~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Patch Queue Manager
  • Date: 2014-02-12 18:22:22 UTC
  • mfrom: (6589.2.1 trunk)
  • Revision ID: pqm@pqm.ubuntu.com-20140212182222-beouo25gaf1cny76
(vila) The XDG Base Directory Specification uses the XDG_CACHE_HOME,
 not XDG_CACHE_DIR. (Andrew Starr-Bochicchio)

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006 Canonical Ltd
 
1
# Copyright (C) 2007-2012 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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 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
 
25
27
from cStringIO import StringIO
26
28
import os
27
29
import sys
28
30
 
29
31
from bzrlib.lazy_import import lazy_import
30
32
lazy_import(globals(), """
31
 
from bisect import bisect_left
32
 
import collections
33
 
from copy import deepcopy
34
33
import errno
35
 
import itertools
36
 
import operator
37
34
import stat
38
 
from time import time
39
 
import warnings
40
35
 
41
 
import bzrlib
42
36
from bzrlib import (
43
37
    bzrdir,
44
38
    cache_utf8,
 
39
    config,
45
40
    conflicts as _mod_conflicts,
46
 
    delta,
 
41
    controldir,
 
42
    debug,
47
43
    dirstate,
48
44
    errors,
 
45
    filters as _mod_filters,
49
46
    generate_ids,
50
 
    globbing,
51
 
    hashcache,
52
 
    ignores,
53
 
    merge,
54
47
    osutils,
 
48
    revision as _mod_revision,
55
49
    revisiontree,
56
 
    textui,
 
50
    trace,
57
51
    transform,
58
 
    urlutils,
59
 
    xml5,
60
 
    xml6,
 
52
    views,
61
53
    )
62
 
import bzrlib.branch
63
 
from bzrlib.transport import get_transport
64
 
import bzrlib.ui
65
54
""")
66
55
 
67
 
from bzrlib import symbol_versioning
68
56
from bzrlib.decorators import needs_read_lock, needs_write_lock
69
 
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
70
 
from bzrlib.lockable_files import LockableFiles, TransportLock
 
57
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
58
from bzrlib.lock import LogicalLockResult
 
59
from bzrlib.lockable_files import LockableFiles
71
60
from bzrlib.lockdir import LockDir
72
 
import bzrlib.mutabletree
73
 
from bzrlib.mutabletree import needs_tree_write_lock
 
61
from bzrlib.mutabletree import (
 
62
    MutableTree,
 
63
    needs_tree_write_lock,
 
64
    )
74
65
from bzrlib.osutils import (
75
 
    compact_date,
76
66
    file_kind,
77
67
    isdir,
78
 
    normpath,
79
68
    pathjoin,
80
 
    rand_chars,
81
69
    realpath,
82
70
    safe_unicode,
83
 
    splitpath,
84
 
    supports_executable,
85
 
    )
86
 
from bzrlib.trace import mutter, note
 
71
    )
 
72
from bzrlib.symbol_versioning import (
 
73
    deprecated_in,
 
74
    deprecated_method,
 
75
    )
87
76
from bzrlib.transport.local import LocalTransport
88
 
from bzrlib.tree import InterTree
89
 
from bzrlib.progress import DummyProgress, ProgressPhase
90
 
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
91
 
from bzrlib.rio import RioReader, rio_file, Stanza
92
 
from bzrlib.symbol_versioning import (deprecated_passed,
93
 
        deprecated_method,
94
 
        deprecated_function,
95
 
        DEPRECATED_PARAMETER,
96
 
        zero_eight,
97
 
        zero_eleven,
98
 
        zero_thirteen,
99
 
        )
100
 
from bzrlib.tree import Tree
101
 
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
102
 
 
103
 
 
104
 
class WorkingTree4(WorkingTree3):
105
 
    """This is the Format 4 working tree.
106
 
 
107
 
    This differs from WorkingTree3 by:
108
 
     - having a consolidated internal dirstate.
109
 
     - not having a regular inventory attribute.
110
 
 
111
 
    This is new in bzr TODO FIXME SETMEBEFORE MERGE.
112
 
    """
 
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):
113
89
 
114
90
    def __init__(self, basedir,
115
91
                 branch,
125
101
        """
126
102
        self._format = _format
127
103
        self.bzrdir = _bzrdir
128
 
        from bzrlib.hashcache import HashCache
129
 
        from bzrlib.trace import note, mutter
130
 
        assert isinstance(basedir, basestring), \
131
 
            "base directory %r is not a string" % basedir
132
104
        basedir = safe_unicode(basedir)
133
 
        mutter("opening working tree %r", basedir)
 
105
        trace.mutter("opening working tree %r", basedir)
134
106
        self._branch = branch
135
 
        assert isinstance(self.branch, bzrlib.branch.Branch), \
136
 
            "branch %r is not a Branch" % self.branch
137
107
        self.basedir = realpath(basedir)
138
108
        # if branch is at our basedir and is a format 6 or less
139
109
        # assume all other formats have their own control files.
140
 
        assert isinstance(_control_files, LockableFiles), \
141
 
            "_control_files must be a LockableFiles, not %r" % _control_files
142
110
        self._control_files = _control_files
143
 
        # update the whole cache up front and write to disk if anything changed;
144
 
        # in the future we might want to do this more selectively
145
 
        # two possible ways offer themselves : in self._unlock, write the cache
146
 
        # if needed, or, when the cache sees a change, append it to the hash
147
 
        # cache file, and have the parser take the most recent entry for a
148
 
        # given path only.
149
 
        cache_filename = self.bzrdir.get_workingtree_transport(None).local_abspath('stat-cache')
150
 
        hc = self._hashcache = HashCache(basedir, cache_filename, self._control_files._file_mode)
151
 
        hc.read()
152
 
        # is this scan needed ? it makes things kinda slow.
153
 
        #hc.scan()
154
 
 
155
 
        if hc.needs_write:
156
 
            mutter("write hc")
157
 
            hc.write()
158
 
 
 
111
        self._transport = self._control_files._transport
159
112
        self._dirty = None
160
113
        #-------------
161
114
        # during a read or write lock these objects are set, and are
163
116
        self._dirstate = None
164
117
        self._inventory = None
165
118
        #-------------
 
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
166
125
 
167
126
    @needs_tree_write_lock
168
127
    def _add(self, files, ids, kinds):
170
129
        state = self.current_dirstate()
171
130
        for f, file_id, kind in zip(files, ids, kinds):
172
131
            f = f.strip('/')
173
 
            assert '//' not in f
174
 
            assert '..' not in f
 
132
            if self.path2id(f):
 
133
                # special case tree root handling.
 
134
                if f == '' and self.path2id(f) == ROOT_ID:
 
135
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
136
                continue
175
137
            if file_id is None:
176
138
                file_id = generate_ids.gen_file_id(f)
177
139
            # deliberately add the file with no cached stat or sha1
178
140
            # - on the first access it will be gathered, and we can
179
141
            # always change this once tests are all passing.
180
142
            state.add(f, file_id, kind, None, '')
 
143
        self._make_dirty(reset_inventory=True)
 
144
 
 
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
    def _make_dirty(self, reset_inventory):
 
150
        """Make the tree state dirty.
 
151
 
 
152
        :param reset_inventory: True if the cached inventory should be removed
 
153
            (presuming there is one).
 
154
        """
181
155
        self._dirty = True
 
156
        if reset_inventory and self._inventory is not None:
 
157
            self._inventory = None
 
158
 
 
159
    @needs_tree_write_lock
 
160
    def add_reference(self, sub_tree):
 
161
        # use standard implementation, which calls back to self._add
 
162
        #
 
163
        # So we don't store the reference_revision in the working dirstate,
 
164
        # it's just recorded at the moment of commit.
 
165
        self._add_reference(sub_tree)
182
166
 
183
167
    def break_lock(self):
184
168
        """Break a lock if one is present from another instance.
217
201
        self._control_files.break_lock()
218
202
        self.branch.break_lock()
219
203
 
 
204
    def _comparison_data(self, entry, path):
 
205
        kind, executable, stat_value = \
 
206
            WorkingTree._comparison_data(self, entry, path)
 
207
        # it looks like a plain directory, but it's really a reference -- see
 
208
        # also kind()
 
209
        if (self._repo_supports_tree_reference and kind == 'directory'
 
210
            and entry is not None and entry.kind == 'tree-reference'):
 
211
            kind = 'tree-reference'
 
212
        return kind, executable, stat_value
 
213
 
 
214
    @needs_write_lock
 
215
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
216
        # mark the tree as dirty post commit - commit
 
217
        # can change the current versioned list by doing deletes.
 
218
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
219
        self._make_dirty(reset_inventory=True)
 
220
        return result
 
221
 
220
222
    def current_dirstate(self):
221
 
        """Return the current dirstate object. 
 
223
        """Return the current dirstate object.
222
224
 
223
225
        This is not part of the tree interface and only exposed for ease of
224
226
        testing.
225
227
 
226
 
        :raises errors.NotWriteLocked: when not in a lock. 
 
228
        :raises errors.NotWriteLocked: when not in a lock.
227
229
        """
228
 
        if not self._control_files._lock_count:
229
 
            raise errors.ObjectNotLocked(self)
 
230
        self._must_be_locked()
230
231
        return self._current_dirstate()
231
232
 
232
233
    def _current_dirstate(self):
233
234
        """Internal function that does not check lock status.
234
 
        
 
235
 
235
236
        This is needed for break_lock which also needs the dirstate.
236
237
        """
237
238
        if self._dirstate is not None:
238
239
            return self._dirstate
239
240
        local_path = self.bzrdir.get_workingtree_transport(None
240
241
            ).local_abspath('dirstate')
241
 
        self._dirstate = dirstate.DirState.on_file(local_path)
 
242
        self._dirstate = dirstate.DirState.on_file(local_path,
 
243
            self._sha1_provider(), self._worth_saving_limit())
242
244
        return self._dirstate
243
245
 
 
246
    def _sha1_provider(self):
 
247
        """A function that returns a SHA1Provider suitable for this tree.
 
248
 
 
249
        :return: None if content filtering is not supported by this tree.
 
250
          Otherwise, a SHA1Provider is returned that sha's the canonical
 
251
          form of files, i.e. after read filters are applied.
 
252
        """
 
253
        if self.supports_content_filtering():
 
254
            return ContentFilterAwareSHA1Provider(self)
 
255
        else:
 
256
            return None
 
257
 
 
258
    def _worth_saving_limit(self):
 
259
        """How many hash changes are ok before we must save the dirstate.
 
260
 
 
261
        :return: an integer. -1 means never save.
 
262
        """
 
263
        conf = self.get_config_stack()
 
264
        return conf.get('bzr.workingtree.worth_saving_limit')
 
265
 
244
266
    def filter_unversioned_files(self, paths):
245
 
        """Filter out paths that are not versioned.
 
