1
# Copyright (C) 2005, 2006 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""WorkingTree4 format and implementation.
19
WorkingTree4 provides the dirstate based working tree logic.
21
To get a WorkingTree, call bzrdir.open_workingtree() or
22
WorkingTree.open(dir).
25
from cStringIO import StringIO
29
from bzrlib.lazy_import import lazy_import
30
lazy_import(globals(), """
31
from bisect import bisect_left
33
from copy import deepcopy
45
conflicts as _mod_conflicts,
63
from bzrlib.transport import get_transport
67
from bzrlib import symbol_versioning
68
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
71
from bzrlib.lockdir import LockDir
72
import bzrlib.mutabletree
73
from bzrlib.mutabletree import needs_tree_write_lock
74
from bzrlib.osutils import (
86
from bzrlib.trace import mutter, note
87
from bzrlib.transport.local import LocalTransport
88
from bzrlib.tree import InterTree
89
from bzrlib.progress import DummyProgress, ProgressPhase
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
91
from bzrlib.rio import RioReader, rio_file, Stanza
92
from bzrlib.symbol_versioning import (deprecated_passed,
100
from bzrlib.tree import Tree
101
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
104
class WorkingTree4(WorkingTree3):
105
"""This is the Format 4 working tree.
107
This differs from WorkingTree3 by:
108
- having a consolidated internal dirstate.
109
- not having a regular inventory attribute.
111
This is new in bzr TODO FIXME SETMEBEFORE MERGE.
114
def __init__(self, basedir,
119
"""Construct a WorkingTree for basedir.
121
If the branch is not supplied, it is opened automatically.
122
If the branch is supplied, it must be the branch for this basedir.
123
(branch.base is not cross checked, because for remote branches that
124
would be meaningless).
126
self._format = _format
127
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
basedir = safe_unicode(basedir)
133
mutter("opening working tree %r", basedir)
134
self._branch = branch
135
assert isinstance(self.branch, bzrlib.branch.Branch), \
136
"branch %r is not a Branch" % self.branch
137
self.basedir = realpath(basedir)
138
# if branch is at our basedir and is a format 6 or less
139
# 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
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
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)
152
# is this scan needed ? it makes things kinda slow.
161
# during a read or write lock these objects are set, and are
162
# None the rest of the time.
163
self._dirstate = None
164
self._inventory = None
167
@needs_tree_write_lock
168
def _add(self, files, ids, kinds):
169
"""See MutableTree._add."""
170
state = self.current_dirstate()
171
for f, file_id, kind in zip(files, ids, kinds):
176
file_id = generate_ids.gen_file_id(f)
177
# deliberately add the file with no cached stat or sha1
178
# - on the first access it will be gathered, and we can
179
# always change this once tests are all passing.
180
state.add(f, file_id, kind, None, '')
183
def break_lock(self):
184
"""Break a lock if one is present from another instance.
186
Uses the ui factory to ask for confirmation if the lock may be from
189
This will probe the repository for its lock as well.
191
# if the dirstate is locked by an active process, reject the break lock
194
if self._dirstate is None:
198
state = self._current_dirstate()
199
if state._lock_token is not None:
200
# we already have it locked. sheese, cant break our own lock.
201
raise errors.LockActive(self.basedir)
204
# try for a write lock - need permission to get one anyhow
207
except errors.LockContention:
208
# oslocks fail when a process is still live: fail.
209
# TODO: get the locked lockdir info and give to the user to
210
# assist in debugging.
211
raise errors.LockActive(self.basedir)
216
self._dirstate = None
217
self._control_files.break_lock()
218
self.branch.break_lock()
220
def current_dirstate(self):
221
"""Return the current dirstate object.
223
This is not part of the tree interface and only exposed for ease of
226
:raises errors.NotWriteLocked: when not in a lock.
228
if not self._control_files._lock_count:
229
raise errors.ObjectNotLocked(self)
230
return self._current_dirstate()
232
def _current_dirstate(self):
233
"""Internal function that does not check lock status.
235
This is needed for break_lock which also needs the dirstate.
237
if self._dirstate is not None:
238
return self._dirstate
239
local_path = self.bzrdir.get_workingtree_transport(None
240
).local_abspath('dirstate')
241
self._dirstate = dirstate.DirState.on_file(local_path)
242
return self._dirstate
244
def filter_unversioned_files(self, paths):
245
"""Filter out paths that are not versioned.
247
:return: set of paths.
249
# TODO: make a generic multi-bisect routine roughly that should list
250
# the paths, then process one half at a time recursively, and feed the
251
# results of each bisect in further still
252
paths = sorted(paths)
254
state = self.current_dirstate()
255
# TODO we want a paths_to_dirblocks helper I think
257
dirname, basename = os.path.split(path.encode('utf8'))
258
_, _, _, path_is_versioned = state._get_block_entry_index(
259
dirname, basename, 0)
260
if path_is_versioned:
265
"""Write all cached data to disk."""
266
if self._control_files._lock_mode != 'w':
267
raise errors.NotWriteLocked(self)
268
self.current_dirstate().save()
269
self._inventory = None
272
def _generate_inventory(self):
273
"""Create and set self.inventory from the dirstate object.
275
This is relatively expensive: we have to walk the entire dirstate.
276
Ideally we would not, and can deprecate this function.
278
#: uncomment to trap on inventory requests.
