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[block[0]]
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
323
assert name_unicode not in parent_ie.children
324
inv_byid[file_id] = inv_entry
325
parent_ie.children[name_unicode] = inv_entry
326
self._inventory = inv
328
def _get_entry(self, file_id=None, path=None):
329
"""Get the dirstate row for file_id or path.
331
If either file_id or path is supplied, it is used as the key to lookup.
332
If both are supplied, the fastest lookup is used, and an error is
333
raised if they do not both point at the same row.
335
:param file_id: An optional unicode file_id to be looked up.
336
:param path: An optional unicode path to be looked up.
337
:return: The dirstate row tuple for path/file_id, or (None, None)
339
if file_id is None and path is None:
340
raise errors.BzrError('must supply file_id or path')
341
state = self.current_dirstate()
343
path = path.encode('utf8')
344
return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
346
def get_file_sha1(self, file_id, path=None, stat_value=None):
347
# check file id is valid unconditionally.
348
key, details = self._get_entry(file_id=file_id, path=path)
349
assert key is not None, 'what error should this raise'
351
# if row stat is valid, use cached sha1, else, get a new sha1.
353
path = os.path.join(*key[0:2]).decode('utf8')
354
return self._hashcache.get_sha1(path, stat_value)
356
def _get_inventory(self):
357
"""Get the inventory for the tree. This is only valid within a lock."""
358
if self._inventory is not None:
359
return self._inventory
360
self._generate_inventory()
361
return self._inventory
363
inventory = property(_get_inventory,
364
doc="Inventory of this Tree")
367
def get_parent_ids(self):
368
"""See Tree.get_parent_ids.
370
This implementation requests the ids list from the dirstate file.
372
return self.current_dirstate().get_parent_ids()
375
def get_root_id(self):
376
"""Return the id of this trees root"""
377
return self._get_entry(path='')[0][2]
379
def has_id(self, file_id):
380
state = self.current_dirstate()
381
file_id = osutils.safe_file_id(file_id)
382
row, parents = self._get_entry(file_id=file_id)
385
return osutils.lexists(pathjoin(
386
self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
389
def id2path(self, file_id):
390
file_id = osutils.safe_file_id(file_id)
391
state = self.current_dirstate()
392
# XXX: possible_dir_name_ids is unused - is this just called for
393
# side-effects? mbp 20070226
394
possible_dir_name_ids = state._get_id_index().get(file_id, None)
395
entry = self._get_entry(file_id=file_id)
396
if entry == (None, None):
398
path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
399
return path_utf8.decode('utf8')
403
"""Iterate through file_ids for this tree.
405
file_ids are in a WorkingTree if they are in the working inventory
406
and the working file exists.
409
for key, tree_details in self.current_dirstate()._iter_entries():
410
if tree_details[0][0] in ('a', 'r'): # absent, relocated
411
# not relevant to the working tree
413
path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
414
if osutils.lexists(path):
415
result.append(key[2])
419
def _last_revision(self):
420
"""See Mutable.last_revision."""
421
parent_ids = self.current_dirstate().get_parent_ids()
428
super(WorkingTree4, self).lock_read()
429
if self._dirstate is None:
430
self.current_dirstate()
431
self._dirstate.lock_read()
433
def lock_tree_write(self):
434
super(WorkingTree4, self).lock_tree_write()
435
if self._dirstate is None:
436
self.current_dirstate()
437
self._dirstate.lock_write()
439
def lock_write(self):
440
super(WorkingTree4, self).lock_write()
441
if self._dirstate is None:
442
self.current_dirstate()
443
self._dirstate.lock_write()
445
@needs_tree_write_lock
446
def move(self, from_paths, to_dir, after=False):
447
"""See WorkingTree.move()."""