267
        """Filter out paths that are versioned.
246
268
 
247
269
        :return: set of paths.
248
270
        """
257
279
            dirname, basename = os.path.split(path.encode('utf8'))
258
280
            _, _, _, path_is_versioned = state._get_block_entry_index(
259
281
                dirname, basename, 0)
260
 
            if path_is_versioned:
 
282
            if not path_is_versioned:
261
283
                result.add(path)
262
284
        return result
263
285
 
269
291
        self._inventory = None
270
292
        self._dirty = False
271
293
 
 
294
    @needs_tree_write_lock
 
295
    def _gather_kinds(self, files, kinds):
 
296
        """See MutableTree._gather_kinds."""
 
297
        for pos, f in enumerate(files):
 
298
            if kinds[pos] is None:
 
299
                kinds[pos] = self._kind(f)
 
300
 
272
301
    def _generate_inventory(self):
273
302
        """Create and set self.inventory from the dirstate object.
274
 
        
 
303
 
275
304
        This is relatively expensive: we have to walk the entire dirstate.
276
305
        Ideally we would not, and can deprecate this function.
277
306
        """
281
310
        state._read_dirblocks_if_needed()
282
311
        root_key, current_entry = self._get_entry(path='')
283
312
        current_id = root_key[2]
284
 
        assert current_entry[0][0] == 'd' # directory
 
313
        if not (current_entry[0][0] == 'd'): # directory
 
314
            raise AssertionError(current_entry)
285
315
        inv = Inventory(root_id=current_id)
286
316
        # Turn some things into local variables
287
317
        minikind_to_kind = dirstate.DirState._minikind_to_kind
310
340
                inv_entry = factory[kind](file_id, name_unicode,
311
341
                                          parent_ie.file_id)
312
342
                if kind == 'file':
 
343
                    # This is only needed on win32, where this is the only way
 
344
                    # we know the executable bit.
 
345
                    inv_entry.executable = executable
313
346
                    # not strictly needed: working tree
314
 
                    #entry.executable = executable
315
 
                    #entry.text_size = size
316
 
                    #entry.text_sha1 = sha1
317
 
                    pass
 
347
                    #inv_entry.text_size = size
 
348
                    #inv_entry.text_sha1 = sha1
318
349
                elif kind == 'directory':
319
350
                    # add this entry to the parent map.
320
351
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
352
                elif kind == 'tree-reference':
 
353
                    if not self._repo_supports_tree_reference:
 
354
                        raise errors.UnsupportedOperation(
 
355
                            self._generate_inventory,
 
356
                            self.branch.repository)
 
357
                    inv_entry.reference_revision = link_or_sha1 or None
 
358
                elif kind != 'symlink':
 
359
                    raise AssertionError("unknown kind %r" % kind)
321
360
                # These checks cost us around 40ms on a 55k entry tree
322
 
                assert file_id not in inv_byid, ('file_id %s already in'
323
 
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
324
 
                assert name_unicode not in parent_ie.children
 
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,))
325
367
                inv_byid[file_id] = inv_entry
326
368
                parent_ie.children[name_unicode] = inv_entry
327
369
        self._inventory = inv
332
374
        If either file_id or path is supplied, it is used as the key to lookup.
333
375
        If both are supplied, the fastest lookup is used, and an error is
334
376
        raised if they do not both point at the same row.
335
 
        
 
377
 
336
378
        :param file_id: An optional unicode file_id to be looked up.
337
379
        :param path: An optional unicode path to be looked up.
338
380
        :return: The dirstate row tuple for path/file_id, or (None, None)
346
388
 
347
389
    def get_file_sha1(self, file_id, path=None, stat_value=None):
348
390
        # check file id is valid unconditionally.
349
 
        key, details = self._get_entry(file_id=file_id, path=path)
350
 
        assert key is not None, 'what error should this raise'
351
 
        # TODO:
352
 
        # if row stat is valid, use cached sha1, else, get a new sha1.
 
391
        entry = self._get_entry(file_id=file_id, path=path)
 
392
        if entry[0] is None:
 
393
            raise errors.NoSuchId(self, file_id)
353
394
        if path is None:
354
 
            path = pathjoin(key[0], key[1]).decode('utf8')
355
 
        return self._hashcache.get_sha1(path, stat_value)
356
 
 
357
 
    def _get_inventory(self):
 
395
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
396
 
 
397
        file_abspath = self.abspath(path)
 
398
        state = self.current_dirstate()
 
399
        if stat_value is None:
 
400
            try:
 
401
                stat_value = osutils.lstat(file_abspath)
 
402
            except OSError, e:
 
403
                if e.errno == errno.ENOENT:
 
404
                    return None
 
405
                else:
 
406
                    raise
 
407
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
 
408
            stat_value=stat_value)
 
409
        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
 
420
        return None
 
421
 
 
422
    def _get_root_inventory(self):
358
423
        """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.")
359
427
        if self._inventory is not None:
360
428
            return self._inventory
 
429
        self._must_be_locked()
361
430
        self._generate_inventory()
362
431
        return self._inventory
363
432
 
 
433
    @deprecated_method(deprecated_in((2, 5, 0)))
 
434
    def _get_inventory(self):
 
435
        return self.root_inventory
 
436
 
364
437
    inventory = property(_get_inventory,
365
438
                         doc="Inventory of this Tree")
366
439
 
 
440
    root_inventory = property(_get_root_inventory,
 
441
        "Root inventory of this tree")
 
442
 
367
443
    @needs_read_lock
368
444
    def get_parent_ids(self):
369
445
        """See Tree.get_parent_ids.
370
 
        
 
446
 
371
447
        This implementation requests the ids list from the dirstate file.
372
448
        """
373
449
        return self.current_dirstate().get_parent_ids()
374
450
 
 
451
    def get_reference_revision(self, file_id, path=None):
 
452
        # referenced tree's revision is whatever's currently there
 
453
        return self.get_nested_tree(file_id, path).last_revision()
 
454
 
 
455
    def get_nested_tree(self, file_id, path=None):
 
456
        if path is None:
 
457
            path = self.id2path(file_id)
 
458
        # else: check file_id is at path?
 
459
        return WorkingTree.open(self.abspath(path))
 
460
 
375
461
    @needs_read_lock
376
462
    def get_root_id(self):
377
463
        """Return the id of this trees root"""
379
465
 
380
466
    def has_id(self, file_id):
381
467
        state = self.current_dirstate()
382
 
        file_id = osutils.safe_file_id(file_id)
383
468
        row, parents = self._get_entry(file_id=file_id)
384
469
        if row is None:
385
470
            return False
386
471
        return osutils.lexists(pathjoin(
387
472
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
388
473
 
 
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
 
389
479
    @needs_read_lock
390
480
    def id2path(self, file_id):
391
 
        file_id = osutils.safe_file_id(file_id)
 
481
        "Convert a file-id to a path."
392
482
        state = self.current_dirstate()
393
483
        entry = self._get_entry(file_id=file_id)
394
484
        if entry == (None, None):
395
 
            return None
 
485
            raise errors.NoSuchId(tree=self, file_id=file_id)
396
486
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
397
487
        return path_utf8.decode('utf8')
398
488
 
 
489
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
490
        entry = self._get_entry(path=path)
 
491
        if entry == (None, None):
 
492
            return False # Missing entries are not executable
 
493
        return entry[1][0][3] # Executable?
 
494
 
 
495
    def is_executable(self, file_id, path=None):
 
496
        """Test if a file is executable or not.
 
497
 
 
498
        Note: The caller is expected to take a read-lock before calling this.
 
499
        """
 
500
        if not self._supports_executable():
 
501
            entry = self._get_entry(file_id=file_id, path=path)
 
502
            if entry == (None, None):
 
503
                return False
 
504
            return entry[1][0][3]
 
505
        else:
 
506
            self._must_be_locked()
 
507
            if not path:
 
508
                path = self.id2path(file_id)
 
509
            mode = osutils.lstat(self.abspath(path)).st_mode
 
510
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
511
 
 
512
    def all_file_ids(self):
 
513
        """See Tree.iter_all_file_ids"""
 
514
        self._must_be_locked()
 
515
        result = set()
 
516
        for key, tree_details in self.current_dirstate()._iter_entries():
 
517
            if tree_details[0][0] in ('a', 'r'): # relocated
 
518
                continue
 
519
            result.add(key[2])
 
520
        return result
 
521
 
399
522
    @needs_read_lock
400
523
    def __iter__(self):
401
524
        """Iterate through file_ids for this tree.
413
536
                result.append(key[2])
414
537
        return iter(result)
415
538
 
 
539
    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
        for key, tree_details in self.current_dirstate()._iter_entries():
 
545
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
546
                # not relevant to the working tree
 
547
                continue
 
548
            if not key[1]:
 
549
                # the root is not a reference.
 
550
                continue
 
551
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
552
            try:
 
553
                if self._kind(relpath) == 'tree-reference':
 
554
                    yield relpath, key[2]
 
555
            except errors.NoSuchFile:
 
556
                # path is missing on disk.
 
557
                continue
 
558
 
 
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
    def kind(self, file_id):
 
566
        """Return the kind of a file.
 
567
 
 
568
        This is always the actual kind that's on disk, regardless of what it
 
569
        was added as.
 
570
 
 
571
        Note: The caller is expected to take a read-lock before calling this.
 
572
        """
 
573
        relpath = self.id2path(file_id)
 
574
        if relpath is None:
 
575
            raise AssertionError(
 
576
                "path for id {%s} is None!" % file_id)
 
577
        return self._kind(relpath)
 
578
 
 
579
    def _kind(self, relpath):
 
580
        abspath = self.abspath(relpath)
 
581
        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'
 
587
        return kind
 
588
 
416
589
    @needs_read_lock
417
590
    def _last_revision(self):
418
591
        """See Mutable.last_revision."""
420
593
        if parent_ids:
421
594
            return parent_ids[0]
422
595
        else:
423
 
            return None
 
596
            return _mod_revision.NULL_REVISION
424
597
 
425
598
    def lock_read(self):
426
 
        super(WorkingTree4, self).lock_read()
427
 
        if self._dirstate is None:
428
 
            self.current_dirstate()
429
 
            self._dirstate.lock_read()
 
599
        """See Branch.lock_read, and WorkingTree.unlock.
 
600
 
 
601
        :return: A bzrlib.lock.LogicalLockResult.
 
602
        """
 
603
        self.branch.lock_read()
 
604
        try:
 
605
            self._control_files.lock_read()
 
606
            try:
 
607
                state = self.current_dirstate()
 
608
                if not state._lock_token:
 
609
                    state.lock_read()
 
610
                # set our support for tree references from the repository in
 
611
                # use.
 
