parent_index][0] != 'l':
1945
1512
            return None
1946
1513
        else:
1947
 
            target = entry[1][parent_index][1]
1948
 
            target = target.decode('utf8')
1949
 
            return target
 
1514
            # At present, none of the tree implementations supports non-ascii
 
1515
            # symlink targets. So we will just assume that the dirstate path is
 
1516
            # correct.
 
1517
            return entry[1][parent_index][1]
1950
1518
 
1951
1519
    def get_revision_id(self):
1952
1520
        """Return the revision id for this tree."""
1953
1521
        return self._revision_id
1954
1522
 
1955
 
    def _get_root_inventory(self):
 
1523
    def _get_inventory(self):
1956
1524
        if self._inventory is not None:
1957
1525
            return self._inventory
1958
1526
        self._must_be_locked()
1959
1527
        self._generate_inventory()
1960
1528
        return self._inventory
1961
1529
 
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
1530
    inventory = property(_get_inventory,
1970
1531
                         doc="Inventory of this Tree")
1971
1532
 
1977
1538
        return bool(self.path2id(filename))
1978
1539
 
1979
1540
    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)
 
1541
        return self.inventory[file_id].kind
2003
1542
 
2004
1543
    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]
 
1544
        ie = self.inventory[file_id]
2007
1545
        if ie.kind != "file":
2008
 
            return False
 
1546
            return None
2009
1547
        return ie.executable
2010
1548
 
2011
 
    def is_locked(self):
2012
 
        return self._locked
2013
 
 
2014
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1549
    def list_files(self, include_root=False):
2015
1550
        # We use a standard implementation, because DirStateRevisionTree is
2016
1551
        # 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:
 
1552
        inv = self._get_inventory()
 
1553
        entries = inv.iter_entries()
 
1554
        if self.inventory.root is not None and not include_root:
2028
1555
            entries.next()
2029
1556
        for path, entry in entries:
2030
1557
            yield path, 'V', entry.kind, entry.file_id, entry
2031
1558
 
2032
1559
    def lock_read(self):
2033
 
        """Lock the tree for a set of operations.
2034
 
 
2035
 
        :return: A bzrlib.lock.LogicalLockResult.
2036
 
        """
 
1560
        """Lock the tree for a set of operations."""
2037
1561
        if not self._locked:
2038
1562
            self._repository.lock_read()
2039
1563
            if self._dirstate._lock_token is None:
2040
1564
                self._dirstate.lock_read()
2041
1565
                self._dirstate_locked = True
2042
1566
        self._locked += 1
2043
 
        return LogicalLockResult(self.unlock)
2044
1567
 
2045
1568
    def _must_be_locked(self):
2046
1569
        if not self._locked:
2050
1573
    def path2id(self, path):
2051
1574
        """Return the id for path in this tree."""
2052
1575
        # 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
1576
        entry = self._get_entry(path=path)
2058
1577
        if entry == (None, None):
2059
1578
            return None
2071
1590
                self._dirstate_locked = False
2072
1591
            self._repository.unlock()
2073
1592
 
2074
 
    @needs_read_lock
2075
 
    def supports_tree_reference(self):
2076
 
        return self._repo_supports_tree_reference
2077
 
 
2078
1593
    def walkdirs(self, prefix=""):
2079
1594
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2080
 
        # implementation based on an inventory.
 
1595
        # implementation based on an inventory.  
2081
1596
        # This should be cleaned up to use the much faster Dirstate code
2082
1597
        # So for now, we just build up the parent inventory, and extract
2083
1598
        # it the same way RevisionTree does.
2084
1599
        _directory = 'directory'
2085
 
        inv = self._get_root_inventory()
 
1600
        inv = self._get_inventory()
2086
1601
        top_id = inv.path2id(prefix)
2087
1602
        if top_id is None:
2088
1603
            pending = []
2112
1627
 
2113
1628
class InterDirStateTree(InterTree):
2114
1629
    """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
 
1630
    
 
1631
    This is used only when both trees are in the dirstate working file, and 
 
1632
    the source is any parent within the dirstate, and the destination is 
2118
1633
    the current working tree of the same dirstate.
2119
1634
    """
2120
1635
    # this could be generalized to allow comparisons between any trees in the
2129
1644
    def make_source_parent_tree(source, target):
2130
1645
        """Change the source tree into a parent of the target."""
2131
1646
        revid = source.commit('record tree')
2132
 
        target.branch.fetch(source.branch, revid)
 