451
state = self.current_dirstate()
453
assert not isinstance(from_paths, basestring)
454
to_dir_utf8 = to_dir.encode('utf8')
455
to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
456
id_index = state._get_id_index()
457
# check destination directory
458
# get the details for it
459
to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
460
state._get_block_entry_index(to_entry_dirname, to_basename, 0)
461
if not entry_present:
462
raise errors.BzrMoveFailedError('', to_dir,
463
errors.NotInWorkingDirectory(to_dir))
464
to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
465
# get a handle on the block itself.
466
to_block_index = state._ensure_block(
467
to_entry_block_index, to_entry_entry_index, to_dir_utf8)
468
to_block = state._dirblocks[to_block_index]
469
to_abs = self.abspath(to_dir)
470
if not isdir(to_abs):
471
raise errors.BzrMoveFailedError('',to_dir,
472
errors.NotADirectory(to_abs))
474
if to_entry[1][0][0] != 'd':
475
raise errors.BzrMoveFailedError('',to_dir,
476
errors.NotADirectory(to_abs))
478
if self._inventory is not None:
479
update_inventory = True
481
to_dir_ie = inv[to_dir_id]
482
to_dir_id = to_entry[0][2]
484
update_inventory = False
487
def move_one(old_entry, from_path_utf8, minikind, executable,
488
fingerprint, packed_stat, size,
489
to_block, to_key, to_path_utf8):
490
state._make_absent(old_entry)
491
from_key = old_entry[0]
493
lambda:state.update_minimal(from_key,
495
executable=executable,
496
fingerprint=fingerprint,
497
packed_stat=packed_stat,
499
path_utf8=from_path_utf8))
500
state.update_minimal(to_key,
502
executable=executable,
503
fingerprint=fingerprint,
504
packed_stat=packed_stat,
506
path_utf8=to_path_utf8)
507
added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
508
new_entry = to_block[1][added_entry_index]
509
rollbacks.append(lambda:state._make_absent(new_entry))
511
# create rename entries and tuples
512
for from_rel in from_paths:
513
# from_rel is 'pathinroot/foo/bar'
514
from_rel_utf8 = from_rel.encode('utf8')
515
from_dirname, from_tail = osutils.split(from_rel)
516
from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
517
from_entry = self._get_entry(path=from_rel)
518
if from_entry == (None, None):
519
raise errors.BzrMoveFailedError(from_rel,to_dir,
520
errors.NotVersionedError(path=str(from_rel)))
522
from_id = from_entry[0][2]
523
to_rel = pathjoin(to_dir, from_tail)
524
to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
525
item_to_entry = self._get_entry(path=to_rel)
526
if item_to_entry != (None, None):
527
raise errors.BzrMoveFailedError(from_rel, to_rel,
528
"Target is already versioned.")
530
if from_rel == to_rel:
531
raise errors.BzrMoveFailedError(from_rel, to_rel,
532
"Source and target are identical.")
534
from_missing = not self.has_filename(from_rel)
535
to_missing = not self.has_filename(to_rel)
542
raise errors.BzrMoveFailedError(from_rel, to_rel,
543
errors.NoSuchFile(path=to_rel,
544
extra="New file has not been created yet"))
546
# neither path exists
547
raise errors.BzrRenameFailedError(from_rel, to_rel,
548
errors.PathsDoNotExist(paths=(from_rel, to_rel)))
550
if from_missing: # implicitly just update our path mapping
553
raise errors.RenameFailedFilesExist(from_rel, to_rel,
554
extra="(Use --after to update the Bazaar id)")
557
def rollback_rename():
558
"""A single rename has failed, roll it back."""
560
for rollback in reversed(rollbacks):
564
import pdb;pdb.set_trace()
565
exc_info = sys.exc_info()
567
raise exc_info[0], exc_info[1], exc_info[2]
569
# perform the disk move first - its the most likely failure point.
571
from_rel_abs = self.abspath(from_rel)
572
to_rel_abs = self.abspath(to_rel)
574
osutils.rename(from_rel_abs, to_rel_abs)
576
raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
577
rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
579
# perform the rename in the inventory next if needed: its easy
583
from_entry = inv[from_id]
584
current_parent = from_entry.parent_id
585
inv.rename(from_id, to_dir_id, from_tail)
587
lambda: inv.rename(from_id, current_parent, from_tail))