612
                self._repo_supports_tree_reference = getattr(
 
613
                    self.branch.repository._format, "supports_tree_reference",
 
614
                    False)
 
615
            except:
 
616
                self._control_files.unlock()
 
617
                raise
 
618
        except:
 
619
            self.branch.unlock()
 
620
            raise
 
621
        return LogicalLockResult(self.unlock)
 
622
 
 
623
    def _lock_self_write(self):
 
624
        """This should be called after the branch is locked."""
 
625
        try:
 
626
            self._control_files.lock_write()
 
627
            try:
 
628
                state = self.current_dirstate()
 
629
                if not state._lock_token:
 
630
                    state.lock_write()
 
631
                # set our support for tree references from the repository in
 
632
                # use.
 
633
                self._repo_supports_tree_reference = getattr(
 
634
                    self.branch.repository._format, "supports_tree_reference",
 
635
                    False)
 
636
            except:
 
637
                self._control_files.unlock()
 
638
                raise
 
639
        except:
 
640
            self.branch.unlock()
 
641
            raise
 
642
        return LogicalLockResult(self.unlock)
430
643
 
431
644
    def lock_tree_write(self):
432
 
        super(WorkingTree4, self).lock_tree_write()
433
 
        if self._dirstate is None:
434
 
            self.current_dirstate()
435
 
            self._dirstate.lock_write()
 
645
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
 
646
 
 
647
        :return: A bzrlib.lock.LogicalLockResult.
 
648
        """
 
649
        self.branch.lock_read()
 
650
        return self._lock_self_write()
436
651
 
437
652
    def lock_write(self):
438
 
        super(WorkingTree4, self).lock_write()
439
 
        if self._dirstate is None:
440
 
            self.current_dirstate()
441
 
            self._dirstate.lock_write()
 
653
        """See MutableTree.lock_write, and WorkingTree.unlock.
 
654
 
 
655
        :return: A bzrlib.lock.LogicalLockResult.
 
656
        """
 
657
        self.branch.lock_write()
 
658
        return self._lock_self_write()
442
659
 
443
660
    @needs_tree_write_lock
444
661
    def move(self, from_paths, to_dir, after=False):
446
663
        result = []
447
664
        if not from_paths:
448
665
            return result
449
 
 
450
666
        state = self.current_dirstate()
451
 
 
452
 
        assert not isinstance(from_paths, basestring)
 
667
        if isinstance(from_paths, basestring):
 
668
            raise ValueError()
453
669
        to_dir_utf8 = to_dir.encode('utf8')
454
670
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
455
671
        id_index = state._get_id_index()
459
675
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
460
676
        if not entry_present:
461
677
            raise errors.BzrMoveFailedError('', to_dir,
462
 
                errors.NotInWorkingDirectory(to_dir))
 
678
                errors.NotVersionedError(to_dir))
463
679
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
464
680
        # get a handle on the block itself.
465
681
        to_block_index = state._ensure_block(
476
692
 
477
693
        if self._inventory is not None:
478
694
            update_inventory = True
479
 
            inv = self.inventory
 
695
            inv = self.root_inventory
 
696
            to_dir_id = to_entry[0][2]
480
697
            to_dir_ie = inv[to_dir_id]
481
 
            to_dir_id = to_entry[0][2]
482
698
        else:
483
699
            update_inventory = False
484
700
 
507
723
            new_entry = to_block[1][added_entry_index]
508
724
            rollbacks.append(lambda:state._make_absent(new_entry))
509
725
 
510
 
        # create rename entries and tuples
511
726
        for from_rel in from_paths:
512
727
            # from_rel is 'pathinroot/foo/bar'
513
728
            from_rel_utf8 = from_rel.encode('utf8')
516
731
            from_entry = self._get_entry(path=from_rel)
517
732
            if from_entry == (None, None):
518
733
                raise errors.BzrMoveFailedError(from_rel,to_dir,
519
 
                    errors.NotVersionedError(path=str(from_rel)))
 
734
                    errors.NotVersionedError(path=from_rel))
520
735
 
521
736
            from_id = from_entry[0][2]
522
737
            to_rel = pathjoin(to_dir, from_tail)
549
764
                if from_missing: # implicitly just update our path mapping
550
765
                    move_file = False
551
766
                elif not after:
552
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
553
 
                        extra="(Use --after to update the Bazaar id)")
 
767
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
554
768
 
555
769
            rollbacks = []
556
770
            def rollback_rename():
557
771
                """A single rename has failed, roll it back."""
 
772
                # roll back everything, even if we encounter trouble doing one
 
773
                # of them.
 
774
                #
 
775
                # TODO: at least log the other exceptions rather than just
 
776
                # losing them mbp 20070307
558
777
                exc_info = None
559
778
                for rollback in reversed(rollbacks):
560
779
                    try:
561
780
                        rollback()
562
781
                    except Exception, e:
563
 
                        import pdb;pdb.set_trace()
564
782
                        exc_info = sys.exc_info()
565
783
                if exc_info:
566
784
                    raise exc_info[0], exc_info[1], exc_info[2]
607
825
 
608
826
                if minikind == 'd':
609
827
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
610
 
                        """all entries in this block need updating.
611
 
 
612
 
                        TODO: This is pretty ugly, and doesn't support
613
 
                        reverting, but it works.
614
 
                        """
615
 
                        assert from_dir != '', "renaming root not supported"
 
828
                        """Recursively update all entries in this dirblock."""
 
829
                        if from_dir == '':
 
830
                            raise AssertionError("renaming root not supported")
616
831
                        from_key = (from_dir, '')
617
832
                        from_block_idx, present = \
618
833
                            state._find_block_index_from_key(from_key)
628
843
                        to_block_index = state._ensure_block(
629
844
                            to_block_index, to_entry_index, to_dir_utf8)
630
845
                        to_block = state._dirblocks[to_block_index]
631
 
                        for entry in from_block[1]:
632
 
                            assert entry[0][0] == from_dir
 
846
 
 
847
                        # Grab a copy since move_one may update the list.
 
848
                        for entry in from_block[1][:]:
 
849
                            if not (entry[0][0] == from_dir):
 
850
                                raise AssertionError()
633
851
                            cur_details = entry[1][0]
634
852
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
635
853
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
636
854
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
637
855
                            minikind = cur_details[0]
 
856
                            if minikind in 'ar':
 
857
                                # Deleted children of a renamed directory
 
858
                                # Do not need to be updated.
 
859
                                # Children that have been renamed out of this
 
860
                                # directory should also not be updated
 
861
                                continue
638
862
                            move_one(entry, from_path_utf8=from_path_utf8,
639
863
                                     minikind=minikind,
640
864
                                     executable=cur_details[3],
643
867
                                     size=cur_details[2],
644
868
                                     to_block=to_block,
645
869
                                     to_key=to_key,
646
 
                                     to_path_utf8=to_rel_utf8)
 
870
                                     to_path_utf8=to_path_utf8)
647
871
                            if minikind == 'd':
648
872
                                # We need to move all the children of this
649
873
                                # entry
654
878
                rollback_rename()
655
879
                raise
656
880
            result.append((from_rel, to_rel))
657
 
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
658
 
            self._dirty = True
 
881
            state._mark_modified()
 
882
            self._make_dirty(reset_inventory=False)
659
883
 
660
884
        return result
661
885
 
 
886
    def _must_be_locked(self):
 
887
        if not self._control_files._lock_count:
 
888
            raise errors.ObjectNotLocked(self)
 
889
 
662
890
    def _new_tree(self):
663
891
        """Initialize the state in this tree to be a new tree."""
664
892
        self._dirty = True
666
894
    @needs_read_lock
667
895
    def path2id(self, path):
668
896
        """Return the id for path in this tree."""
 
897
        if isinstance(path, list):
 
898
            if path == []:
 
899
                path = [""]
 
900
            path = osutils.pathjoin(*path)
669
901
        path = path.strip('/')
670
902
        entry = self._get_entry(path=path)
671
903
        if entry == (None, None):
684
916
        for tree in trees:
685
917
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
686
918
                parents):
687
 
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
919
                return super(DirStateWorkingTree, self).paths2ids(paths,
 
920
                    trees, require_versioned)
688
921
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
689
922
        # -- make all paths utf8 --
690
923
        paths_utf8 = set()
748
981
                    all_versioned = False
749
982
                    break
750
983
            if not all_versioned:
751
 
                raise errors.PathsNotVersionedError(paths)
 
984
                raise errors.PathsNotVersionedError(
 
985
                    [p.decode('utf-8') for p in paths])
752
986
        # -- remove redundancy in supplied paths to prevent over-scanning --
753
 
        search_paths = set()
754
 
        for path in paths:
755
 
            other_paths = paths.difference(set([path]))
756
 
            if not osutils.is_inside_any(other_paths, path):
757
 
                # this is a top level path, we must check it.
758
 
                search_paths.add(path)
759
 
        # sketch: 
 
987
        search_paths = osutils.minimum_path_selection(paths)
 
988
        # sketch:
760
989
        # for all search_indexs in each path at or under each element of
761
990
        # search_paths, if the detail is relocated: add the id, and add the
762
991
        # relocated path as one to search if its not searched already. If the
808
1037
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
809
1038
            for dir_name in split_paths:
810
1039
                if dir_name not in found_dir_names:
811
 
                    raise errors.PathsNotVersionedError(paths)
 
1040
                    raise errors.PathsNotVersionedError(
 
1041
                        [p.decode('utf-8') for p in paths])
812
1042
 
813
1043
        for dir_name_id, trees_info in found.iteritems():
814
1044
            for index in search_indexes:
818
1048
 
819
1049
    def read_working_inventory(self):
820
1050
        """Read the working inventory.
821
 
        
 
1051
 
822
1052
        This is a meaningless operation for dirstate, but we obey it anyhow.
823
1053
        """
824
 
        return self.inventory
 
1054
        return self.root_inventory
825
1055
 
826
1056
    @needs_read_lock
827
1057
    def revision_tree(self, revision_id):
829
1059
 
830
1060
        WorkingTree4 supplies revision_trees for any basis tree.
831
1061
        """
832
 
        revision_id = osutils.safe_revision_id(revision_id)
833
1062
        dirstate = self.current_dirstate()
834
1063
        parent_ids = dirstate.get_parent_ids()
835
1064
        if revision_id not in parent_ids:
842
1071
    @needs_tree_write_lock
843
1072
    def set_last_revision(self, new_revision):
844
1073
        """Change the last revision in the working tree."""
845
 
        new_revision = osutils.safe_revision_id(new_revision)
846
1074
        parents = self.get_parent_ids()
847
 
        if new_revision in (NULL_REVISION, None):
848
 
            assert len(parents) < 2, (
849
 
                "setting the last parent to none with a pending merge is "
850
 
                "unsupported.")
 