279
# import pdb;pdb.set_trace()
280
state = self.current_dirstate()
281
state._read_dirblocks_if_needed()
282
root_key, current_entry = self._get_entry(path='')
283
current_id = root_key[2]
284
assert current_entry[0][0] == 'd' # directory
285
inv = Inventory(root_id=current_id)
286
# Turn some things into local variables
287
minikind_to_kind = dirstate.DirState._minikind_to_kind
288
factory = entry_factory
289
utf8_decode = cache_utf8._utf8_decode
291
# we could do this straight out of the dirstate; it might be fast
292
# and should be profiled - RBC 20070216
293
parent_ies = {'' : inv.root}
294
for block in state._dirblocks[1:]: # skip the root
297
parent_ie = parent_ies[dirname]
299
# all the paths in this block are not versioned in this tree
301
for key, entry in block[1]:
302
minikind, link_or_sha1, size, executable, stat = entry[0]
303
if minikind in ('a', 'r'): # absent, relocated
304
# a parent tree only entry
307
name_unicode = utf8_decode(name)[0]
309
kind = minikind_to_kind[minikind]
310
inv_entry = factory[kind](file_id, name_unicode,
313
# not strictly needed: working tree
314
#entry.executable = executable
315
#entry.text_size = size
316
#entry.text_sha1 = sha1
318
elif kind == 'directory':
319
# add this entry to the parent map.
320
parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
321
# 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
325
inv_byid[file_id] = inv_entry
326
parent_ie.children[name_unicode] = inv_entry
327
self._inventory = inv
329
def _get_entry(self, file_id=None, path=None):
330
"""Get the dirstate row for file_id or path.
332
If either file_id or path is supplied, it is used as the key to lookup.
333
If both are supplied, the fastest lookup is used, and an error is
334
raised if they do not both point at the same row.
336
:param file_id: An optional unicode file_id to be looked up.
337
:param path: An optional unicode path to be looked up.
338
:return: The dirstate row tuple for path/file_id, or (None, None)
340
if file_id is None and path is None:
341
raise errors.BzrError('must supply file_id or path')
342
state = self.current_dirstate()
344
path = path.encode('utf8')
345
return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
347
def get_file_sha1(self, file_id, path=None, stat_value=None):
348
# 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'
352
# if row stat is valid, use cached sha1, else, get a new sha1.
354
path = pathjoin(key[0], key[1]).decode('utf8')
355
return self._hashcache.get_sha1(path, stat_value)
357
def _get_inventory(self):
358
"""Get the inventory for the tree. This is only valid within a lock."""
359
if self._inventory is not None:
360
return self._inventory
361
self._generate_inventory()
362
return self._inventory
364
inventory = property(_get_inventory,
365
doc="Inventory of this Tree")
368
def get_parent_ids(self):
369
"""See Tree.get_parent_ids.
371
This implementation requests the ids list from the dirstate file.
373
return self.current_dirstate().get_parent_ids()
376
def get_root_id(self):
377
"""Return the id of this trees root"""
378
return self._get_entry(path='')[0][2]
380
def has_id(self, file_id):
381
state = self.current_dirstate()
382
file_id = osutils.safe_file_id(file_id)
383
row, parents = self._get_entry(file_id=file_id)
386
return osutils.lexists(pathjoin(
387
self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
390
def id2path(self, file_id):
391
file_id = osutils.safe_file_id(file_id)
392
state = self.current_dirstate()
393
entry = self._get_entry(file_id=file_id)
394
if entry == (None, None):
396
path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
397
return path_utf8.decode('utf8')
401
"""Iterate through file_ids for this tree.
403
file_ids are in a WorkingTree if they are in the working inventory
404
and the working file exists.
407
for key, tree_details in self.current_dirstate()._iter_entries():
408
if tree_details[0][0] in ('a', 'r'): # absent, relocated
409
# not relevant to the working tree
411
path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
412
if osutils.lexists(path):
413
result.append(key[2])
417
def _last_revision(self):
418
"""See Mutable.last_revision."""
419
parent_ids = self.current_dirstate().get_parent_ids()
426
super(WorkingTree4, self).lock_read()
427
if self._dirstate is None:
428
self.current_dirstate()
429
self._dirstate.lock_read()
431
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()
437
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()
443
@needs_tree_write_lock
444
def move(self, from_paths, to_dir, after=False):
445
"""See WorkingTree.move()."""
450
state = self.current_dirstate()
452
assert not isinstance(from_paths, basestring)
453
to_dir_utf8 = to_dir.encode('utf8')
454
to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
455
id_index = state._get_id_index()
456
# check destination directory
457
# get the details for it
458
to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
459
state._get_block_entry_index(to_entry_dirname, to_basename, 0)
460
if not entry_present:
461
raise errors.BzrMoveFailedError('', to_dir,
462
errors.NotInWorkingDirectory(to_dir))