1647
        target.branch.repository.fetch(source.branch.repository, revid)
2133
1648
        target.set_parent_ids([revid])
2134
1649
        return target.basis_tree(), target
2135
1650
 
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
1651
    _matching_from_tree_format = WorkingTreeFormat4()
2154
1652
    _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,
 
1653
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1654
 
 
1655
    def _iter_changes(self, include_unchanged=False,
2163
1656
                      specific_files=None, pb=None, extra_trees=[],
2164
1657
                      require_versioned=True, want_unversioned=False):
2165
1658
        """Return the changes from source to target.
2166
1659
 
2167
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1660
        :return: An iterator that yields tuples. See InterTree._iter_changes
2168
1661
            for details.
2169
1662
        :param specific_files: An optional list of file paths to restrict the
2170
1663
            comparison to. When mapping filenames to ids, all matches in all
2181
1674
            output. An unversioned file is defined as one with (False, False)
2182
1675
            for the versioned pair.
2183
1676
        """
 
1677
        utf8_decode = cache_utf8._utf8_decode
 
1678
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1679
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1680
        # NB: show_status depends on being able to pass in non-versioned files
 
1681
        # and report them as unknown
2184
1682
        # TODO: handle extra trees in the dirstate.
2185
 
        if (extra_trees or specific_files == []):
 
1683
        # TODO: handle comparisons as an empty tree as a different special
 
1684
        # case? mbp 20070226
 
1685
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2186
1686
            # we can't fast-path these cases (yet)
2187
 
            return super(InterDirStateTree, self).iter_changes(
 
1687
            for f in super(InterDirStateTree, self)._iter_changes(
2188
1688
                include_unchanged, specific_files, pb, extra_trees,
2189
 
                require_versioned, want_unversioned=want_unversioned)
 
1689
                require_versioned, want_unversioned=want_unversioned):
 
1690
                yield f
 
1691
            return
2190
1692
        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))
 
1693
        assert (self.source._revision_id in parent_ids), \
 
1694
                "revision {%s} is not stored in {%s}, but %s " \
 
1695
                "can only be used for trees stored in the dirstate" \
 
1696
                % (self.source._revision_id, self.target, self._iter_changes)
2197
1697
        target_index = 0
2198
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1698
        if self.source._revision_id == NULL_REVISION:
2199
1699
            source_index = None
2200
1700
            indices = (target_index,)
2201
1701
        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))
 
1702
            assert (self.source._revision_id in parent_ids), \
 
1703
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1704
                self.source._revision_id, parent_ids)
2206
1705
            source_index = 1 + parent_ids.index(self.source._revision_id)
2207
 
            indices = (source_index, target_index)
 
1706
            indices = (source_index,target_index)
2208
1707
        # -- make all specific_files utf8 --
2209
1708
        if specific_files:
2210
1709
            specific_files_utf8 = set()
2211
1710
            for path in specific_files:
2212
 
                # Note, if there are many specific files, using cache_utf8
2213
 
                # would be good here.
2214
1711
                specific_files_utf8.add(path.encode('utf8'))
2215
1712
            specific_files = specific_files_utf8
2216
1713
        else:
2217
1714
            specific_files = set([''])
2218
1715
        # -- specific_files is now a utf8 path set --
2219
 
 
2220
1716
        # -- get the state object and prepare it.
2221
1717
        state = self.target.current_dirstate()
2222
1718
        state._read_dirblocks_if_needed()
 
1719
        def _entries_for_path(path):
 
1720
            """Return a list with all the entries that match path for all ids.
 
1721
            """
 
1722
            dirname, basename = os.path.split(path)
 
1723
            key = (dirname, basename, '')
 
1724
            block_index, present = state._find_block_index_from_key(key)
 
1725
            if not present:
 
1726
                # the block which should contain path is absent.
 
1727
                return []
 
1728
            result = []
 
1729
            block = state._dirblocks[block_index][1]
 
1730
            entry_index, _ = state._find_entry_index(key, block)
 
1731
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1732
            while (entry_index < len(block) and
 
1733
                block[entry_index][0][0:2] == key[0:2]):
 
1734
                result.append(block[entry_index])
 
1735
                entry_index += 1
 
1736
            return result
2223
1737
        if require_versioned:
2224
1738
            # -- check all supplied paths are versioned in a search tree. --
2225
 
            not_versioned = []
 
1739
            all_versioned = True
2226
1740
            for path in specific_files:
2227
 
                path_entries = state._entries_for_path(path)
 