588
# finally do the rename in the dirstate, which is a little
589
# tricky to rollback, but least likely to need it.
590
old_block_index, old_entry_index, dir_present, file_present = \
591
state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
592
old_block = state._dirblocks[old_block_index][1]
593
old_entry = old_block[old_entry_index]
594
from_key, old_entry_details = old_entry
595
cur_details = old_entry_details[0]
597
to_key = ((to_block[0],) + from_key[1:3])
598
minikind = cur_details[0]
599
move_one(old_entry, from_path_utf8=from_rel_utf8,
601
executable=cur_details[3],
602
fingerprint=cur_details[1],
603
packed_stat=cur_details[4],
607
to_path_utf8=to_rel_utf8)
610
def update_dirblock(from_dir, to_key, to_dir_utf8):
611
"""all entries in this block need updating.
613
TODO: This is pretty ugly, and doesn't support
614
reverting, but it works.
616
assert from_dir != '', "renaming root not supported"
617
from_key = (from_dir, '')
618
from_block_idx, present = \
619
state._find_block_index_from_key(from_key)
621
# This is the old record, if it isn't present, then
622
# there is theoretically nothing to update.
623
# (Unless it isn't present because of lazy loading,
624
# but we don't do that yet)
626
from_block = state._dirblocks[from_block_idx]
627
to_block_index, to_entry_index, _, _ = \
628
state._get_block_entry_index(to_key[0], to_key[1], 0)
629
to_block_index = state._ensure_block(
630
to_block_index, to_entry_index, to_dir_utf8)
631
to_block = state._dirblocks[to_block_index]
632
for entry in from_block[1]:
633
assert entry[0][0] == from_dir
634
cur_details = entry[1][0]
635
to_key = (to_dir_utf8, entry[0][1], entry[0][2])
636
from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
637
to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
638
minikind = cur_details[0]
639
move_one(entry, from_path_utf8=from_path_utf8,
641
executable=cur_details[3],
642
fingerprint=cur_details[1],
643
packed_stat=cur_details[4],
647
to_path_utf8=to_rel_utf8)
649
# We need to move all the children of this
651
update_dirblock(from_path_utf8, to_key,
653
update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
657
state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
660
return #rename_tuples
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
entry = self._get_entry(path=path)
670
if entry == (None, None):
674
def paths2ids(self, paths, trees=[], require_versioned=True):
675
"""See Tree.paths2ids().
677
This specialisation fast-paths the case where all the trees are in the
682
parents = self.get_parent_ids()
684
if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
686
return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
687
search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
688
# -- make all paths utf8 --
691
paths_utf8.add(path.encode('utf8'))
693
# -- paths is now a utf8 path set --
694
# -- get the state object and prepare it.
695
state = self.current_dirstate()
696
if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
697
and '' not in paths):
698
paths2ids = self._paths2ids_using_bisect
700
paths2ids = self._paths2ids_in_memory
701
return paths2ids(paths, search_indexes,
702
require_versioned=require_versioned)
704
def _paths2ids_in_memory(self, paths, search_indexes,
705
require_versioned=True):
706
state = self.current_dirstate()
707
state._read_dirblocks_if_needed()
708
def _entries_for_path(path):
709
"""Return a list with all the entries that match path for all ids.
711
dirname, basename = os.path.split(path)
712
key = (dirname, basename, '')
713
block_index, present = state._find_block_index_from_key(key)
715
# the block which should contain path is absent.
718
block = state._dirblocks[block_index][1]
719
entry_index, _ = state._find_entry_index(key, block)