463
to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
464
# get a handle on the block itself.
465
to_block_index = state._ensure_block(
466
to_entry_block_index, to_entry_entry_index, to_dir_utf8)
467
to_block = state._dirblocks[to_block_index]
468
to_abs = self.abspath(to_dir)
469
if not isdir(to_abs):
470
raise errors.BzrMoveFailedError('',to_dir,
471
errors.NotADirectory(to_abs))
473
if to_entry[1][0][0] != 'd':
474
raise errors.BzrMoveFailedError('',to_dir,
475
errors.NotADirectory(to_abs))
477
if self._inventory is not None:
478
update_inventory = True
480
to_dir_ie = inv[to_dir_id]
481
to_dir_id = to_entry[0][2]
483
update_inventory = False
486
def move_one(old_entry, from_path_utf8, minikind, executable,
487
fingerprint, packed_stat, size,
488
to_block, to_key, to_path_utf8):
489
state._make_absent(old_entry)
490
from_key = old_entry[0]
492
lambda:state.update_minimal(from_key,
494
executable=executable,
495
fingerprint=fingerprint,
496
packed_stat=packed_stat,
498
path_utf8=from_path_utf8))
499
state.update_minimal(to_key,
501
executable=executable,
502
fingerprint=fingerprint,
503
packed_stat=packed_stat,
505
path_utf8=to_path_utf8)
506
added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
507
new_entry = to_block[1][added_entry_index]
508
rollbacks.append(lambda:state._make_absent(new_entry))
510
# create rename entries and tuples
511
for from_rel in from_paths:
512
# from_rel is 'pathinroot/foo/bar'
513
from_rel_utf8 = from_rel.encode('utf8')
514
from_dirname, from_tail = osutils.split(from_rel)
515
from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
516
from_entry = self._get_entry(path=from_rel)
517
if from_entry == (None, None):
518
raise errors.BzrMoveFailedError(from_rel,to_dir,
519
errors.NotVersionedError(path=str(from_rel)))
521
from_id = from_entry[0][2]
522
to_rel = pathjoin(to_dir, from_tail)
523
to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
524
item_to_entry = self._get_entry(path=to_rel)
525
if item_to_entry != (None, None):
526
raise errors.BzrMoveFailedError(from_rel, to_rel,
527
"Target is already versioned.")
529
if from_rel == to_rel:
530
raise errors.BzrMoveFailedError(from_rel, to_rel,
531
"Source and target are identical.")
533
from_missing = not self.has_filename(from_rel)
534
to_missing = not self.has_filename(to_rel)
541
raise errors.BzrMoveFailedError(from_rel, to_rel,
542
errors.NoSuchFile(path=to_rel,
543
extra="New file has not been created yet"))
545
# neither path exists
546
raise errors.BzrRenameFailedError(from_rel, to_rel,
547
errors.PathsDoNotExist(paths=(from_rel, to_rel)))
549
if from_missing: # implicitly just update our path mapping
552
raise errors.RenameFailedFilesExist(from_rel, to_rel,
553
extra="(Use --after to update the Bazaar id)")
556
def rollback_rename():
557
"""A single rename has failed, roll it back."""
559
for rollback in reversed(rollbacks):
563
import pdb;pdb.set_trace()
564
exc_info = sys.exc_info()
566
raise exc_info[0], exc_info[1], exc_info[2]
568
# perform the disk move first - its the most likely failure point.
570
from_rel_abs = self.abspath(from_rel)
571
to_rel_abs = self.abspath(to_rel)
573
osutils.rename(from_rel_abs, to_rel_abs)
575
raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
576
rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
578
# perform the rename in the inventory next if needed: its easy
582
from_entry = inv[from_id]
583
current_parent = from_entry.parent_id
584
inv.rename(from_id, to_dir_id, from_tail)
586
lambda: inv.rename(from_id, current_parent, from_tail))
587
# finally do the rename in the dirstate, which is a little
588
# tricky to rollback, but least likely to need it.
589
old_block_index, old_entry_index, dir_present, file_present = \
590
state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
591
old_block = state._dirblocks[old_block_index][1]
592
old_entry = old_block[old_entry_index]
593
from_key, old_entry_details = old_entry
594
cur_details = old_entry_details[0]
596
to_key = ((to_block[0],) + from_key[1:3])
597
minikind = cur_details[0]
598
move_one(old_entry, from_path_utf8=from_rel_utf8,
600
executable=cur_details[3],
601
fingerprint=cur_details[1],
602
packed_stat=cur_details[4],
606
to_path_utf8=to_rel_utf8)
609
def update_dirblock(from_dir, to_key, to_dir_utf8):
610
"""all entries in this block need updating.
612
TODO: This is pretty ugly, and doesn't support
613
reverting, but it works.
615
assert from_dir != '', "renaming root not supported"
616
from_key = (from_dir, '')
617
from_block_idx, present = \
618
state._find_block_index_from_key(from_key)
620
# This is the old record, if it isn't present, then
621
# there is theoretically nothing to update.
622
# (Unless it isn't present because of lazy loading,
623
# but we don't do that yet)
625
from_block = state._dirblocks[from_block_idx]
626
to_block_index, to_entry_index, _, _ = \
627
state._get_block_entry_index(to_key[0], to_key[1], 0)
628
to_block_index = state._ensure_block(
629
to_block_index, to_entry_index, to_dir_utf8)
630
to_block = state._dirblocks[to_block_index]
631
for entry in from_block[1]:
632
assert entry[0][0] == from_dir
633
cur_details = entry[1][0]
634
to_key = (to_dir_utf8, entry[0][1], entry[0][2])
635
from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
636
to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
637
minikind = cur_details[0]
638
move_one(entry, from_path_utf8=from_path_utf8,
640
executable=cur_details[3],
641
fingerprint=cur_details[1],
642
packed_stat=cur_details[4],
646
to_path_utf8=to_rel_utf8)
648
# We need to move all the children of this
650
update_dirblock(from_path_utf8, to_key,
652
update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
656
result.append((from_rel, to_rel))
657
state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
663
"""Initialize the state in this tree to be a new tree."""
667
def path2id(self, path):
668
"""Return the id for path in this tree."""