1741
                path_entries = _entries_for_path(path)
2228
1742
                if not path_entries:
2229
1743
                    # this specified path is not present at all: error
2230
 
                    not_versioned.append(path.decode('utf-8'))
2231
 
                    continue
 
1744
                    all_versioned = False
 
1745
                    break
2232
1746
                found_versioned = False
2233
1747
                # for each id at this path
2234
1748
                for entry in path_entries:
2241
1755
                if not found_versioned:
2242
1756
                    # none of the indexes was not 'absent' at all ids for this
2243
1757
                    # path.
2244
 
                    not_versioned.append(path.decode('utf-8'))
2245
 
            if len(not_versioned) > 0:
2246
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1758
                    all_versioned = False
 
1759
                    break
 
1760
            if not all_versioned:
 
1761
                raise errors.PathsNotVersionedError(specific_files)
2247
1762
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2248
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1763
        search_specific_files = set()
 
1764
        for path in specific_files:
 
1765
            other_specific_files = specific_files.difference(set([path]))
 
1766
            if not osutils.is_inside_any(other_specific_files, path):
 
1767
                # this is a top level path, we must check it.
 
1768
                search_specific_files.add(path)
 
1769
        # sketch: 
 
1770
        # compare source_index and target_index at or under each element of search_specific_files.
 
1771
        # follow the following comparison table. Note that we only want to do diff operations when
 
1772
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1773
        # for the target.
 
1774
        # cases:
 
1775
        # 
 
1776
        # Source | Target | disk | action
 
1777
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1778
        #        |        |      | diff check on source-target
 
1779
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1780
        #        |        |      | ???
 
1781
        #   r    |  a     |      | add source to search
 
1782
        #   r    |  a     |  a   | 
 
1783
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1784
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1785
        #   a    | fdlt   |      | add new id
 
1786
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1787
        #   a    |  a     |      | not present in either tree, skip
 
1788
        #   a    |  a     |  a   | not present in any tree, skip
 
1789
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1790
        #        |        |      | may not be selected by the users list of paths.
 
1791
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1792
        #        |        |      | may not be selected by the users list of paths.
 
1793
        #  fdlt  | fdlt   |      | content in both: diff them
 
1794
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1795
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1796
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1797
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1798
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1799
        #        |        |      | this id at the other path.
 
1800
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1801
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1802
        #        |        |      | this id at the other path.
 
1803
 
 
1804
        # for all search_indexs in each path at or under each element of
 
1805
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1806
        # relocated path as one to search if its not searched already. If the
 
1807
        # detail is not relocated, add the id.
 
1808
        searched_specific_files = set()
 
1809
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1810
        # Using a list so that we can access the values and change them in
 
1811
        # nested scope. Each one is [path, file_id, entry]
 
1812
        last_source_parent = [None, None]
 
1813
        last_target_parent = [None, None]
2249
1814
 
2250
1815
        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()
 
1816
 
 
1817
        # Just a sentry, so that _process_entry can say that this
 
1818
        # record is handled, but isn't interesting to process (unchanged)
 
1819
        uninteresting = object()
 
1820
 
 
1821
 
 
1822
        old_dirname_to_file_id = {}
 
1823
        new_dirname_to_file_id = {}
 
1824
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1825
        #       keeping a cache of directories that we have seen.
 
1826
 
 
1827
        def _process_entry(entry, path_info):
 
1828
            """Compare an entry and real disk to generate delta information.
 
1829
 
 
1830
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1831
                the path of entry. If None, then the path is considered absent.
 
1832
                (Perhaps we should pass in a concrete entry for this ?)
 
1833
                Basename is returned as a utf8 string because we expect this
 
1834
                tuple will be ignored, and don't want to take the time to
 
1835
                decode.
 
1836
            :return: None if these don't match
 
1837
                     A tuple of information about the change, or
 
1838
                     the object 'uninteresting' if these match, but are
 
1839
                     basically identical.
 
1840
            """
 
1841
            if source_index is None:
 
1842
                source_details = NULL_PARENT_DETAILS
 
1843
            else:
 
1844
                source_details = entry[1][source_index]
 
1845
            target_details = entry[1][target_index]
 
1846
            target_minikind = target_details[0]
 
1847
            if path_info is not None and target_minikind in 'fdlt':
 
1848
                assert target_index == 0
 
1849
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1850
                                                  stat_value=path_info[3])
 
1851
                # The entry may have been modified by update_entry
 
1852
                target_details = entry[1][target_index]
 
1853
                target_minikind = target_details[0]
 
1854
            else:
 
1855
                link_or_sha1 = None
 
1856
            file_id = entry[0][2]
 
1857
            source_minikind = source_details[0]
 
1858
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1859
                # claimed content in both: diff
 
1860
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1861
                #        |        |      | diff check on source-target
 
1862
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1863
                #        |        |      | ???
 