720
# we may need to look at multiple entries at this path: walk while the paths match.
721
while (entry_index < len(block) and
722
block[entry_index][0][0:2] == key[0:2]):
723
result.append(block[entry_index])
726
if require_versioned:
727
# -- check all supplied paths are versioned in a search tree. --
730
path_entries = _entries_for_path(path)
732
# this specified path is not present at all: error
733
all_versioned = False
735
found_versioned = False
736
# for each id at this path
737
for entry in path_entries:
739
for index in search_indexes:
740
if entry[1][index][0] != 'a': # absent
741
found_versioned = True
742
# all good: found a versioned cell
744
if not found_versioned:
745
# none of the indexes was not 'absent' at all ids for this
747
all_versioned = False
749
if not all_versioned:
750
raise errors.PathsNotVersionedError(paths)
751
# -- remove redundancy in supplied paths to prevent over-scanning --
754
other_paths = paths.difference(set([path]))
755
if not osutils.is_inside_any(other_paths, path):
756
# this is a top level path, we must check it.
757
search_paths.add(path)
759
# for all search_indexs in each path at or under each element of
760
# search_paths, if the detail is relocated: add the id, and add the
761
# relocated path as one to search if its not searched already. If the
762
# detail is not relocated, add the id.
763
searched_paths = set()
765
def _process_entry(entry):
766
"""Look at search_indexes within entry.
768
If a specific tree's details are relocated, add the relocation
769
target to search_paths if not searched already. If it is absent, do
770
nothing. Otherwise add the id to found_ids.
772
for index in search_indexes:
773
if entry[1][index][0] == 'r': # relocated
774
if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
775
search_paths.add(entry[1][index][1])
776
elif entry[1][index][0] != 'a': # absent
777
found_ids.add(entry[0][2])
779
current_root = search_paths.pop()
780
searched_paths.add(current_root)
781
# process the entries for this containing directory: the rest will be
782
# found by their parents recursively.
783
root_entries = _entries_for_path(current_root)
785
# this specified path is not present at all, skip it.
787
for entry in root_entries:
788
_process_entry(entry)
789
initial_key = (current_root, '', '')
790
block_index, _ = state._find_block_index_from_key(initial_key)
791
while (block_index < len(state._dirblocks) and
792
osutils.is_inside(current_root, state._dirblocks[block_index][0])):
793
for entry in state._dirblocks[block_index][1]:
794
_process_entry(entry)
798
def _paths2ids_using_bisect(self, paths, search_indexes,
799
require_versioned=True):
800
state = self.current_dirstate()
803
split_paths = sorted(osutils.split(p) for p in paths)
804
found = state._bisect_recursive(split_paths)
806
if require_versioned:
807
found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
808
for dir_name in split_paths:
809
if dir_name not in found_dir_names:
810
raise errors.PathsNotVersionedError(paths)
812
for dir_name_id, trees_info in found.iteritems():
813
for index in search_indexes:
814
if trees_info[index][0] not in ('r', 'a'):
815
found_ids.add(dir_name_id[2])
818
def read_working_inventory(self):
819
"""Read the working inventory.
821
This is a meaningless operation for dirstate, but we obey it anyhow.
823
return self.inventory
826
def revision_tree(self, revision_id):
827
"""See Tree.revision_tree.
829
WorkingTree4 supplies revision_trees for any basis tree.
831
revision_id = osutils.safe_revision_id(revision_id)
832
dirstate = self.current_dirstate()
833
parent_ids = dirstate.get_parent_ids()
834
if revision_id not in parent_ids:
835
raise errors.NoSuchRevisionInTree(self, revision_id)
836
if revision_id in dirstate.get_ghosts():
837
raise errors.NoSuchRevisionInTree(self, revision_id)
838
return DirStateRevisionTree(dirstate, revision_id,
839
self.branch.repository)
841
@needs_tree_write_lock
842
def set_last_revision(self, new_revision):
843
"""Change the last revision in the working tree."""
844
new_revision = osutils.safe_revision_id(new_revision)
845
parents = self.get_parent_ids()
846
if new_revision in (NULL_REVISION, None):
847
assert len(parents) < 2, (
848
"setting the last parent to none with a pending merge is "
850
self.set_parent_ids([])
852
self.set_parent_ids([new_revision] + parents[1:],
853
allow_leftmost_as_ghost=True)
855
@needs_tree_write_lock
856
def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
857
"""Set the parent ids to revision_ids.