669
path = path.strip('/')
670
entry = self._get_entry(path=path)
671
if entry == (None, None):
675
def paths2ids(self, paths, trees=[], require_versioned=True):
676
"""See Tree.paths2ids().
678
This specialisation fast-paths the case where all the trees are in the
683
parents = self.get_parent_ids()
685
if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
687
return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
688
search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
689
# -- make all paths utf8 --
692
paths_utf8.add(path.encode('utf8'))
694
# -- paths is now a utf8 path set --
695
# -- get the state object and prepare it.
696
state = self.current_dirstate()
697
if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
698
and '' not in paths):
699
paths2ids = self._paths2ids_using_bisect
701
paths2ids = self._paths2ids_in_memory
702
return paths2ids(paths, search_indexes,
703
require_versioned=require_versioned)
705
def _paths2ids_in_memory(self, paths, search_indexes,
706
require_versioned=True):
707
state = self.current_dirstate()
708
state._read_dirblocks_if_needed()
709
def _entries_for_path(path):
710
"""Return a list with all the entries that match path for all ids.
712
dirname, basename = os.path.split(path)
713
key = (dirname, basename, '')
714
block_index, present = state._find_block_index_from_key(key)
716
# the block which should contain path is absent.
719
block = state._dirblocks[block_index][1]
720
entry_index, _ = state._find_entry_index(key, block)
721
# we may need to look at multiple entries at this path: walk while the paths match.
722
while (entry_index < len(block) and
723
block[entry_index][0][0:2] == key[0:2]):
724
result.append(block[entry_index])
727
if require_versioned:
728
# -- check all supplied paths are versioned in a search tree. --
731
path_entries = _entries_for_path(path)
733
# this specified path is not present at all: error
734
all_versioned = False
736
found_versioned = False
737
# for each id at this path
738
for entry in path_entries:
740
for index in search_indexes:
741
if entry[1][index][0] != 'a': # absent
742
found_versioned = True
743
# all good: found a versioned cell
745
if not found_versioned:
746
# none of the indexes was not 'absent' at all ids for this
748
all_versioned = False
750
if not all_versioned:
751
raise errors.PathsNotVersionedError(paths)
752
# -- remove redundancy in supplied paths to prevent over-scanning --
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)
760
# for all search_indexs in each path at or under each element of
761
# search_paths, if the detail is relocated: add the id, and add the
762
# relocated path as one to search if its not searched already. If the
763
# detail is not relocated, add the id.
764
searched_paths = set()
766
def _process_entry(entry):
767
"""Look at search_indexes within entry.
769
If a specific tree's details are relocated, add the relocation
770
target to search_paths if not searched already. If it is absent, do
771
nothing. Otherwise add the id to found_ids.
773
for index in search_indexes:
774
if entry[1][index][0] == 'r': # relocated
775
if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
776
search_paths.add(entry[1][index][1])
777
elif entry[1][index][0] != 'a': # absent
778
found_ids.add(entry[0][2])
780
current_root = search_paths.pop()
781
searched_paths.add(current_root)
782
# process the entries for this containing directory: the rest will be
783
# found by their parents recursively.
784
root_entries = _entries_for_path(current_root)
786
# this specified path is not present at all, skip it.
788
for entry in root_entries:
789
_process_entry(entry)
790
initial_key = (current_root, '', '')
791
block_index, _ = state._find_block_index_from_key(initial_key)
792
while (block_index < len(state._dirblocks) and
793
osutils.is_inside(current_root, state._dirblocks[block_index][0])):
794
for entry in state._dirblocks[block_index][1]:
795
_process_entry(entry)
799
def _paths2ids_using_bisect(self, paths, search_indexes,
800
require_versioned=True):
801
state = self.current_dirstate()
804
split_paths = sorted(osutils.split(p) for p in paths)
805
found = state._bisect_recursive(split_paths)
807
if require_versioned:
808
found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
809
for dir_name in split_paths:
810
if dir_name not in found_dir_names:
811
raise errors.PathsNotVersionedError(paths)
813
for dir_name_id, trees_info in found.iteritems():
814
for index in search_indexes:
815
if trees_info[index][0] not in ('r', 'a'):
816
found_ids.add(dir_name_id[2])
819
def read_working_inventory(self):
820
"""Read the working inventory.
822
This is a meaningless operation for dirstate, but we obey it anyhow.
824
return self.inventory
827
def revision_tree(self, revision_id):
828
"""See Tree.revision_tree.
830
WorkingTree4 supplies revision_trees for any basis tree.
832
revision_id = osutils.safe_revision_id(revision_id)
833
dirstate = self.current_dirstate()
834
parent_ids = dirstate.get_parent_ids()
835
if revision_id not in parent_ids:
836
raise errors.NoSuchRevisionInTree(self, revision_id)
837
if revision_id in dirstate.get_ghosts():
838
raise errors.NoSuchRevisionInTree(self, revision_id)
839
return DirStateRevisionTree(dirstate, revision_id,
840
self.branch.repository)
842
@needs_tree_write_lock
843
def set_last_revision(self, new_revision):
844
"""Change the last revision in the working tree."""
845
new_revision = osutils.safe_revision_id(new_revision)
846
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 "
851
self.set_parent_ids([])
853
self.set_parent_ids([new_revision] + parents[1:],
854
allow_leftmost_as_ghost=True)
856
@needs_tree_write_lock
857
def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
858
"""Set the parent ids to revision_ids.