1864
                if source_minikind in 'r':
 
1865
                    # add the source to the search path to find any children it
 
1866
                    # has.  TODO ? : only add if it is a container ?
 
1867
                    if not osutils.is_inside_any(searched_specific_files,
 
1868
                                                 source_details[1]):
 
1869
                        search_specific_files.add(source_details[1])
 
1870
                    # generate the old path; this is needed for stating later
 
1871
                    # as well.
 
1872
                    old_path = source_details[1]
 
1873
                    old_dirname, old_basename = os.path.split(old_path)
 
1874
                    path = pathjoin(entry[0][0], entry[0][1])
 
1875
                    old_entry = state._get_entry(source_index,
 
1876
                                                 path_utf8=old_path)
 
1877
                    # update the source details variable to be the real
 
1878
                    # location.
 
1879
                    source_details = old_entry[1][source_index]
 
1880
                    source_minikind = source_details[0]
 
1881
                else:
 
1882
                    old_dirname = entry[0][0]
 
1883
                    old_basename = entry[0][1]
 
1884
                    old_path = path = None
 
1885
                if path_info is None:
 
1886
                    # the file is missing on disk, show as removed.
 
1887
                    content_change = True
 
1888
                    target_kind = None
 
1889
                    target_exec = False
 
1890
                else:
 
1891
                    # source and target are both versioned and disk file is present.
 
1892
                    target_kind = path_info[2]
 
1893
                    if target_kind == 'directory':
 
1894
                        if path is None:
 
1895
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1896
                        new_dirname_to_file_id[path] = file_id
 
1897
                        if source_minikind != 'd':
 
1898
                            content_change = True
 
1899
                        else:
 
1900
                            # directories have no fingerprint
 
1901
                            content_change = False
 
1902
                        target_exec = False
 
1903
                    elif target_kind == 'file':
 
1904
                        if source_minikind != 'f':
 
1905
                            content_change = True
 
1906
                        else:
 
1907
                            # We could check the size, but we already have the
 
1908
                            # sha1 hash.
 
1909
                            content_change = (link_or_sha1 != source_details[1])
 
1910
                        # Target details is updated at update_entry time
 
1911
                        if use_filesystem_for_exec:
 
1912
                            # We don't need S_ISREG here, because we are sure
 
1913
                            # we are dealing with a file.
 
1914
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1915
                        else:
 
1916
                            target_exec = target_details[3]
 
1917
                    elif target_kind == 'symlink':
 
1918
                        if source_minikind != 'l':
 
1919
                            content_change = True
 
1920
                        else:
 
1921
                            content_change = (link_or_sha1 != source_details[1])
 
1922
                        target_exec = False
 
1923
                    elif target_kind == 'tree-reference':
 
1924
                        if source_minikind != 't':
 
1925
                            content_change = True
 
1926
                        else:
 
1927
                            content_change = False
 
1928
                        target_exec = False
 
1929
                    else:
 
1930
                        raise Exception, "unknown kind %s" % path_info[2]
 
1931
                if source_minikind == 'd':
 
1932
                    if path is None:
 
1933
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1934
                    old_dirname_to_file_id[old_path] = file_id
 
1935
                # parent id is the entry for the path in the target tree
 
1936
                if old_dirname == last_source_parent[0]:
 
1937
                    source_parent_id = last_source_parent[1]
 
1938
                else:
 
1939
                    try:
 
1940
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1941
                    except KeyError:
 
1942
                        source_parent_entry = state._get_entry(source_index,
 
1943
                                                               path_utf8=old_dirname)
 
1944
                        source_parent_id = source_parent_entry[0][2]
 
1945
                    if source_parent_id == entry[0][2]:
 
1946
                        # This is the root, so the parent is None
 
1947
                        source_parent_id = None
 
1948
                    else:
 
1949
                        last_source_parent[0] = old_dirname
 
1950
                        last_source_parent[1] = source_parent_id
 
1951
                new_dirname = entry[0][0]
 
1952
                if new_dirname == last_target_parent[0]:
 
1953
                    target_parent_id = last_target_parent[1]
 
1954
                else:
 
1955
                    try:
 
1956
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1957
                    except KeyError:
 
1958
                        # TODO: We don't always need to do the lookup, because the
 
1959
                        #       parent entry will be the same as the source entry.
 