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.")
851
1080
            self.set_parent_ids([])
852
1081
        else:
853
1082
            self.set_parent_ids([new_revision] + parents[1:],
856
1085
    @needs_tree_write_lock
857
1086
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
858
1087
        """Set the parent ids to revision_ids.
859
 
        
 
1088
 
860
1089
        See also set_parent_trees. This api will try to retrieve the tree data
861
1090
        for each element of revision_ids from the trees repository. If you have
862
1091
        tree data already available, it is more efficient to use
866
1095
        :param revision_ids: The revision_ids to set as the parent ids of this
867
1096
            working tree. Any of these may be ghosts.
868
1097
        """
869
 
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
870
1098
        trees = []
871
1099
        for revision_id in revision_ids:
872
1100
            try:
895
1123
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
896
1124
        real_trees = []
897
1125
        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
 
898
1132
        # convert absent trees to the null tree, which we convert back to
899
1133
        # missing on access.
900
1134
        for rev_id, tree in parents_list:
901
 
            rev_id = osutils.safe_revision_id(rev_id)
 
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)
902
1142
            if tree is not None:
903
1143
                real_trees.append((rev_id, tree))
904
1144
            else:
905
1145
                real_trees.append((rev_id,
906
 
                    self.branch.repository.revision_tree(None)))
 
1146
                    self.branch.repository.revision_tree(
 
1147
                        _mod_revision.NULL_REVISION)))
907
1148
                ghosts.append(rev_id)
908
 
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
909
 
        self._dirty = True
 
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)
 
1174
        self._make_dirty(reset_inventory=False)
910
1175
 
911
1176
    def _set_root_id(self, file_id):
912
1177
        """See WorkingTree.set_root_id."""
913
1178
        state = self.current_dirstate()
914
1179
        state.set_path_id('', file_id)
915
 
        self._dirty = state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED
 
1180
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1181
            self._make_dirty(reset_inventory=True)
 
1182
 
 
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
    @needs_read_lock
 
1194
    def supports_tree_reference(self):
 
1195
        return self._repo_supports_tree_reference
916
1196
 
917
1197
    def unlock(self):
918
1198
        """Unlock in format 4 trees needs to write the entire dirstate."""
919
1199
        if self._control_files._lock_count == 1:
920
 
            self._write_hashcache_if_dirty()
 
1200
            # do non-implementation specific cleanup
 
1201
            self._cleanup()
 
1202
 
921
1203
            # eventually we should do signature checking during read locks for
922
1204
            # dirstate updates.
923
1205
            if self._control_files._lock_mode == 'w':
924
1206
                if self._dirty:
925
1207
                    self.flush()
926
1208
            if self._dirstate is not None:
 
1209
                # This is a no-op if there are no modifications.
 
1210
                self._dirstate.save()
927
1211
                self._dirstate.unlock()
 
1212
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1213
            #       point. Instead, it could check if the header has been
 
1214
            #       modified when it is locked, and if not, it can hang on to
 
1215
            #       the data it has in memory.
928
1216
            self._dirstate = None
929
1217
            self._inventory = None
930
1218
        # reverse order of locking.
947
1235
            return
948
1236
        state = self.current_dirstate()
949
1237
        state._read_dirblocks_if_needed()
950
 
        ids_to_unversion = set()
951
 
        for file_id in file_ids:
952
 
            ids_to_unversion.add(osutils.safe_file_id(file_id))
 
1238
        ids_to_unversion = set(file_ids)
953
1239
        paths_to_unversion = set()
954
1240
        # sketch:
955
1241
        # check if the root is to be unversioned, if so, assert for now.
982
1268
                # just forget the whole block.
983
1269
                entry_index = 0
984
1270
                while entry_index < len(block[1]):
985
 
                    # Mark this file id as having been removed
986
 
                    ids_to_unversion.discard(block[1][entry_index][0][2])
987
 
                    if not state._make_absent(block[1][entry_index]):
 
1271
                    entry = block[1][entry_index]
 
1272
                    if entry[1][0][0] in 'ar':
 
1273
                        # don't remove absent or renamed entries
988
1274
                        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
989
1281
                # go to the next block. (At the moment we dont delete empty
990
1282
                # dirblocks)
991
1283
                block_index += 1
1008
1300
            block_index += 1
1009
1301
        if ids_to_unversion:
1010
1302
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
1011
 
        self._dirty = True
 
1303
        self._make_dirty(reset_inventory=False)
1012
1304
        # have to change the legacy inventory too.
1013
1305
        if self._inventory is not None:
1014
1306
            for file_id in file_ids:
1015
 
                self._inventory.remove_recursive_id(file_id)
 
1307
                if self._inventory.has_id(file_id):
 
1308
                    self._inventory.remove_recursive_id(file_id)
 
1309
 
 
1310
    @needs_tree_write_lock
 
1311
    def rename_one(self, from_rel, to_rel, after=False):
 
1312
        """See WorkingTree.rename_one"""
 
1313
        self.flush()
 
1314
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
 
1315
 
 
1316
    @needs_tree_write_lock
 
1317
    def apply_inventory_delta(self, changes):
 
1318
        """See MutableTree.apply_inventory_delta"""
 
1319
        state = self.current_dirstate()
 
1320
        state.update_by_delta(changes)
 
1321
        self._make_dirty(reset_inventory=True)
 
1322
 
 
1323
    def update_basis_by_delta(self, new_revid, delta):
 
1324
        """See MutableTree.update_basis_by_delta."""
 
1325
        if self.last_revision() == new_revid:
 
1326
            raise AssertionError()
 
1327
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1328
 
 
1329
    @needs_read_lock
 
1330
    def _validate(self):
 
1331
        self._dirstate._validate()
1016
1332
 
1017
1333
    @needs_tree_write_lock
1018
1334
    def _write_inventory(self, inv):
1019
1335
        """Write inventory as the current inventory."""
1020
 
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
1021
 
        self.current_dirstate().set_state_from_inventory(inv)
1022
 
        self._dirty = True
 
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:
 
1350
            self._inventory = inv
1023
1351
        self.flush()
1024
1352
 
1025
 
 
1026
 
class WorkingTreeFormat4(WorkingTreeFormat3):
1027
 
    """The first consolidated dirstate working tree format.
1028
 
 
1029
 
    This format:
1030
 
        - exists within a metadir controlling .bzr
1031
 
        - includes an explicit version marker for the workingtree control
1032
 
          files, separate from the BzrDir format
1033
 
        - modifies the hash cache format
1034
 
        - is new in bzr TODO FIXME SETBEFOREMERGE
1035
 
        - uses a LockDir to guard access to it.
1036
 
    """
1037
 
 
1038
 
    def get_format_string(self):
1039
 
        """See WorkingTreeFormat.get_format_string()."""
1040
 
        return "Bazaar Working Tree format 4\n"
1041
 
 
1042
 
    def get_format_description(self):
1043
 
        """See WorkingTreeFormat.get_format_description()."""
1044
 
        return "Working tree format 4"
1045
 
 
1046
 
    def initialize(self, a_bzrdir, revision_id=None):
 
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):
1047
1480
        """See WorkingTreeFormat.initialize().
1048
1481
 
1049
 
        revision_id allows creating a working tree at a different
1050
 
        revision than the branch is at.
 
1482
        :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.
 
1490
 
 
1491
        These trees get an initial random root id, if their repository supports
 
1492
        rich root data, TREE_ROOT otherwise.
1051
1493
        """
1052
 
        revision_id = osutils.safe_revision_id(revision_id)
1053
1494
        if not isinstance(a_bzrdir.transport, LocalTransport):
1054
1495
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1055
1496
        transport = a_bzrdir.get_workingtree_transport(self)
1056
1497
        control_files = self._open_control_files(a_bzrdir)
1057
1498
        control_files.create_lock()
1058
1499
        control_files.lock_write()
1059
 
        control_files.put_utf8('format', self.get_format_string())
1060
 
        branch = a_bzrdir.open_branch()
 
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()
1061
1506
        if revision_id is None:
1062
1507
            revision_id = branch.last_revision()
1063
1508
        local_path = transport.local_abspath('dirstate')
 
1509
        # write out new dirstate (must exist when we create the tree)
1064
1510
        state = dirstate.DirState.initialize(local_path)
1065
1511
        state.unlock()
1066
 
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1512
        del state
 
1513
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1067
1514
                         branch,
1068
1515
                         _format=self,
1069
1516
                         _bzrdir=a_bzrdir,
1071
1518
        wt._new_tree()
1072
1519
        wt.lock_tree_write()
1073
1520
        try:
1074
 
            #wt.current_dirstate().set_path_id('', NEWROOT)
1075
 
            wt.set_last_revision(revision_id)
1076
 
            wt.flush()
1077
 
            basis = wt.basis_tree()
 
1521
            self._init_custom_control_files(wt)
 
1522
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1523
                if branch.repository.supports_rich_root():
 
1524
                    wt._set_root_id(generate_ids.gen_root_id())
 
1525
                else:
 
1526
                    wt._set_root_id(ROOT_ID)
 
1527
                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)]
1078
1543
            basis.lock_read()
1079
 
            transform.build_tree(basis, wt)
1080
 
            basis.unlock()
 
1544
            try:
 
1545
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1546
                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()
1081
1570
        finally:
1082
1571
            control_files.unlock()
1083
1572
            wt.unlock()
1084
1573
        return wt
1085
1574
 
 
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
1086
1597
 
1087
1598
    def _open(self, a_bzrdir, control_files):
1088
1599
        """Open the tree itself.
1090
1601
        :param a_bzrdir: the dir for the tree.
1091
1602
        :param control_files: the control files for the tree.
1092
1603
        """
1093
 
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1604
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1094
1605
                           branch=a_bzrdir.open_branch(),
1095
1606
                           _format=self,
1096
1607
                           _bzrdir=a_bzrdir,
1097
1608
                           _control_files=control_files)
1098
1609
 
1099
 
 
1100
 
class DirStateRevisionTree(Tree):
1101
 
    """A revision tree pulling the inventory from a dirstate."""
 