860
See also set_parent_trees. This api will try to retrieve the tree data
861
for each element of revision_ids from the trees repository. If you have
862
tree data already available, it is more efficient to use
863
set_parent_trees rather than set_parent_ids. set_parent_ids is however
864
an easier API to use.
866
:param revision_ids: The revision_ids to set as the parent ids of this
867
working tree. Any of these may be ghosts.
869
revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
871
for revision_id in revision_ids:
873
revtree = self.branch.repository.revision_tree(revision_id)
874
# TODO: jam 20070213 KnitVersionedFile raises
875
# RevisionNotPresent rather than NoSuchRevision if a
876
# given revision_id is not present. Should Repository be
877
# catching it and re-raising NoSuchRevision?
878
except (errors.NoSuchRevision, errors.RevisionNotPresent):
880
trees.append((revision_id, revtree))
881
self.set_parent_trees(trees,
882
allow_leftmost_as_ghost=allow_leftmost_as_ghost)
884
@needs_tree_write_lock
885
def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
886
"""Set the parents of the working tree.
888
:param parents_list: A list of (revision_id, tree) tuples.
889
If tree is None, then that element is treated as an unreachable
890
parent tree - i.e. a ghost.
892
dirstate = self.current_dirstate()
893
if len(parents_list) > 0:
894
if not allow_leftmost_as_ghost and parents_list[0][1] is None:
895
raise errors.GhostRevisionUnusableHere(parents_list[0][0])
898
# convert absent trees to the null tree, which we convert back to
900
for rev_id, tree in parents_list:
901
rev_id = osutils.safe_revision_id(rev_id)
903
real_trees.append((rev_id, tree))
905
real_trees.append((rev_id,
906
self.branch.repository.revision_tree(None)))
907
ghosts.append(rev_id)
908
dirstate.set_parent_trees(real_trees, ghosts=ghosts)
911
def _set_root_id(self, file_id):
912
"""See WorkingTree.set_root_id."""
913
state = self.current_dirstate()
914
state.set_path_id('', file_id)
915
self._dirty = state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED
918
"""Unlock in format 4 trees needs to write the entire dirstate."""
919
if self._control_files._lock_count == 1:
920
self._write_hashcache_if_dirty()
921
# eventually we should do signature checking during read locks for
923
if self._control_files._lock_mode == 'w':
926
if self._dirstate is not None:
927
self._dirstate.unlock()
928
self._dirstate = None
929
self._inventory = None
930
# reverse order of locking.
932
return self._control_files.unlock()
936
@needs_tree_write_lock
937
def unversion(self, file_ids):
938
"""Remove the file ids in file_ids from the current versioned set.
940
When a file_id is unversioned, all of its children are automatically
943
:param file_ids: The file ids to stop versioning.
944
:raises: NoSuchId if any fileid is not currently versioned.
948
state = self.current_dirstate()
949
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))
953
paths_to_unversion = set()
955
# check if the root is to be unversioned, if so, assert for now.
956
# walk the state marking unversioned things as absent.
957
# if there are any un-unversioned ids at the end, raise
958
for key, details in state._dirblocks[0][1]:
959
if (details[0][0] not in ('a', 'r') and # absent or relocated
960
key[2] in ids_to_unversion):
961
# I haven't written the code to unversion / yet - it should be
963
raise errors.BzrError('Unversioning the / is not currently supported')
965
while block_index < len(state._dirblocks):
966
# process one directory at a time.
967
block = state._dirblocks[block_index]
968
# first check: is the path one to remove - it or its children
970
for path in paths_to_unversion:
971
if (block[0].startswith(path) and
972
(len(block[0]) == len(path) or
973
block[0][len(path)] == '/')):
974
# this entire block should be deleted - its the block for a
975
# path to unversion; or the child of one
978
# TODO: trim paths_to_unversion as we pass by paths
980
# this block is to be deleted: process it.
981
# TODO: we can special case the no-parents case and
982
# just forget the whole block.
984
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]):
989
# go to the next block. (At the moment we dont delete empty
994
while entry_index < len(block[1]):
995
entry = block[1][entry_index]
996
if (entry[1][0][0] in ('a', 'r') or # absent, relocated
998
entry[0][2] not in ids_to_unversion):
999
# ^ not an id to unversion
1002
if entry[1][0][0] == 'd':
1003
paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
1004
if not state._make_absent(entry):
1006
# we have unversioned this id
1007
ids_to_unversion.remove(entry[0][2])
1009
if ids_to_unversion:
1010
raise errors.NoSuchId(self, iter(ids_to_unversion).next())
1012
# have to change the legacy inventory too.
1013
if self._inventory is not None:
1014
for file_id in file_ids:
1015
self._inventory.remove_recursive_id(file_id)
1017
@needs_tree_write_lock
1018
def _write_inventory(self, inv):
1019
"""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)
1026
class WorkingTreeFormat4(WorkingTreeFormat3):
1027
"""The first consolidated dirstate working tree 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.
1038
def get_format_string(self):
1039
"""See WorkingTreeFormat.get_format_string()."""
1040
return "Bazaar Working Tree format 4\n"
1042
def get_format_description(self):
1043
"""See WorkingTreeFormat.get_format_description()."""
1044
return "Working tree format 4"
1046
def initialize(self, a_bzrdir, revision_id=None):
1047
"""See WorkingTreeFormat.initialize().
1049
revision_id allows creating a working tree at a different
1050
revision than the branch is at.