1960
                        target_parent_entry = state._get_entry(target_index,
 
1961
                                                               path_utf8=new_dirname)
 
1962
                        assert target_parent_entry != (None, None), (
 
1963
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1964
                            % (new_dirname, entry))
 
1965
                        target_parent_id = target_parent_entry[0][2]
 
1966
                    if target_parent_id == entry[0][2]:
 
1967
                        # This is the root, so the parent is None
 
1968
                        target_parent_id = None
 
1969
                    else:
 
1970
                        last_target_parent[0] = new_dirname
 
1971
                        last_target_parent[1] = target_parent_id
 
1972
 
 
1973
                source_exec = source_details[3]
 
1974
                if (include_unchanged
 
1975
                    or content_change
 
1976
                    or source_parent_id != target_parent_id
 
1977
                    or old_basename != entry[0][1]
 
1978
                    or source_exec != target_exec
 
1979
                    ):
 
1980
                    if old_path is None:
 
1981
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1982
                        old_path_u = utf8_decode(old_path)[0]
 
1983
                        path_u = old_path_u
 
1984
                    else:
 
1985
                        old_path_u = utf8_decode(old_path)[0]
 
1986
                        if old_path == path:
 
1987
                            path_u = old_path_u
 
1988
                        else:
 
1989
                            path_u = utf8_decode(path)[0]
 
1990
                    source_kind = _minikind_to_kind[source_minikind]
 
1991
                    return (entry[0][2],
 
1992
                           (old_path_u, path_u),
 
1993
                           content_change,
 
1994
                           (True, True),
 
1995
                           (source_parent_id, target_parent_id),
 
1996
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
1997
                           (source_kind, target_kind),
 
1998
                           (source_exec, target_exec))
 
1999
                else:
 
2000
                    return uninteresting
 
2001
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2002
                # looks like a new file
 
2003
                if path_info is not None:
 
2004
                    path = pathjoin(entry[0][0], entry[0][1])
 
2005
                    # parent id is the entry for the path in the target tree
 
2006
                    # TODO: these are the same for an entire directory: cache em.
 
2007
                    parent_id = state._get_entry(target_index,
 
2008
                                                 path_utf8=entry[0][0])[0][2]
 
2009
                    if parent_id == entry[0][2]:
 
2010
                        parent_id = None
 
2011
                    if use_filesystem_for_exec:
 
2012
                        # We need S_ISREG here, because we aren't sure if this
 
2013
                        # is a file or not.
 
2014
                        target_exec = bool(
 
2015
                            stat.S_ISREG(path_info[3].st_mode)
 
2016
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2017
                    else:
 
2018
                        target_exec = target_details[3]
 
2019
                    return (entry[0][2],
 
2020
                           (None, utf8_decode(path)[0]),
 
2021
                           True,
 
2022
                           (False, True),
 
2023
                           (None, parent_id),
 
2024
                           (None, utf8_decode(entry[0][1])[0]),
 
2025
                           (None, path_info[2]),
 
2026
                           (None, target_exec))
 
2027
                else:
 
2028
                    # but its not on disk: we deliberately treat this as just
 
2029
                    # never-present. (Why ?! - RBC 20070224)
 
2030
                    pass
 
2031
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2032
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2033
                # if its still on disk, *and* theres no other entry at this
 
2034
                # path [we dont know this in this routine at the moment -
 
2035
                # perhaps we should change this - then it would be an unknown.
 
2036
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2037
                # parent id is the entry for the path in the target tree
 
2038
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2039
                if parent_id == entry[0][2]:
 
2040
                    parent_id = None
 
2041
                return (entry[0][2],
 
2042
                       (utf8_decode(old_path)[0], None),
 
2043
                       True,
 
2044
                       (True, False),
 
2045
                       (parent_id, None),
 
2046
                       (utf8_decode(entry[0][1])[0], None),
 
2047
                       (_minikind_to_kind[source_minikind], None),
 
2048
                       (source_details[3], None))
 
2049
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2050
                # a rename; could be a true rename, or a rename inherited from
 
2051
                # a renamed parent. TODO: handle this efficiently. Its not
 
2052
                # common case to rename dirs though, so a correct but slow
 
2053
                # implementation will do.
 
2054
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2055
                    search_specific_files.add(target_details[1])
 
2056
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2057
                # neither of the selected trees contain this file,
 
2058
                # so skip over it. This is not currently directly tested, but
 
2059
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2060
                pass
 
2061
            else:
 
2062
                raise AssertionError("don't know how to compare "
 
2063
                    "source_minikind=%r, target_minikind=%r"
 
2064
                    % (source_minikind, target_minikind))
 
2065
                ## import pdb;pdb.set_trace()
 
2066
            return None
 
2067
 
 
2068
        while search_specific_files:
 
2069
            # TODO: the pending list should be lexically sorted?  the
 
2070
            # interface doesn't require it.
 