1610
    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
        # please test against something that will let us do tree references
 
1616
        return controldir.format_registry.make_bzrdir(
 
1617
            'development-subtree')
 
1618
 
 
1619
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1620
 
 
1621
 
 
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
    """
1102
1709
 
1103
1710
    def __init__(self, dirstate, revision_id, repository):
1104
1711
        self._dirstate = dirstate
1105
 
        self._revision_id = osutils.safe_revision_id(revision_id)
 
1712
        self._revision_id = revision_id
1106
1713
        self._repository = repository
1107
1714
        self._inventory = None
1108
1715
        self._locked = 0
1109
1716
        self._dirstate_locked = False
1110
 
 
1111
 
    def annotate_iter(self, file_id):
 
1717
        self._repo_supports_tree_reference = getattr(
 
1718
            repository._format, "supports_tree_reference",
 
1719
            False)
 
1720
 
 
1721
    def __repr__(self):
 
1722
        return "<%s of %s in %s>" % \
 
1723
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1724
 
 
1725
    def annotate_iter(self, file_id,
 
1726
                      default_revision=_mod_revision.CURRENT_REVISION):
1112
1727
        """See Tree.annotate_iter"""
1113
 
        w = self._repository.weave_store.get_weave(file_id,
1114
 
                           self._repository.get_transaction())
1115
 
        return w.annotate_iter(self.inventory[file_id].revision)
 
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]
1116
1731
 
1117
1732
    def _comparison_data(self, entry, path):
1118
1733
        """See Tree._comparison_data."""
1133
1748
        pred = self.has_filename
1134
1749
        return set((p for p in paths if not pred(p)))
1135
1750
 
 
1751
    def get_root_id(self):
 
1752
        return self.path2id('')
 
1753
 
 
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
 
1136
1770
    def _get_parent_index(self):
1137
1771
        """Return the index in the dirstate referenced by this tree."""
1138
1772
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1143
1777
        If either file_id or path is supplied, it is used as the key to lookup.
1144
1778
        If both are supplied, the fastest lookup is used, and an error is
1145
1779
        raised if they do not both point at the same row.
1146
 
        
 
1780
 
1147
1781
        :param file_id: An optional unicode file_id to be looked up.
1148
1782
        :param path: An optional unicode path to be looked up.
1149
1783
        :return: The dirstate row tuple for path/file_id, or (None, None)
1150
1784
        """
1151
1785
        if file_id is None and path is None:
1152
1786
            raise errors.BzrError('must supply file_id or path')
1153
 
        file_id = osutils.safe_file_id(file_id)
1154
1787
        if path is not None:
1155
1788
            path = path.encode('utf8')
1156
1789
        parent_index = self._get_parent_index()
1157
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1790
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
 
1791
            path_utf8=path)
1158
1792
 
1159
1793
    def _generate_inventory(self):
1160
1794
        """Create and set self.inventory from the dirstate object.
1161
1795
 
 
1796
        (So this is only called the first time the inventory is requested for
 
1797
        this tree; it then remains in memory until it's out of date.)
 
1798
 
1162
1799
        This is relatively expensive: we have to walk the entire dirstate.
1163
 
        Ideally we would not, and instead would """
1164
 
        assert self._locked, 'cannot generate inventory of an unlocked '\
1165
 
            'dirstate revision tree'
 
1800
        """
 
1801
        if not self._locked:
 
1802
            raise AssertionError(
 
1803
                'cannot generate inventory of an unlocked '
 
1804
                'dirstate revision tree')
1166
1805
        # separate call for profiling - makes it clear where the costs are.
1167
1806
        self._dirstate._read_dirblocks_if_needed()
1168
 
        assert self._revision_id in self._dirstate.get_parent_ids(), \
1169
 
            'parent %s has disappeared from %s' % (
1170
 
            self._revision_id, self._dirstate.get_parent_ids())
 
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()))
1171
1811
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1172
1812
        # This is identical now to the WorkingTree _generate_inventory except
1173
1813
        # for the tree index use.
1174
1814
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1175
1815
        current_id = root_key[2]
1176
 
        assert current_entry[parent_index][0] == 'd'
 
1816
        if current_entry[parent_index][0] != 'd':
 
1817
            raise AssertionError()
1177
1818
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1178
1819
        inv.root.revision = current_entry[parent_index][4]
1179
1820
        # Turn some things into local variables
1192
1833
                # all the paths in this block are not versioned in this tree
1193
1834
                continue
1194
1835
            for key, entry in block[1]:
1195
 
                minikind, link_or_sha1, size, executable, revid = entry[parent_index]
 
1836
                minikind, fingerprint, size, executable, revid = entry[parent_index]
1196
1837
                if minikind in ('a', 'r'): # absent, relocated
1197
1838
                    # not this tree
1198
1839
                    continue
1206
1847
                if kind == 'file':
1207
1848
                    inv_entry.executable = executable
1208
1849
                    inv_entry.text_size = size
1209
 
                    inv_entry.text_sha1 = link_or_sha1
 
1850
                    inv_entry.text_sha1 = fingerprint
1210
1851
                elif kind == 'directory':
1211
1852
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1212
1853
                elif kind == 'symlink':
1213
 
                    inv_entry.executable = False
1214
 
                    inv_entry.text_size = size
1215
 
                    inv_entry.symlink_target = utf8_decode(link_or_sha1)[0]
 
1854
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1855
                elif kind == 'tree-reference':
 
1856
                    inv_entry.reference_revision = fingerprint or None
1216
1857
                else:
1217
 
                    raise Exception, kind
 
1858
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1859
                            % entry)
1218
1860
                # These checks cost us around 40ms on a 55k entry tree
1219
 
                assert file_id not in inv_byid
1220
 
                assert name_unicode not in parent_ie.children
 
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,))
1221
1867
                inv_byid[file_id] = inv_entry
1222
1868
                parent_ie.children[name_unicode] = inv_entry
1223
1869
        self._inventory = inv
1230
1876
        # Make sure the file exists
1231
1877
        entry = self._get_entry(file_id, path=path)
1232
1878
        if entry == (None, None): # do we raise?
1233
 
            return None
 
1879
            raise errors.NoSuchId(self, file_id)
1234
1880
        parent_index = self._get_parent_index()
1235
1881
        last_changed_revision = entry[1][parent_index][4]
1236
 
        return self._repository.get_revision(last_changed_revision).timestamp
 
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
1237
1887
 
1238
1888
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1239
 
        # TODO: if path is present, fast-path on that, as inventory
1240
 
        # might not be present
1241
 
        ie = self.inventory[file_id]
1242
 
        if ie.kind == "file":
1243
 
            return ie.text_sha1
 
1889
        entry = self._get_entry(file_id=file_id, path=path)
 
1890
        parent_index = self._get_parent_index()
 
1891
        parent_details = entry[1][parent_index]
 
1892
        if parent_details[0] == 'f':
 
1893
            return parent_details[1]
1244
1894
        return None
1245
1895
 
1246
 
    def get_file(self, file_id):
 
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):
1247
1902
        return StringIO(self.get_file_text(file_id))
1248
1903
 
1249
 
    def get_file_lines(self, file_id):
1250
 
        ie = self.inventory[file_id]
1251
 
        return self._repository.weave_store.get_weave(file_id,
1252
 
                self._repository.get_transaction()).get_lines(ie.revision)
1253
 
 
1254
1904
    def get_file_size(self, file_id):
1255
 
        return self.inventory[file_id].text_size
1256
 
 
1257
 
    def get_file_text(self, file_id):
1258
 
        return ''.join(self.get_file_lines(file_id))
1259
 
 
1260
 
    def get_symlink_target(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
 
1908
 
 
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
 
1922
 
 
1923
    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):
1261
1942
        entry = self._get_entry(file_id=file_id)
1262
1943
        parent_index = self._get_parent_index()
1263
1944
        if entry[1][parent_index][0] != 'l':
1264
1945
            return None
1265
1946
        else:
1266
 
            # At present, none of the tree implementations supports non-ascii
1267
 
            # symlink targets. So we will just assume that the dirstate path is
1268
 
            # correct.
1269
 
            return entry[1][parent_index][1]
 
1947
            target = entry[1][parent_index][1]
 
1948
            target = target.decode('utf8')
 
1949
            return target
1270
1950
 
1271
1951
    def get_revision_id(self):
1272
1952
        """Return the revision id for this tree."""
1273
1953
        return self._revision_id
1274
1954
 
1275
 
    def _get_inventory(self):
 
1955
    def _get_root_inventory(self):
1276
1956
        if self._inventory is not None:
1277
1957
            return self._inventory
 
1958
        self._must_be_locked()
1278
1959
        self._generate_inventory()
1279
1960
        return self._inventory
1280
1961
 
 
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
 
1281
1969
    inventory = property(_get_inventory,
1282
1970
                         doc="Inventory of this Tree")
1283
1971
 
1289
1977
        return bool(self.path2id(filename))
1290
1978
 
1291
1979
    def kind(self, file_id):
1292
 
        return self.inventory[file_id].kind
 
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)
1293
2003
 
1294
2004
    def is_executable(self, file_id, path=None):
1295
 
        ie = self.inventory[file_id]
 
2005
        inv, inv_file_id = self._unpack_file_id(file_id)
 
2006
        ie = inv[inv_file_id]
1296
2007
        if ie.kind != "file":
1297
 
            return None
 
2008
            return False
1298
2009
        return ie.executable
1299
2010
 
1300
 
    def list_files(self, include_root=False):
 
2011
    def is_locked(self):
 
2012
        return self._locked
 
2013
 
 
2014
    def list_files(self, include_root=False, from_dir=None, recursive=True):
1301
2015
        # We use a standard implementation, because DirStateRevisionTree is
1302
2016
        # dealing with one of the parents of the current state
1303
 
        inv = self._get_inventory()
1304
 
        entries = inv.iter_entries()
1305
 
        if self.inventory.root is not None and not include_root:
 
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:
1306
2028
            entries.next()
1307
2029
        for path, entry in entries:
1308
2030
            yield path, 'V', entry.kind, entry.file_id, entry
1309
2031
 
1310
2032
    def lock_read(self):
1311
 
        """Lock the tree for a set of operations."""
 
2033
        """Lock the tree for a set of operations.
 
2034
 
 
2035
        :return: A bzrlib.lock.LogicalLockResult.
 
2036
        """
1312
2037
        if not self._locked:
1313
2038
            self._repository.lock_read()
1314
2039
            if self._dirstate._lock_token is None:
1315
2040
                self._dirstate.lock_read()
1316
2041
                self._dirstate_locked = True
1317
2042
        self._locked += 1
 
2043
        return LogicalLockResult(self.unlock)
 
2044
 
 
2045
    def _must_be_locked(self):
 
2046
        if not self._locked:
 
2047
            raise errors.ObjectNotLocked(self)
1318
2048
 
1319
2049
    @needs_read_lock
1320
2050
    def path2id(self, path):
1321
2051
        """Return the id for path in this tree."""
1322
2052
        # 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)
1323
2057
        entry = self._get_entry(path=path)
1324
2058
        if entry == (None, None):
1325
2059
            return None
1337
2071
                self._dirstate_locked = False
1338
2072
            self._repository.unlock()
1339
2073
 
 
2074
    @needs_read_lock
 
2075
    def supports_tree_reference(self):
 
2076
        return self._repo_supports_tree_reference
 
2077
 
1340
2078
    def walkdirs(self, prefix=""):
1341
 
        # TODO: jam 20070215 This is the cheap way by cheating and using the
1342
 
        #       RevisionTree implementation.
1343
 
        #       This should be cleaned up to use the much faster Dirstate code
1344
 
        #       This is a little tricky, though, because the dirstate is
1345
 
        #       indexed by current path, not by parent path.
1346
 
        #       So for now, we just build up the parent inventory, and extract
1347
 
        #       it the same way RevisionTree does.
 
2079
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
2080
        # implementation based on an inventory.
 