859
See also set_parent_trees. This api will try to retrieve the tree data
860
for each element of revision_ids from the trees repository. If you have
861
tree data already available, it is more efficient to use
862
set_parent_trees rather than set_parent_ids. set_parent_ids is however
863
an easier API to use.
865
:param revision_ids: The revision_ids to set as the parent ids of this
866
working tree. Any of these may be ghosts.
868
revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
870
for revision_id in revision_ids:
872
revtree = self.branch.repository.revision_tree(revision_id)
873
# TODO: jam 20070213 KnitVersionedFile raises
874
# RevisionNotPresent rather than NoSuchRevision if a
875
# given revision_id is not present. Should Repository be
876
# catching it and re-raising NoSuchRevision?
877
except (errors.NoSuchRevision, errors.RevisionNotPresent):
879
trees.append((revision_id, revtree))
880
self.set_parent_trees(trees,
881
allow_leftmost_as_ghost=allow_leftmost_as_ghost)
883
@needs_tree_write_lock
884
def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
885
"""Set the parents of the working tree.
887
:param parents_list: A list of (revision_id, tree) tuples.
888
If tree is None, then that element is treated as an unreachable
889
parent tree - i.e. a ghost.
891
dirstate = self.current_dirstate()
892
if len(parents_list) > 0:
893
if not allow_leftmost_as_ghost and parents_list[0][1] is None:
894
raise errors.GhostRevisionUnusableHere(parents_list[0][0])
897
# convert absent trees to the null tree, which we convert back to
899
for rev_id, tree in parents_list:
900
rev_id = osutils.safe_revision_id(rev_id)
902
real_trees.append((rev_id, tree))
904
real_trees.append((rev_id,
905
self.branch.repository.revision_tree(None)))
906
ghosts.append(rev_id)
907
dirstate.set_parent_trees(real_trees, ghosts=ghosts)
910
def _set_root_id(self, file_id):
911
"""See WorkingTree.set_root_id."""
912
state = self.current_dirstate()
913
state.set_path_id('', file_id)
914
self._dirty = state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED
917
"""Unlock in format 4 trees needs to write the entire dirstate."""
918
if self._control_files._lock_count == 1:
919
self._write_hashcache_if_dirty()
920
# eventually we should do signature checking during read locks for
922
if self._control_files._lock_mode == 'w':
925
if self._dirstate is not None:
926
self._dirstate.unlock()
927
self._dirstate = None
928
self._inventory = None
929
# reverse order of locking.
931
return self._control_files.unlock()
935
@needs_tree_write_lock
936
def unversion(self, file_ids):
937
"""Remove the file ids in file_ids from the current versioned set.
939
When a file_id is unversioned, all of its children are automatically
942
:param file_ids: The file ids to stop versioning.
943
:raises: NoSuchId if any fileid is not currently versioned.
947
state = self.current_dirstate()
948
state._read_dirblocks_if_needed()
949
ids_to_unversion = set()
950
for file_id in file_ids:
951
ids_to_unversion.add(osutils.safe_file_id(file_id))
952
paths_to_unversion = set()
954
# check if the root is to be unversioned, if so, assert for now.
955
# walk the state marking unversioned things as absent.
956
# if there are any un-unversioned ids at the end, raise
957
for key, details in state._dirblocks[0][1]:
958
if (details[0][0] not in ('a', 'r') and # absent or relocated
959
key[2] in ids_to_unversion):
960
# I haven't written the code to unversion / yet - it should be
962
raise errors.BzrError('Unversioning the / is not currently supported')
963
details_length = len(state._dirblocks[0][1][0][1])
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
if not state._make_absent(block[1][entry_index]):
987
# go to the next block. (At the moment we dont delete empty
992
while entry_index < len(block[1]):
993
entry = block[1][entry_index]
994
if (entry[1][0][0] in ('a', 'r') or # absent, relocated
996
entry[0][2] not in ids_to_unversion):
997
# ^ not an id to unversion
1000
if entry[1][0][0] == 'd':
1001
paths_to_unversion.add(os.path.join(*entry[0][0:2]))
1002
if not state._make_absent(entry):
1004
# we have unversioned this id
1005
ids_to_unversion.remove(entry[0][2])
1007
if ids_to_unversion:
1008
raise errors.NoSuchId(self, iter(ids_to_unversion).next())