2071
            current_root = search_specific_files.pop()
 
2072
            current_root_unicode = current_root.decode('utf8')
 
2073
            searched_specific_files.add(current_root)
 
2074
            # process the entries for this containing directory: the rest will be
 
2075
            # found by their parents recursively.
 
2076
            root_entries = _entries_for_path(current_root)
 
2077
            root_abspath = self.target.abspath(current_root_unicode)
 
2078
            try:
 
2079
                root_stat = os.lstat(root_abspath)
 
2080
            except OSError, e:
 
2081
                if e.errno == errno.ENOENT:
 
2082
                    # the path does not exist: let _process_entry know that.
 
2083
                    root_dir_info = None
 
2084
                else:
 
2085
                    # some other random error: hand it up.
 
2086
                    raise
 
2087
            else:
 
2088
                root_dir_info = ('', current_root,
 
2089
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2090
                    root_abspath)
 
2091
                if root_dir_info[2] == 'directory':
 
2092
                    if self.target._directory_is_tree_reference(
 
2093
                        current_root.decode('utf8')):
 
2094
                        root_dir_info = root_dir_info[:2] + \
 
2095
                            ('tree-reference',) + root_dir_info[3:]
 
2096
 
 
2097
            if not root_entries and not root_dir_info:
 
2098
                # this specified path is not present at all, skip it.
 
2099
                continue
 
2100
            path_handled = False
 
2101
            for entry in root_entries:
 
2102
                result = _process_entry(entry, root_dir_info)
 
2103
                if result is not None:
 
2104
                    path_handled = True
 
2105
                    if result is not uninteresting:
 
2106
                        yield result
 
2107
            if want_unversioned and not path_handled and root_dir_info:
 
2108
                new_executable = bool(
 
2109
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2110
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2111
                yield (None,
 
2112
                       (None, current_root_unicode),
 
2113
                       True,
 
2114
                       (False, False),
 
2115
                       (None, None),
 
2116
                       (None, splitpath(current_root_unicode)[-1]),
 
2117
                       (None, root_dir_info[2]),
 
2118
                       (None, new_executable)
 
2119
                      )
 
2120
            initial_key = (current_root, '', '')
 
2121
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2122
            if block_index == 0:
 
2123
                # we have processed the total root already, but because the
 
2124
                # initial key matched it we should skip it here.
 
2125
                block_index +=1
 
2126
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2127
                current_dir_info = None
 
2128
            else:
 
2129
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2130
                try:
 
2131
                    current_dir_info = dir_iterator.next()
 
2132
                except OSError, e:
 
2133
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2134
                    # python 2.5 has e.errno == EINVAL,
 
2135
                    #            and e.winerror == ERROR_DIRECTORY
 
2136
                    e_winerror = getattr(e, 'winerror', None)
 
2137
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2138
                    # there may be directories in the inventory even though
 
2139
                    # this path is not a file on disk: so mark it as end of
 
2140
                    # iterator
 
2141
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2142
                        current_dir_info = None
 
2143
                    elif (sys.platform == 'win32'
 
2144
                          and (e.errno in win_errors
 
2145
                               or e_winerror in win_errors)):
 
2146
                        current_dir_info = None
 
2147
                    else:
 
2148
                        raise
 
2149
                else:
 
2150
                    if current_dir_info[0][0] == '':
 
2151
                        # remove .bzr from iteration
 
2152
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2153
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2154
                        del current_dir_info[1][bzr_index]
 
2155
            # walk until both the directory listing and the versioned metadata
 
2156
            # are exhausted. 
 
2157
            if (block_index < len(state._dirblocks) and
 
2158
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2159
                current_block = state._dirblocks[block_index]
 
2160
            else:
 
2161
                current_block = None
 
2162
            while (current_dir_info is not None or
 
2163
                   current_block is not None):
 
2164
                if (current_dir_info and current_block
 
2165
                    and current_dir_info[0][0] != current_block[0]):
 
2166
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2167
                        # filesystem data refers to paths not covered by the dirblock.
 
2168
                        # this has two possibilities:
 
2169
                        # A) it is versioned but empty, so there is no block for it
 
2170
                        # B) it is not versioned.
 
2171
 
 
2172
                        # if (A) then we need to recurse into it to check for
 
2173
                        # new unknown files or directories.
 