2081
        # This should be cleaned up to use the much faster Dirstate code
 
2082
        # So for now, we just build up the parent inventory, and extract
 
2083
        # it the same way RevisionTree does.
1348
2084
        _directory = 'directory'
1349
 
        inv = self._get_inventory()
 
2085
        inv = self._get_root_inventory()
1350
2086
        top_id = inv.path2id(prefix)
1351
2087
        if top_id is None:
1352
2088
            pending = []
1375
2111
 
1376
2112
 
1377
2113
class InterDirStateTree(InterTree):
1378
 
    """Fast path optimiser for changes_from with dirstate trees."""
 
2114
    """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
 
2118
    the current working tree of the same dirstate.
 
2119
    """
 
2120
    # this could be generalized to allow comparisons between any trees in the
 
2121
    # dirstate, and possibly between trees stored in different dirstates.
1379
2122
 
1380
2123
    def __init__(self, source, target):
1381
2124
        super(InterDirStateTree, self).__init__(source, target)
1386
2129
    def make_source_parent_tree(source, target):
1387
2130
        """Change the source tree into a parent of the target."""
1388
2131
        revid = source.commit('record tree')
1389
 
        target.branch.repository.fetch(source.branch.repository, revid)
 
2132
        target.branch.fetch(source.branch, revid)
1390
2133
        target.set_parent_ids([revid])
1391
2134
        return target.basis_tree(), target
1392
2135
 
 
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
 
1393
2153
    _matching_from_tree_format = WorkingTreeFormat4()
1394
2154
    _matching_to_tree_format = WorkingTreeFormat4()
1395
 
    _test_mutable_trees_to_test_trees = make_source_parent_tree
1396
 
 
1397
 
    def _iter_changes(self, include_unchanged=False,
 
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,
1398
2163
                      specific_files=None, pb=None, extra_trees=[],
1399
 
                      require_versioned=True):
 
2164
                      require_versioned=True, want_unversioned=False):
1400
2165
        """Return the changes from source to target.
1401
2166
 
1402
 
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
2167
        :return: An iterator that yields tuples. See InterTree.iter_changes
1403
2168
            for details.
1404
2169
        :param specific_files: An optional list of file paths to restrict the
1405
2170
            comparison to. When mapping filenames to ids, all matches in all
1412
2177
        :param require_versioned: If True, all files in specific_files must be
1413
2178
            versioned in one of source, target, extra_trees or
1414
2179
            PathsNotVersionedError is raised.
 
2180
        :param want_unversioned: Should unversioned files be returned in the
 
2181
            output. An unversioned file is defined as one with (False, False)
 
2182
            for the versioned pair.
1415
2183
        """
1416
 
        utf8_decode = cache_utf8._utf8_decode
1417
 
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
1418
 
        # NB: show_status depends on being able to pass in non-versioned files
1419
 
        # and report them as unknown
1420
 
            # TODO: handle extra trees in the dirstate.
1421
 
        if extra_trees:
1422
 
            for f in super(InterDirStateTree, self)._iter_changes(
 
2184
        # TODO: handle extra trees in the dirstate.
 
2185
        if (extra_trees or specific_files == []):
 
2186
            # we can't fast-path these cases (yet)
 
2187
            return super(InterDirStateTree, self).iter_changes(
1423
2188
                include_unchanged, specific_files, pb, extra_trees,
1424
 
                require_versioned):
1425
 
                yield f
1426
 
            return
 
2189
                require_versioned, want_unversioned=want_unversioned)
1427
2190
        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))
1428
2197
        target_index = 0
1429
 
        if self.source._revision_id == NULL_REVISION:
 
2198
        if self.source._revision_id == _mod_revision.NULL_REVISION:
1430
2199
            source_index = None
1431
2200
            indices = (target_index,)
1432
2201
        else:
1433
 
            assert (self.source._revision_id in parent_ids), \
1434
 
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
1435
 
                self.source._revision_id, parent_ids)
 
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))
1436
2206
            source_index = 1 + parent_ids.index(self.source._revision_id)
1437
 
            indices = (source_index,target_index)
 
2207
            indices = (source_index, target_index)
1438
2208
        # -- make all specific_files utf8 --
1439
2209
        if specific_files:
1440
2210
            specific_files_utf8 = set()
1441
2211
            for path in specific_files:
 
2212
                # Note, if there are many specific files, using cache_utf8
 
2213
                # would be good here.
1442
2214
                specific_files_utf8.add(path.encode('utf8'))
1443
2215
            specific_files = specific_files_utf8
1444
2216
        else:
1445
2217
            specific_files = set([''])
1446
2218
        # -- specific_files is now a utf8 path set --
 
2219
 
1447
2220
        # -- get the state object and prepare it.
1448
2221
        state = self.target.current_dirstate()
1449
2222
        state._read_dirblocks_if_needed()
1450
 
        def _entries_for_path(path):
1451
 
            """Return a list with all the entries that match path for all ids.
1452
 
            """
1453
 
            dirname, basename = os.path.split(path)
1454
 
            key = (dirname, basename, '')
1455
 
            block_index, present = state._find_block_index_from_key(key)
1456
 
            if not present:
1457
 
                # the block which should contain path is absent.
1458
 
                return []
1459
 
            result = []
1460
 
            block = state._dirblocks[block_index][1]
1461
 
            entry_index, _ = state._find_entry_index(key, block)
1462
 
            # we may need to look at multiple entries at this path: walk while the specific_files match.
1463
 
            while (entry_index < len(block) and
1464
 
                block[entry_index][0][0:2] == key[0:2]):
1465
 
                result.append(block[entry_index])
1466
 
                entry_index += 1
1467
 
            return result
1468
2223
        if require_versioned:
1469
2224
            # -- check all supplied paths are versioned in a search tree. --
1470
 
            all_versioned = True
 
2225
            not_versioned = []
1471
2226
            for path in specific_files:
1472
 
                path_entries = _entries_for_path(path)
 
2227
                path_entries = state._entries_for_path(path)
1473
2228
                if not path_entries:
1474
2229
                    # this specified path is not present at all: error
1475
 
                    all_versioned = False
1476
 
                    break
 
2230
                    not_versioned.append(path.decode('utf-8'))
 
2231
                    continue
1477
2232
                found_versioned = False
1478
2233
                # for each id at this path
1479
2234
                for entry in path_entries:
1486
2241
                if not found_versioned:
1487
2242
                    # none of the indexes was not 'absent' at all ids for this
1488
2243
                    # path.
1489
 
                    all_versioned = False
1490
 
                    break
1491
 
            if not all_versioned:
1492
 
                raise errors.PathsNotVersionedError(paths)
 
2244
                    not_versioned.append(path.decode('utf-8'))
 
2245
            if len(not_versioned) > 0:
 
2246
                raise errors.PathsNotVersionedError(not_versioned)
1493
2247
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
1494
 
        search_specific_files = set()
1495
 
        for path in specific_files:
1496
 
            other_specific_files = specific_files.difference(set([path]))
1497
 
            if not osutils.is_inside_any(other_specific_files, path):
1498
 
                # this is a top level path, we must check it.
1499
 
                search_specific_files.add(path)
1500
 
        # sketch: 
1501
 
        # compare source_index and target_index at or under each element of search_specific_files.
1502
 
        # follow the following comparison table. Note that we only want to do diff operations when
1503
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1504
 
        # for the target.
1505
 
        # cases:
1506
 
        # 
1507
 
        # Source | Target | disk | action
1508
 
        #   r    | fdl    |      | add source to search, add id path move and perform
1509
 
        #        |        |      | diff check on source-target
1510
 
        #   r    | fdl    |  a   | dangling file that was present in the basis. 
1511
 
        #        |        |      | ???
1512
 
        #   r    |  a     |      | add source to search
1513
 
        #   r    |  a     |  a   | 
1514
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1515
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1516
 
        #   a    | fdl    |      | add new id
1517
 
        #   a    | fdl    |  a   | dangling locally added file, skip
1518
 
        #   a    |  a     |      | not present in either tree, skip
1519
 
        #   a    |  a     |  a   | not present in any tree, skip
1520
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1521
 
        #        |        |      | may not be selected by the users list of paths.
1522
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1523
 
        #        |        |      | may not be selected by the users list of paths.
1524
 
        #  fdl   | fdl    |      | content in both: diff them
1525
 
        #  fdl   | fdl    |  a   | deleted locally, but not unversioned - show as deleted ?
1526
 
        #  fdl   |  a     |      | unversioned: output deleted id for now
1527
 
        #  fdl   |  a     |  a   | unversioned and deleted: output deleted id
1528
 
        #  fdl   |  r     |      | relocated in this tree, so add target to search.
1529
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1530
 
        #        |        |      | this id at the other path.
1531
 
        #  fdl   |  r     |  a   | relocated in this tree, so add target to search.
1532
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1533
 