1052
revision_id = osutils.safe_revision_id(revision_id)
1053
if not isinstance(a_bzrdir.transport, LocalTransport):
1054
raise errors.NotLocalUrl(a_bzrdir.transport.base)
1055
transport = a_bzrdir.get_workingtree_transport(self)
1056
control_files = self._open_control_files(a_bzrdir)
1057
control_files.create_lock()
1058
control_files.lock_write()
1059
control_files.put_utf8('format', self.get_format_string())
1060
branch = a_bzrdir.open_branch()
1061
if revision_id is None:
1062
revision_id = branch.last_revision()
1063
local_path = transport.local_abspath('dirstate')
1064
state = dirstate.DirState.initialize(local_path)
1066
wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1070
_control_files=control_files)
1072
wt.lock_tree_write()
1074
#wt.current_dirstate().set_path_id('', NEWROOT)
1075
wt.set_last_revision(revision_id)
1077
basis = wt.basis_tree()
1079
transform.build_tree(basis, wt)
1082
control_files.unlock()
1087
def _open(self, a_bzrdir, control_files):
1088
"""Open the tree itself.
1090
:param a_bzrdir: the dir for the tree.
1091
:param control_files: the control files for the tree.
1093
return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1094
branch=a_bzrdir.open_branch(),
1097
_control_files=control_files)
1100
class DirStateRevisionTree(Tree):
1101
"""A revision tree pulling the inventory from a dirstate."""
1103
def __init__(self, dirstate, revision_id, repository):
1104
self._dirstate = dirstate
1105
self._revision_id = osutils.safe_revision_id(revision_id)
1106
self._repository = repository
1107
self._inventory = None
1109
self._dirstate_locked = False
1111
def annotate_iter(self, file_id):
1112
"""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)
1117
def _comparison_data(self, entry, path):
1118
"""See Tree._comparison_data."""
1120
return None, False, None
1121
# trust the entry as RevisionTree does, but this may not be
1122
# sensible: the entry might not have come from us?
1123
return entry.kind, entry.executable, None
1125
def _file_size(self, entry, stat_value):
1126
return entry.text_size
1128
def filter_unversioned_files(self, paths):
1129
"""Filter out paths that are not versioned.
1131
:return: set of paths.
1133
pred = self.has_filename
1134
return set((p for p in paths if not pred(p)))
1136
def _get_parent_index(self):
1137
"""Return the index in the dirstate referenced by this tree."""
1138
return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1140
def _get_entry(self, file_id=None, path=None):
1141
"""Get the dirstate row for file_id or path.
1143
If either file_id or path is supplied, it is used as the key to lookup.
1144
If both are supplied, the fastest lookup is used, and an error is
1145
raised if they do not both point at the same row.
1147
:param file_id: An optional unicode file_id to be looked up.
1148
:param path: An optional unicode path to be looked up.
1149
:return: The dirstate row tuple for path/file_id, or (None, None)
1151
if file_id is None and path is None:
1152
raise errors.BzrError('must supply file_id or path')
1153
file_id = osutils.safe_file_id(file_id)
1154
if path is not None:
1155
path = path.encode('utf8')
1156
parent_index = self._get_parent_index()
1157
return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1159
def _generate_inventory(self):
1160
"""Create and set self.inventory from the dirstate object.
1162
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'
1166
# separate call for profiling - makes it clear where the costs are.
1167
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())
1171
parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1172
# This is identical now to the WorkingTree _generate_inventory except
1173
# for the tree index use.
1174
root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1175
current_id = root_key[2]
1176
assert current_entry[parent_index][0] == 'd'
1177
inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1178
inv.root.revision = current_entry[parent_index][4]
1179
# Turn some things into local variables
1180
minikind_to_kind = dirstate.DirState._minikind_to_kind
1181
factory = entry_factory
1182
utf8_decode = cache_utf8._utf8_decode
1183
inv_byid = inv._byid
1184
# we could do this straight out of the dirstate; it might be fast
1185
# and should be profiled - RBC 20070216
1186
parent_ies = {'' : inv.root}
1187
for block in self._dirstate._dirblocks[1:]: #skip root
1190
parent_ie = parent_ies[dirname]
1192
# all the paths in this block are not versioned in this tree
1194
for key, entry in block[1]:
1195
minikind, link_or_sha1, size, executable, revid = entry[parent_index]
1196
if minikind in ('a', 'r'): # absent, relocated
1200
name_unicode = utf8_decode(name)[0]
1202
kind = minikind_to_kind[minikind]
1203
inv_entry = factory[kind](file_id, name_unicode,
1205
inv_entry.revision = revid
1207
inv_entry.executable = executable
1208
inv_entry.text_size = size
1209
inv_entry.text_sha1 = link_or_sha1
1210
elif kind == 'directory':
1211
parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1212
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]
1217
raise Exception, kind
1218
# 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
1221
inv_byid[file_id] = inv_entry
1222
parent_ie.children[name_unicode] = inv_entry
1223
self._inventory = inv
1225
def get_file_mtime(self, file_id, path=None):
1226
"""Return the modification time for this record.
1228
We return the timestamp of the last-changed revision.
1230
# Make sure the file exists
1231
entry = self._get_entry(file_id, path=path)
1232
if entry == (None, None): # do we raise?