2174
                        # if (B) then we should ignore it, because we don't
 
2175
                        # recurse into unknown directories.
 
2176
                        path_index = 0
 
2177
                        while path_index < len(current_dir_info[1]):
 
2178
                                current_path_info = current_dir_info[1][path_index]
 
2179
                                if want_unversioned:
 
2180
                                    if current_path_info[2] == 'directory':
 
2181
                                        if self.target._directory_is_tree_reference(
 
2182
                                            current_path_info[0].decode('utf8')):
 
2183
                                            current_path_info = current_path_info[:2] + \
 
2184
                                                ('tree-reference',) + current_path_info[3:]
 
2185
                                    new_executable = bool(
 
2186
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2187
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2188
                                    yield (None,
 
2189
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2190
                                        True,
 
2191
                                        (False, False),
 
2192
                                        (None, None),
 
2193
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2194
                                        (None, current_path_info[2]),
 
2195
                                        (None, new_executable))
 
2196
                                # dont descend into this unversioned path if it is
 
2197
                                # a dir
 
2198
                                if current_path_info[2] in ('directory',
 
2199
                                                            'tree-reference'):
 
2200
                                    del current_dir_info[1][path_index]
 
2201
                                    path_index -= 1
 
2202
                                path_index += 1
 
2203
 
 
2204
                        # This dir info has been handled, go to the next
 
2205
                        try:
 
2206
                            current_dir_info = dir_iterator.next()
 
2207
                        except StopIteration:
 
2208
                            current_dir_info = None
 
2209
                    else:
 
2210
                        # We have a dirblock entry for this location, but there
 
2211
                        # is no filesystem path for this. This is most likely
 
2212
                        # because a directory was removed from the disk.
 
2213
                        # We don't have to report the missing directory,
 
2214
                        # because that should have already been handled, but we
 
2215
                        # need to handle all of the files that are contained
 
2216
                        # within.
 
2217
                        for current_entry in current_block[1]:
 
2218
                            # entry referring to file not present on disk.
 
2219
                            # advance the entry only, after processing.
 
2220
                            result = _process_entry(current_entry, None)
 
2221
                            if result is not None:
 
2222
                                if result is not uninteresting:
 
2223
                                    yield result
 
2224
                        block_index +=1
 
2225
                        if (block_index < len(state._dirblocks) and
 
2226
                            osutils.is_inside(current_root,
 
2227
                                              state._dirblocks[block_index][0])):
 
2228
                            current_block = state._dirblocks[block_index]
 
2229
                        else:
 
2230
                            current_block = None
 
2231
                    continue
 
2232
                entry_index = 0
 
2233
                if current_block and entry_index < len(current_block[1]):
 
2234
                    current_entry = current_block[1][entry_index]
 
2235
                else:
 
2236
                    current_entry = None
 
2237
                advance_entry = True
 
2238
                path_index = 0
 
2239
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2240
                    current_path_info = current_dir_info[1][path_index]
 
2241
                    if current_path_info[2] == 'directory':
 
2242
                        if self.target._directory_is_tree_reference(
 
2243
                            current_path_info[0].decode('utf8')):
 
2244
                            current_path_info = current_path_info[:2] + \
 
2245
                                ('tree-reference',) + current_path_info[3:]
 
2246
                else:
 
2247
                    current_path_info = None
 
2248
                advance_path = True
 
2249
                path_handled = False
 
2250
                while (current_entry is not None or
 
2251
                    current_path_info is not None):
 
2252
                    if current_entry is None:
 
2253
                        # the check for path_handled when the path is adnvaced
 
2254
                        # will yield this path if needed.
 
2255
                        pass
 
2256
                    elif current_path_info is None:
 
2257
                        # no path is fine: the per entry code will handle it.
 
2258
                        result = _process_entry(current_entry, current_path_info)
 
2259
                        if result is not None:
 
2260
                            if result is not uninteresting:
 
2261
                                yield result
 
2262
                    elif (current_entry[0][1] != current_path_info[1]
 
2263
                          or current_entry[1][target_index][0] in 'ar'):
 
2264
                        # The current path on disk doesn't match the dirblock
 
2265
                        # record. Either the dirblock is marked as absent, or
 
2266
                        # the file on disk is not present at all in the
 
2267
                        # dirblock. Either way, report about the dirblock
 
2268
                        # entry, and let other code handle the filesystem one.
 
2269
 
 
2270
                        # Compare the basename for these files to determine
 
2271
                        # which comes first
 
2272
                        if current_path_info[1] < current_entry[0][1]:
 
2273
                            # extra file on disk: pass for now, but only
 
2274
                            # increment the path, not the entry
 
2275
                            advance_entry = False
 
2276
                        else:
 
2277
                            # entry referring to file not present on disk.
 