        #        |        |      | this id at the other path.
1534
 
 
1535
 
        # for all search_indexs in each path at or under each element of
1536
 
        # search_specific_files, if the detail is relocated: add the id, and add the
1537
 
        # relocated path as one to search if its not searched already. If the
1538
 
        # detail is not relocated, add the id.
1539
 
        searched_specific_files = set()
1540
 
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
1541
 
        # Using a list so that we can access the values and change them in
1542
 
        # nested scope. Each one is [path, file_id, entry]
1543
 
        last_source_parent = [None, None, None]
1544
 
        last_target_parent = [None, None, None]
1545
 
 
1546
 
        def _process_entry(entry, path_info):
1547
 
            """Compare an entry and real disk to generate delta information.
1548
 
 
1549
 
            :param path_info: top_relpath, basename, kind, lstat, abspath for
1550
 
                the path of entry. If None, then the path is considered absent.
1551
 
                (Perhaps we should pass in a concrete entry for this ?)
1552
 
            """
1553
 
            # TODO: when a parent has been renamed, dont emit path renames for children,
1554
 
            if source_index is None:
1555
 
                source_details = NULL_PARENT_DETAILS
1556
 
            else:
1557
 
                source_details = entry[1][source_index]
1558
 
            target_details = entry[1][target_index]
1559
 
            source_minikind = source_details[0]
1560
 
            target_minikind = target_details[0]
1561
 
            if source_minikind in 'fdlr' and target_minikind in 'fdl':
1562
 
                # claimed content in both: diff
1563
 
                #   r    | fdl    |      | add source to search, add id path move and perform
1564
 
                #        |        |      | diff check on source-target
1565
 
                #   r    | fdl    |  a   | dangling file that was present in the basis.
1566
 
                #        |        |      | ???
1567
 
                if source_minikind in 'r':
1568
 
                    # add the source to the search path to find any children it
1569
 
                    # has.  TODO ? : only add if it is a container ?
1570
 
                    if not osutils.is_inside_any(searched_specific_files,
1571
 
                                                 source_details[1]):
1572
 
                        search_specific_files.add(source_details[1])
1573
 
                    # generate the old path; this is needed for stating later
1574
 
                    # as well.
1575
 
                    old_path = source_details[1]
1576
 
                    old_dirname, old_basename = os.path.split(old_path)
1577
 
                    path = pathjoin(entry[0][0], entry[0][1])
1578
 
                    old_entry = state._get_entry(source_index,
1579
 
                                                 path_utf8=old_path)
1580
 
                    # update the source details variable to be the real
1581
 
                    # location.
1582
 
                    source_details = old_entry[1][source_index]
1583
 
                    source_minikind = source_details[0]
1584
 
                else:
1585
 
                    old_dirname = entry[0][0]
1586
 
                    old_basename = entry[0][1]
1587
 
                    old_path = path = pathjoin(old_dirname, old_basename)
1588
 
                if path_info is None:
1589
 
                    # the file is missing on disk, show as removed.
1590
 
                    old_path = pathjoin(entry[0][0], entry[0][1])
1591
 
                    content_change = True
1592
 
                    target_kind = None
1593
 
                    target_exec = False
1594
 
                else:
1595
 
                    # source and target are both versioned and disk file is present.
1596
 
                    target_kind = path_info[2]
1597
 
                    if target_kind == 'directory':
1598
 
                        if source_minikind != 'd':
1599
 
                            content_change = True
1600
 
                        else:
1601
 
                            # directories have no fingerprint
1602
 
                            content_change = False
1603
 
                        target_exec = False
1604
 
                    elif target_kind == 'file':
1605
 
                        if source_minikind != 'f':
1606
 
                            content_change = True
1607
 
                        else:
1608
 
                            # has it changed? fast path: size, slow path: sha1.
1609
 
                            if source_details[2] != path_info[3].st_size:
1610
 
                                content_change = True
1611
 
                            else:
1612
 
                                # maybe the same. Get the hash
1613
 
                                new_hash = self.target._hashcache.get_sha1(
1614
 
                                                            path, path_info[3])
1615
 
                                content_change = (new_hash != source_details[1])
1616
 
                        target_exec = bool(
1617
 
                            stat.S_ISREG(path_info[3].st_mode)
1618
 
                            and stat.S_IEXEC & path_info[3].st_mode)
1619
 
                    elif target_kind == 'symlink':
1620
 
                        if source_minikind != 'l':
1621
 
                            content_change = True
1622
 
                        else:
1623
 
                            # TODO: check symlink supported for windows users
1624
 
                            # and grab from target state here.
1625
 
                            link_target = os.readlink(path_info[4])
1626
 
                            content_change = (link_target != source_details[1])
1627
 
                        target_exec = False
1628
 
                    else:
1629
 
                        raise Exception, "unknown kind %s" % path_info[2]
1630
 
                # parent id is the entry for the path in the target tree
1631
 
                if old_dirname == last_source_parent[0]:
1632
 
                    source_parent_id = last_source_parent[1]
1633
 
                else:
1634
 
                    source_parent_entry = state._get_entry(source_index,
1635
 
                                                           path_utf8=old_dirname)
1636
 
                    source_parent_id = source_parent_entry[0][2]
1637
 
                    if source_parent_id == entry[0][2]:
1638
 
                        # This is the root, so the parent is None
1639
 
                        source_parent_id = None
1640
 
                    else:
1641
 
                        last_source_parent[0] = old_dirname
1642
 
                        last_source_parent[1] = source_parent_id
1643
 
                        last_source_parent[2] = source_parent_entry
1644
 
 
1645
 
                new_dirname = entry[0][0]
1646
 
                if new_dirname == last_target_parent[0]:
1647
 
                    target_parent_id = last_target_parent[1]
1648
 
                else:
1649
 
                    # TODO: We don't always need to do the lookup, because the
1650
 
                    #       parent entry will be the same as the source entry.
1651
 
                    target_parent_entry = state._get_entry(target_index,
1652
 
                                                           path_utf8=new_dirname)
1653
 
                    target_parent_id = target_parent_entry[0][2]
1654
 
                    if target_parent_id == entry[0][2]:
1655
 
                        # This is the root, so the parent is None
1656
 
                        target_parent_id = None
1657
 
                    else:
1658
 
                        last_target_parent[0] = new_dirname
1659
 
                        last_target_parent[1] = target_parent_id
1660
 
                        last_target_parent[2] = target_parent_entry
1661
 
 
1662
 
                source_exec = source_details[3]
1663
 
                path_unicode = utf8_decode(path)[0]
1664
 
                return ((entry[0][2], path_unicode, content_change,
1665
 
                        (True, True),
1666
 
                        (source_parent_id, target_parent_id),
1667
 
                        (old_basename, entry[0][1]),
1668
 
                        (_minikind_to_kind[source_minikind], target_kind),
1669
 
                        (source_exec, target_exec)),)
1670
 
            elif source_minikind in 'a' and target_minikind in 'fdl':
1671
 
                # looks like a new file
1672
 
                if path_info is not None:
1673
 
                    path = pathjoin(entry[0][0], entry[0][1])
1674
 
                    # parent id is the entry for the path in the target tree
1675
 
                    # TODO: these are the same for an entire directory: cache em.
1676
 
                    parent_id = state._get_entry(target_index, path_utf8=entry[0][0])[0][2]
1677
 
                    if parent_id == entry[0][2]:
1678
 
                        parent_id = None
1679
 
                    # basename
1680
 
                    new_executable = bool(
1681
 
                        stat.S_ISREG(path_info[3].st_mode)
1682
 
                        and stat.S_IEXEC & path_info[3].st_mode)
1683
 
                    path_unicode = utf8_decode(path)[0]
1684
 
                    return ((entry[0][2], path_unicode, True,
1685
 
                            (False, True),
1686
 
                            (None, parent_id),
1687
 
                            (None, entry[0][1]),
1688
 
                            (None, path_info[2]),
1689
 
                            (None, new_executable)),)
1690
 
                else:
1691
 
                    # but its not on disk: we deliberately treat this as just
1692
 
                    # never-present. (Why ?! - RBC 20070224)
1693
 
                    pass
1694
 
            elif source_minikind in 'fdl' and target_minikind in 'a':
1695
 
                # unversioned, possibly, or possibly not deleted: we dont care.
1696
 
                # if its still on disk, *and* theres no other entry at this
1697
 
                # path [we dont know this in this routine at the moment -
1698
 
                # perhaps we should change this - then it would be an unknown.
1699
 
                old_path = pathjoin(entry[0][0], entry[0][1])
1700
 
                # parent id is the entry for the path in the target tree
1701
 
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
1702
 
                if parent_id == entry[0][2]:
1703
 
                    parent_id = None
1704
 
                old_path_unicode = utf8_decode(old_path)[0]
1705
 
                return ((entry[0][2], old_path_unicode, True,
1706
 
                        (True, False),
1707
 
                        (parent_id, None),
1708
 
                        (entry[0][1], None),
1709
 
                        (_minikind_to_kind[source_minikind], None),
1710
 
                        (source_details[3], None)),)
1711
 
            elif source_minikind in 'fdl' and target_minikind in 'r':
1712
 
                # a rename; could be a true rename, or a rename inherited from
1713
 
                # a renamed parent. TODO: handle this efficiently. Its not
1714
 
                # common case to rename dirs though, so a correct but slow
1715
 
                # implementation will do.
1716
 
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
1717
 
                    search_specific_files.add(target_details[1])
1718
 
            elif source_minikind in 'r' and target_minikind in 'r':
1719
 
                # neither of the selected trees contain this file,
1720
 
                # so skip over it. This is not currently directly tested, but
1721
 
                # is indirectly via test_too_much.TestCommands.test_conflicts.
1722
 
                pass
1723
 
            else:
1724
 
                print "*******", source_minikind, target_minikind
1725
 
                import pdb;pdb.set_trace()
1726
 
            return ()
1727
 
        while search_specific_files:
1728
 
            # TODO: the pending list should be lexically sorted?
1729
 
            current_root = search_specific_files.pop()
1730
 
            searched_specific_files.add(current_root)
1731
 
            # process the entries for this containing directory: the rest will be
1732
 
            # found by their parents recursively.
1733
 
            root_entries = _entries_for_path(current_root)
1734
 
            root_abspath = self.target.abspath(current_root)
1735
 
            try:
1736
 
                root_stat = os.lstat(root_abspath)
1737
 
            except OSError, e:
1738
 
                if e.errno == errno.ENOENT:
1739
 
                    # the path does not exist: let _process_entry know that.
1740
 
                    root_dir_info = None
1741
 
                else:
1742
 
                    # some other random error: hand it up.
1743
 
                    raise
1744
 
            else:
1745
 
                root_dir_info = ('', current_root,
1746
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1747
 