1234
parent_index = self._get_parent_index()
1235
last_changed_revision = entry[1][parent_index][4]
1236
return self._repository.get_revision(last_changed_revision).timestamp
1238
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":
1246
def get_file(self, file_id):
1247
return StringIO(self.get_file_text(file_id))
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)
1254
def get_file_size(self, file_id):
1255
return self.inventory[file_id].text_size
1257
def get_file_text(self, file_id):
1258
return ''.join(self.get_file_lines(file_id))
1260
def get_symlink_target(self, file_id):
1261
entry = self._get_entry(file_id=file_id)
1262
parent_index = self._get_parent_index()
1263
if entry[1][parent_index][0] != 'l':
1266
# At present, none of the tree implementations supports non-ascii
1267
# symlink targets. So we will just assume that the dirstate path is
1269
return entry[1][parent_index][1]
1271
def get_revision_id(self):
1272
"""Return the revision id for this tree."""
1273
return self._revision_id
1275
def _get_inventory(self):
1276
if self._inventory is not None:
1277
return self._inventory
1278
self._generate_inventory()
1279
return self._inventory
1281
inventory = property(_get_inventory,
1282
doc="Inventory of this Tree")
1284
def get_parent_ids(self):
1285
"""The parents of a tree in the dirstate are not cached."""
1286
return self._repository.get_revision(self._revision_id).parent_ids
1288
def has_filename(self, filename):
1289
return bool(self.path2id(filename))
1291
def kind(self, file_id):
1292
return self.inventory[file_id].kind
1294
def is_executable(self, file_id, path=None):
1295
ie = self.inventory[file_id]
1296
if ie.kind != "file":
1298
return ie.executable
1300
def list_files(self, include_root=False):
1301
# We use a standard implementation, because DirStateRevisionTree is
1302
# 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:
1307
for path, entry in entries:
1308
yield path, 'V', entry.kind, entry.file_id, entry
1310
def lock_read(self):
1311
"""Lock the tree for a set of operations."""
1312
if not self._locked:
1313
self._repository.lock_read()
1314
if self._dirstate._lock_token is None:
1315
self._dirstate.lock_read()
1316
self._dirstate_locked = True
1320
def path2id(self, path):
1321
"""Return the id for path in this tree."""
1322
# lookup by path: faster than splitting and walking the ivnentory.
1323
entry = self._get_entry(path=path)
1324
if entry == (None, None):
1329
"""Unlock, freeing any cache memory used during the lock."""
1330
# outside of a lock, the inventory is suspect: release it.
1332
if not self._locked:
1333
self._inventory = None
1335
if self._dirstate_locked:
1336
self._dirstate.unlock()
1337
self._dirstate_locked = False
1338
self._repository.unlock()
1340
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.
1348
_directory = 'directory'
1349
inv = self._get_inventory()
1350
top_id = inv.path2id(prefix)
1354
pending = [(prefix, top_id)]
1357
relpath, file_id = pending.pop()
1358
# 0 - relpath, 1- file-id
1360
relroot = relpath + '/'
1363
# FIXME: stash the node in pending
1364
entry = inv[file_id]
1365
for name, child in entry.sorted_children():
1366
toppath = relroot + name
1367
dirblock.append((toppath, name, child.kind, None,
1368
child.file_id, child.kind
1370
yield (relpath, entry.file_id), dirblock
1371
# push the user specified dirs from dirblock
1372
for dir in reversed(dirblock):
1373
if dir[2] == _directory:
1374
pending.append((dir[0], dir[4]))
1377
class InterDirStateTree(InterTree):
1378
"""Fast path optimiser for changes_from with dirstate trees."""
1380
def __init__(self, source, target):
1381
super(InterDirStateTree, self).__init__(source, target)
1382
if not InterDirStateTree.is_compatible(source, target):
1383
raise Exception, "invalid source %r and target %r" % (source, target)
1386
def make_source_parent_tree(source, target):
1387
"""Change the source tree into a parent of the target."""
1388
revid = source.commit('record tree')
1389
target.branch.repository.fetch(source.branch.repository, revid)
1390
target.set_parent_ids([revid])
1391
return target.basis_tree(), target
1393
_matching_from_tree_format = WorkingTreeFormat4()
1394
_matching_to_tree_format = WorkingTreeFormat4()
1395
_test_mutable_trees_to_test_trees = make_source_parent_tree
1397
def _iter_changes(self, include_unchanged=False,
1398
specific_files=None, pb=None, extra_trees=[],
1399
require_versioned=True):
1400
"""Return the changes from source to target.
1402
:return: An iterator that yields tuples. See InterTree._iter_changes
1404
:param specific_files: An optional list of file paths to restrict the
1405
comparison to. When mapping filenames to ids, all matches in all
1406
trees (including optional extra_trees) are used, and all children of
1407
matched directories are included.
1408
:param include_unchanged: An optional boolean requesting the inclusion of
1409
unchanged entries in the result.
1410
:param extra_trees: An optional list of additional trees to use when
1411
mapping the contents of specific_files (paths) to file_ids.
1412
:param require_versioned: If True, all files in specific_files must be
1413
versioned in one of source, target, extra_trees or
1414
PathsNotVersionedError is raised.
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.
1422
for f in super(InterDirStateTree, self)._iter_changes(
1423
include_unchanged, specific_files, pb, extra_trees,
1427
parent_ids = self.target.get_parent_ids()
1429
if self.source._revision_id == NULL_REVISION:
1431
indices = (target_index,)
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)
1436
source_index = 1 + parent_ids.index(self.source._revision_id)
1437
indices = (source_index,target_index)
1438
# -- make all specific_files utf8 --
1440
specific_files_utf8 = set()
1441
for path in specific_files:
1442
specific_files_utf8.add(path.encode('utf8'))
1443
specific_files = specific_files_utf8
1445
specific_files = set([''])
1446
# -- specific_files is now a utf8 path set --
1447
# -- get the state object and prepare it.
1448
state = self.target.current_dirstate()
1449
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.