2278
                            # advance the entry only, after processing.
 
2279
                            result = _process_entry(current_entry, None)
 
2280
                            if result is not None:
 
2281
                                if result is not uninteresting:
 
2282
                                    yield result
 
2283
                            advance_path = False
 
2284
                    else:
 
2285
                        result = _process_entry(current_entry, current_path_info)
 
2286
                        if result is not None:
 
2287
                            path_handled = True
 
2288
                            if result is not uninteresting:
 
2289
                                yield result
 
2290
                    if advance_entry and current_entry is not None:
 
2291
                        entry_index += 1
 
2292
                        if entry_index < len(current_block[1]):
 
2293
                            current_entry = current_block[1][entry_index]
 
2294
                        else:
 
2295
                            current_entry = None
 
2296
                    else:
 
2297
                        advance_entry = True # reset the advance flaga
 
2298
                    if advance_path and current_path_info is not None:
 
2299
                        if not path_handled:
 
2300
                            # unversioned in all regards
 
2301
                            if want_unversioned:
 
2302
                                new_executable = bool(
 
2303
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2304
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2305
                                yield (None,
 
2306
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2307
                                    True,
 
2308
                                    (False, False),
 
2309
                                    (None, None),
 
2310
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2311
                                    (None, current_path_info[2]),
 
2312
                                    (None, new_executable))
 
2313
                            # dont descend into this unversioned path if it is
 
2314
                            # a dir
 
2315
                            if current_path_info[2] in ('directory'):
 
2316
                                del current_dir_info[1][path_index]
 
2317
                                path_index -= 1
 
2318
                        # dont descend the disk iterator into any tree 
 
2319
                        # paths.
 
2320
                        if current_path_info[2] == 'tree-reference':
 
2321
                            del current_dir_info[1][path_index]
 
2322
                            path_index -= 1
 
2323
                        path_index += 1
 
2324
                        if path_index < len(current_dir_info[1]):
 
2325
                            current_path_info = current_dir_info[1][path_index]
 
2326
                            if current_path_info[2] == 'directory':
 
2327
                                if self.target._directory_is_tree_reference(
 
2328
                                    current_path_info[0].decode('utf8')):
 
2329
                                    current_path_info = current_path_info[:2] + \
 
2330
                                        ('tree-reference',) + current_path_info[3:]
 
2331
                        else:
 
2332
                            current_path_info = None
 
2333
                        path_handled = False
 
2334
                    else:
 
2335
                        advance_path = True # reset the advance flagg.
 
2336
                if current_block is not None:
 
2337
                    block_index += 1
 
2338
                    if (block_index < len(state._dirblocks) and
 
2339
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2340
                        current_block = state._dirblocks[block_index]
 
2341
                    else:
 
2342
                        current_block = None
 
2343
                if current_dir_info is not None:
 
2344
                    try:
 
2345
                        current_dir_info = dir_iterator.next()
 
2346
                    except StopIteration:
 
2347
                        current_dir_info = None
 
2348
 
2255
2349
 
2256
2350
    @staticmethod
2257
2351
    def is_compatible(source, target):
2258
2352
        # the target must be a dirstate working tree
2259
 
        if not isinstance(target, DirStateWorkingTree):
 
2353
        if not isinstance(target, WorkingTree4):
2260
2354
            return False
2261
 
        # the source must be a revtree or dirstate rev tree.
 
2355
        # the source must be a revtreee or dirstate rev tree.
2262
2356
        if not isinstance(source,
2263
2357
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2264
2358
            return False
2265
2359
        # the source revid must be in the target dirstate
2266
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2360
        if not (source._revision_id == NULL_REVISION or
2267
2361
            source._revision_id in target.get_parent_ids()):
2268
 
            # TODO: what about ghosts? it may well need to
 
2362
            # TODO: what about ghosts? it may well need to 
2269
2363
            # check for them explicitly.
2270
2364
            return False
2271
2365
        return True
2281
2375
 
2282
2376
    def convert(self, tree):
2283
2377
        # 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
 
2378
        # on-unlock behaviours, and so that noone else diddles with the 
2285
2379
        # tree during upgrade.
2286
2380
        tree._control_files.lock_write()
2287
2381
        try:
2313
2407
 
2314
2408
    def update_format(self, tree):
2315
2409
        """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())
 
2410
        tree._control_files.put_utf8('format',
 
2411
            self.target_format.get_format_string())