                    root_abspath)
1748
 
            if not root_entries and not root_dir_info:
1749
 
                # this specified path is not present at all, skip it.
1750
 
                continue
1751
 
            for entry in root_entries:
1752
 
                for result in _process_entry(entry, root_dir_info):
1753
 
                    # this check should probably be outside the loop: one
1754
 
                    # 'iterate two trees' api, and then _iter_changes filters
1755
 
                    # unchanged pairs. - RBC 20070226
1756
 
                    if (include_unchanged
1757
 
                        or result[2]                    # content change
1758
 
                        or result[3][0] != result[3][1] # versioned status
1759
 
                        or result[4][0] != result[4][1] # parent id
1760
 
                        or result[5][0] != result[5][1] # name
1761
 
                        or result[6][0] != result[6][1] # kind
1762
 
                        or result[7][0] != result[7][1] # executable
1763
 
                        ):
1764
 
                        yield result
1765
 
            dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
1766
 
            initial_key = (current_root, '', '')
1767
 
            block_index, _ = state._find_block_index_from_key(initial_key)
1768
 
            if block_index == 0:
1769
 
                # we have processed the total root already, but because the
1770
 
                # initial key matched it we should skip it here.
1771
 
                block_index +=1
1772
 
            try:
1773
 
                current_dir_info = dir_iterator.next()
1774
 
            except OSError, e:
1775
 
                if e.errno in (errno.ENOENT, errno.ENOTDIR):
1776
 
                    # there may be directories in the inventory even though
1777
 
                    # this path is not a file on disk: so mark it as end of
1778
 
                    # iterator
1779
 
                    current_dir_info = None
1780
 
                else:
1781
 
                    raise
1782
 
            else:
1783
 
                if current_dir_info[0][0] == '':
1784
 
                    # remove .bzr from iteration
1785
 
                    bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
1786
 
                    assert current_dir_info[1][bzr_index][0] == '.bzr'
1787
 
                    del current_dir_info[1][bzr_index]
1788
 
            # walk until both the directory listing and the versioned metadata
1789
 
            # are exhausted. TODO: reevaluate this, perhaps we should stop when
1790
 
            # the versioned data runs out.
1791
 
            if (block_index < len(state._dirblocks) and
1792
 
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
1793
 
                current_block = state._dirblocks[block_index]
1794
 
            else:
1795
 
                current_block = None
1796
 
            while (current_dir_info is not None or
1797
 
                   current_block is not None):
1798
 
                if (current_dir_info and current_block
1799
 
                    and current_dir_info[0][0] != current_block[0]):
1800
 
                    if current_dir_info[0][0] < current_block[0] :
1801
 
                        # import pdb; pdb.set_trace()
1802
 
                        # print 'unversioned dir'
1803
 
                        # filesystem data refers to paths not covered by the dirblock.
1804
 
                        # this has two possibilities:
1805
 
                        # A) it is versioned but empty, so there is no block for it
1806
 
                        # B) it is not versioned.
1807
 
                        # in either case it was processed by the containing directories walk:
1808
 
                        # if it is root/foo, when we walked root we emitted it,
1809
 
                        # or if we ere given root/foo to walk specifically, we
1810
 
                        # emitted it when checking the walk-root entries
1811
 
                        # advance the iterator and loop - we dont need to emit it.
1812
 
                        try:
1813
 
                            current_dir_info = dir_iterator.next()
1814
 
                        except StopIteration:
1815
 
                            current_dir_info = None
1816
 
                    else:
1817
 
                        # We have a dirblock entry for this location, but there
1818
 
                        # is no filesystem path for this. This is most likely
1819
 
                        # because a directory was removed from the disk.
1820
 
                        # We don't have to report the missing directory,
1821
 
                        # because that should have already been handled, but we
1822
 
                        # need to handle all of the files that are contained
1823
 
                        # within.
1824
 
                        for current_entry in current_block[1]:
1825
 
                            # entry referring to file not present on disk.
1826
 
                            # advance the entry only, after processing.
1827
 
                            for result in _process_entry(current_entry, None):
1828
 
                                # this check should probably be outside the loop: one
1829
 
                                # 'iterate two trees' api, and then _iter_changes filters
1830
 
                                # unchanged pairs. - RBC 20070226
1831
 
                                if (include_unchanged
1832
 
                                    or result[2]                    # content change
1833
 
                                    or result[3][0] != result[3][1] # versioned status
1834
 
                                    or result[4][0] != result[4][1] # parent id
1835
 
                                    or result[5][0] != result[5][1] # name
1836
 
                                    or result[6][0] != result[6][1] # kind
1837
 
                                    or result[7][0] != result[7][1] # executable
1838
 
                                    ):
1839
 
                                    yield result
1840
 
                        block_index +=1
1841
 
                        if (block_index < len(state._dirblocks) and
1842
 
                            osutils.is_inside(current_root,
1843
 
                                              state._dirblocks[block_index][0])):
1844
 
                            current_block = state._dirblocks[block_index]
1845
 
                        else:
1846
 
                            current_block = None
1847
 
                    continue
1848
 
                entry_index = 0
1849
 
                if current_block and entry_index < len(current_block[1]):
1850
 
                    current_entry = current_block[1][entry_index]
1851
 
                else:
1852
 
                    current_entry = None
1853
 
                advance_entry = True
1854
 
                path_index = 0
1855
 
                if current_dir_info and path_index < len(current_dir_info[1]):
1856
 
                    current_path_info = current_dir_info[1][path_index]
1857
 
                else:
1858
 
                    current_path_info = None
1859
 
                advance_path = True
1860
 
                while (current_entry is not None or
1861
 
                    current_path_info is not None):
1862
 
                    if current_entry is None:
1863
 
                        # no more entries: yield current_pathinfo as an
1864
 
                        # unversioned file: its not the same as a path in any
1865
 
                        # tree in the dirstate.
1866
 
                        new_executable = bool(
1867
 
                            stat.S_ISREG(current_path_info[3].st_mode)
1868
 
                            and stat.S_IEXEC & current_path_info[3].st_mode)
1869
 
                        yield (None, current_path_info[0], True,
1870
 
                               (False, False),
1871
 
                               (None, None),
1872
 
                               (None, current_path_info[1]),
1873
 
                               (None, current_path_info[2]),
1874
 
                               (None, new_executable))
1875
 
                    elif current_path_info is None:
1876
 
                        # no path is fine: the per entry code will handle it.
1877
 
                        for result in _process_entry(current_entry, current_path_info):
1878
 
                            # this check should probably be outside the loop: one
1879
 
                            # 'iterate two trees' api, and then _iter_changes filters
1880
 
                            # unchanged pairs. - RBC 20070226
1881
 
                            if (include_unchanged
1882
 
                                or result[2]                    # content change
1883
 
                                or result[3][0] != result[3][1] # versioned status
1884
 
                                or result[4][0] != result[4][1] # parent id
1885
 
                                or result[5][0] != result[5][1] # name
1886
 
                                or result[6][0] != result[6][1] # kind
1887
 
                                or result[7][0] != result[7][1] # executable
1888
 
                                ):
1889
 
                                yield result
1890
 
                    elif current_entry[0][1] != current_path_info[1]:
1891
 
                        if current_path_info[1] < current_entry[0][1]:
1892
 
                            # extra file on disk: pass for now, but only
1893
 
                            # increment the path, not the entry
1894
 
                            # import pdb; pdb.set_trace()
1895
 
                            # print 'unversioned file'
1896
 
                            advance_entry = False
1897
 
                        else:
1898
 
                            # entry referring to file not present on disk.
1899
 
                            # advance the entry only, after processing.
1900
 
                            for result in _process_entry(current_entry, None):
1901
 
                                # this check should probably be outside the loop: one
1902
 
                                # 'iterate two trees' api, and then _iter_changes filters
1903
 
                                # unchanged pairs. - RBC 20070226
1904
 
                                if (include_unchanged
1905
 
                                    or result[2]                    # content change
1906
 
                                    or result[3][0] != result[3][1] # versioned status
1907
 
                                    or result[4][0] != result[4][1] # parent id
1908
 
                                    or result[5][0] != result[5][1] # name
1909
 
                                    or result[6][0] != result[6][1] # kind
1910
 
                                    or result[7][0] != result[7][1] # executable
1911
 
                                    ):
1912
 
                                    yield result
1913
 
                            advance_path = False
1914
 
                    else:
1915
 
                        for result in _process_entry(current_entry, current_path_info):
1916
 
                            # this check should probably be outside the loop: one
1917
 
                            # 'iterate two trees' api, and then _iter_changes filters
1918
 
                            # unchanged pairs. - RBC 20070226
1919
 
                            if (include_unchanged
1920
 
                                or result[2]                    # content change
1921
 
                                or result[3][0] != result[3][1] # versioned status
1922
 
                                or result[4][0] != result[4][1] # parent id
1923
 
                                or result[5][0] != result[5][1] # name
1924
 
                                or result[6][0] != result[6][1] # kind
1925
 
                                or result[7][0] != result[7][1] # executable
1926
 
                                ):
1927
 
                                yield result
1928
 
                    if advance_entry and current_entry is not None:
1929
 
                        entry_index += 1
1930
 
                        if entry_index < len(current_block[1]):
1931
 
                            current_entry = current_block[1][entry_index]
1932
 
                        else:
1933
 
                            current_entry = None
1934
 
                    else:
1935
 
                        advance_entry = True # reset the advance flaga
1936
 
                    if advance_path and current_path_info is not None:
1937
 
                        path_index += 1
1938
 
                        if path_index < len(current_dir_info[1]):
1939
 
                            current_path_info = current_dir_info[1][path_index]
1940
 
                        else:
1941
 
                            current_path_info = None
1942
 
                    else:
1943
 
                        advance_path = True # reset the advance flagg.
1944
 
                if current_block is not None:
1945
 
                    block_index += 1
1946
 
                    if (block_index < len(state._dirblocks) and
1947
 
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
1948
 
                        current_block = state._dirblocks[block_index]
1949
 
                    else:
1950
 
                        current_block = None
1951
 
                if current_dir_info is not None:
1952
 
                    try:
1953
 
                        current_dir_info = dir_iterator.next()
1954
 
                    except StopIteration:
1955
 
                        current_dir_info = None
1956
 
 
 
2248
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
2249
 
 
2250
        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()
1957
2255
 
1958
2256
    @staticmethod
1959
2257
    def is_compatible(source, target):
1960
2258
        # the target must be a dirstate working tree
1961
 
        if not isinstance(target, WorkingTree4):
 
2259
        if not isinstance(target, DirStateWorkingTree):
1962
2260
            return False
1963
 
        # the source must be a revtreee or dirstate rev tree.
 
2261
        # the source must be a revtree or dirstate rev tree.
1964
2262
        if not isinstance(source,
1965
2263
            (revisiontree.RevisionTree, DirStateRevisionTree)):
1966
2264
            return False
1967
2265
        # the source revid must be in the target dirstate
1968
 
        if not (source._revision_id == NULL_REVISION or
 
2266
        if not (source._revision_id == _mod_revision.NULL_REVISION or
1969
2267
            source._revision_id in target.get_parent_ids()):
1970
 
            # TODO: what about ghosts? it may well need to 
 
2268
            # TODO: what about ghosts? it may well need to
1971
2269
            # check for them explicitly.
1972
2270
            return False
1973
2271
        return True
1974
2272
 
1975
2273
InterTree.register_optimiser(InterDirStateTree)
 
2274
 
 
2275
 
 
2276
class Converter3to4(object):
 
2277
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2278
 
 
2279
    def __init__(self):
 
2280
        self.target_format = WorkingTreeFormat4()
 
2281
 
 
2282
    def convert(self, tree):
 
2283
        # 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
 
2285
        # tree during upgrade.
 
2286
        tree._control_files.lock_write()
 
2287
        try:
 
2288
            tree.read_working_inventory()
 
2289
            self.create_dirstate_data(tree)
 
2290
            self.update_format(tree)
 
2291
            self.remove_xml_files(tree)
 
2292
        finally:
 
2293
            tree._control_files.unlock()
 
2294
 
 
2295
    def create_dirstate_data(self, tree):
 
2296
        """Create the dirstate based data for tree."""
 
2297
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2298
            ).local_abspath('dirstate')
 
2299
        state = dirstate.DirState.from_tree(tree, local_path)
 
2300
        state.save()
 
2301
        state.unlock()
 
2302
 
 
2303
    def remove_xml_files(self, tree):
 
2304
        """Remove the oldformat 3 data."""
 
2305
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2306
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2307
            'pending-merges', 'stat-cache']:
 
2308
            try:
 
2309
                transport.delete(path)
 
2310
            except errors.NoSuchFile:
 
2311
                # some files are optional - just deal.
 
2312
                pass
 
2313
 
 
2314
    def update_format(self, tree):
 
2315
        """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())