1010
# have to change the legacy inventory too.
1011
if self._inventory is not None:
1012
for file_id in file_ids:
1013
self._inventory.remove_recursive_id(file_id)
1015
@needs_tree_write_lock
1016
def _write_inventory(self, inv):
1017
"""Write inventory as the current inventory."""
1018
assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
1019
self.current_dirstate().set_state_from_inventory(inv)
1024
class WorkingTreeFormat4(WorkingTreeFormat3):
1025
"""The first consolidated dirstate working tree format.
1028
- exists within a metadir controlling .bzr
1029
- includes an explicit version marker for the workingtree control
1030
files, separate from the BzrDir format
1031
- modifies the hash cache format
1032
- is new in bzr TODO FIXME SETBEFOREMERGE
1033
- uses a LockDir to guard access to it.
1036
def get_format_string(self):
1037
"""See WorkingTreeFormat.get_format_string()."""
1038
return "Bazaar Working Tree format 4\n"
1040
def get_format_description(self):
1041
"""See WorkingTreeFormat.get_format_description()."""
1042
return "Working tree format 4"
1044
def initialize(self, a_bzrdir, revision_id=None):
1045
"""See WorkingTreeFormat.initialize().
1047
revision_id allows creating a working tree at a different
1048
revision than the branch is at.
1050
revision_id = osutils.safe_revision_id(revision_id)
1051
if not isinstance(a_bzrdir.transport, LocalTransport):
1052
raise errors.NotLocalUrl(a_bzrdir.transport.base)
1053
transport = a_bzrdir.get_workingtree_transport(self)
1054
control_files = self._open_control_files(a_bzrdir)
1055
control_files.create_lock()
1056
control_files.lock_write()
1057
control_files.put_utf8('format', self.get_format_string())
1058
branch = a_bzrdir.open_branch()
1059
if revision_id is None:
1060
revision_id = branch.last_revision()
1061
local_path = transport.local_abspath('dirstate')
1062
state = dirstate.DirState.initialize(local_path)
1064
wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1068
_control_files=control_files)
1072
#wt.current_dirstate().set_path_id('', NEWROOT)
1073
wt.set_last_revision(revision_id)
1075
basis = wt.basis_tree()
1077
transform.build_tree(basis, wt)
1080
control_files.unlock()
1084
def _open(self, a_bzrdir, control_files):
1085
"""Open the tree itself.
1087
:param a_bzrdir: the dir for the tree.
1088
:param control_files: the control files for the tree.
1090
return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1091
branch=a_bzrdir.open_branch(),
1094
_control_files=control_files)
1097
class DirStateRevisionTree(Tree):
1098
"""A revision tree pulling the inventory from a dirstate."""
1100
def __init__(self, dirstate, revision_id, repository):
1101
self._dirstate = dirstate
1102
self._revision_id = osutils.safe_revision_id(revision_id)
1103
self._repository = repository
1104
self._inventory = None
1106
self._dirstate_locked = False
1108
def annotate_iter(self, file_id):
1109
"""See Tree.annotate_iter"""
1110
w = self._repository.weave_store.get_weave(file_id,
1111
self._repository.get_transaction())
1112
return w.annotate_iter(self.inventory[file_id].revision)
1114
def _comparison_data(self, entry, path):
1115
"""See Tree._comparison_data."""
1117
return None, False, None
1118
# trust the entry as RevisionTree does, but this may not be
1119
# sensible: the entry might not have come from us?
1120
return entry.kind, entry.executable, None
1122
def _file_size(self, entry, stat_value):
1123
return entry.text_size
1125
def filter_unversioned_files(self, paths):
1126
"""Filter out paths that are not versioned.