1453
dirname, basename = os.path.split(path)
1454
key = (dirname, basename, '')
1455
block_index, present = state._find_block_index_from_key(key)
1457
# the block which should contain path is absent.
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])
1468
if require_versioned:
1469
# -- check all supplied paths are versioned in a search tree. --
1470
all_versioned = True
1471
for path in specific_files:
1472
path_entries = _entries_for_path(path)
1473
if not path_entries:
1474
# this specified path is not present at all: error
1475
all_versioned = False
1477
found_versioned = False
1478
# for each id at this path
1479
for entry in path_entries:
1481
for index in indices:
1482
if entry[1][index][0] != 'a': # absent
1483
found_versioned = True
1484
# all good: found a versioned cell
1486
if not found_versioned:
1487
# none of the indexes was not 'absent' at all ids for this
1489
all_versioned = False
1491
if not all_versioned:
1492
raise errors.PathsNotVersionedError(paths)
1493
# -- 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)
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
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.
1512
# r | a | | add source to search
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.
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]
1546
def _process_entry(entry, path_info):
1547
"""Compare an entry and real disk to generate delta information.
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 ?)
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
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.
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,
1572
search_specific_files.add(source_details[1])
1573
# generate the old path; this is needed for stating later
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,
1580
# update the source details variable to be the real
1582
source_details = old_entry[1][source_index]
1583
source_minikind = source_details[0]
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
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
1601
# directories have no fingerprint
1602
content_change = False
1604
elif target_kind == 'file':
1605
if source_minikind != 'f':
1606
content_change = True
1608
# has it changed? fast path: size, slow path: sha1.
1609
if source_details[2] != path_info[3].st_size:
1610
content_change = True
1612
# maybe the same. Get the hash
1613
new_hash = self.target._hashcache.get_sha1(
1615
content_change = (new_hash != source_details[1])
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
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])
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]
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
1641
last_source_parent[0] = old_dirname
1642
last_source_parent[1] = source_parent_id
1643
last_source_parent[2] = source_parent_entry
1645
new_dirname = entry[0][0]
1646
if new_dirname == last_target_parent[0]:
1647
target_parent_id = last_target_parent[1]
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
1658
last_target_parent[0] = new_dirname
1659
last_target_parent[1] = target_parent_id
1660
last_target_parent[2] = target_parent_entry
1662
source_exec = source_details[3]
1663
path_unicode = utf8_decode(path)[0]
1664
return ((entry[0][2], path_unicode, content_change,
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]:
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,
1687
(None, entry[0][1]),
1688
(None, path_info[2]),
1689
(None, new_executable)),)
1691
# but its not on disk: we deliberately treat this as just
1692
# never-present. (Why ?! - RBC 20070224)
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]:
1704
old_path_unicode = utf8_decode(old_path)[0]
1705
return ((entry[0][2], old_path_unicode, True,
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.
1724
print "*******", source_minikind, target_minikind
1725
import pdb;pdb.set_trace()
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)
1736
root_stat = os.lstat(root_abspath)
1738
if e.errno == errno.ENOENT:
1739
# the path does not exist: let _process_entry know that.
1740
root_dir_info = None
1742
# some other random error: hand it up.
1745
root_dir_info = ('', current_root,
1746
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1748
if not root_entries and not root_dir_info:
1749
# this specified path is not present at all, skip it.
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
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.
1773
current_dir_info = dir_iterator.next()
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
1779
current_dir_info = None
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]
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.
1813
current_dir_info = dir_iterator.next()
1814
except StopIteration:
1815
current_dir_info = None
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
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
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]
1846
current_block = None
1849
if current_block and entry_index < len(current_block[1]):
1850
current_entry = current_block[1][entry_index]
1852
current_entry = None
1853
advance_entry = True
1855
if current_dir_info and path_index < len(current_dir_info[1]):
1856
current_path_info = current_dir_info[1][path_index]
1858
current_path_info = None
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,
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
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
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
1913
advance_path = False
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
1928
if advance_entry and current_entry is not None:
1930
if entry_index < len(current_block[1]):
1931
current_entry = current_block[1][entry_index]
1933
current_entry = None
1935
advance_entry = True # reset the advance flaga
1936
if advance_path and current_path_info is not None:
1938
if path_index < len(current_dir_info[1]):
1939
current_path_info = current_dir_info[1][path_index]
1941
current_path_info = None
1943
advance_path = True # reset the advance flagg.
1944
if current_block is not None:
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]
1950
current_block = None
1951
if current_dir_info is not None:
1953
current_dir_info = dir_iterator.next()
1954
except StopIteration:
1955
current_dir_info = None
1959
def is_compatible(source, target):
1960
# the target must be a dirstate working tree
1961
if not isinstance(target, WorkingTree4):
1963
# the source must be a revtreee or dirstate rev tree.
1964
if not isinstance(source,
1965
(revisiontree.RevisionTree, DirStateRevisionTree)):
1967
# the source revid must be in the target dirstate
1968
if not (source._revision_id == NULL_REVISION or
1969
source._revision_id in target.get_parent_ids()):
1970
# TODO: what about ghosts? it may well need to
1971
# check for them explicitly.
1975
InterTree.register_optimiser(InterDirStateTree)