1128
:return: set of paths.
1130
pred = self.has_filename
1131
return set((p for p in paths if not pred(p)))
1133
def _get_parent_index(self):
1134
"""Return the index in the dirstate referenced by this tree."""
1135
return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1137
def _get_entry(self, file_id=None, path=None):
1138
"""Get the dirstate row for file_id or path.
1140
If either file_id or path is supplied, it is used as the key to lookup.
1141
If both are supplied, the fastest lookup is used, and an error is
1142
raised if they do not both point at the same row.
1144
:param file_id: An optional unicode file_id to be looked up.
1145
:param path: An optional unicode path to be looked up.
1146
:return: The dirstate row tuple for path/file_id, or (None, None)
1148
if file_id is None and path is None:
1149
raise errors.BzrError('must supply file_id or path')
1150
file_id = osutils.safe_file_id(file_id)
1151
if path is not None:
1152
path = path.encode('utf8')
1153
parent_index = self._get_parent_index()
1154
return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1156
def _generate_inventory(self):
1157
"""Create and set self.inventory from the dirstate object.
1159
This is relatively expensive: we have to walk the entire dirstate.
1160
Ideally we would not, and instead would """
1161
assert self._locked, 'cannot generate inventory of an unlocked '\
1162
'dirstate revision tree'
1163
# separate call for profiling - makes it clear where the costs are.
1164
self._dirstate._read_dirblocks_if_needed()
1165
assert self._revision_id in self._dirstate.get_parent_ids(), \
1166
'parent %s has disappeared from %s' % (
1167
self._revision_id, self._dirstate.get_parent_ids())
1168
parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1169
# This is identical now to the WorkingTree _generate_inventory except
1170
# for the tree index use.
1171
root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1172
current_id = root_key[2]
1173
assert current_entry[parent_index][0] == 'd'
1174
inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1175
inv.root.revision = current_entry[parent_index][4]
1176
# Turn some things into local variables
1177
minikind_to_kind = dirstate.DirState._minikind_to_kind
1178
factory = entry_factory
1179
utf8_decode = cache_utf8._utf8_decode
1180
inv_byid = inv._byid
1181
# we could do this straight out of the dirstate; it might be fast
1182
# and should be profiled - RBC 20070216
1183
parent_ies = {'' : inv.root}
1184
for block in self._dirstate._dirblocks[1:]: #skip root
1187
parent_ie = parent_ies[dirname]
1189
# all the paths in this block are not versioned in this tree
1191
for key, entry in block[1]:
1192
minikind, link_or_sha1, size, executable, revid = entry[parent_index]
1193
if minikind in ('a', 'r'): # absent, relocated
1197
name_unicode = utf8_decode(name)[0]
1199
kind = minikind_to_kind[minikind]
1200
inv_entry = factory[kind](file_id, name_unicode,
1202
inv_entry.revision = revid
1204
inv_entry.executable = executable
1205
inv_entry.text_size = size
1206
inv_entry.text_sha1 = link_or_sha1
1207
elif kind == 'directory':
1208
parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1209
elif kind == 'symlink':
1210
inv_entry.executable = False
1211
inv_entry.text_size = size
1212
inv_entry.symlink_target = utf8_decode(link_or_sha1)[0]
1214
raise Exception, kind
1215
# These checks cost us around 40ms on a 55k entry tree
1216
assert file_id not in inv_byid
1217
assert name_unicode not in parent_ie.children
1218
inv_byid[file_id] = inv_entry
1219
parent_ie.children[name_unicode] = inv_entry
1220
self._inventory = inv
1222
def get_file_sha1(self, file_id, path=None, stat_value=None):
1223
# TODO: if path is present, fast-path on that, as inventory
1224
# might not be present
1225
ie = self.inventory[file_id]
1226
if ie.kind == "file":
1230
def get_file(self, file_id):
1231
return StringIO(self.get_file_text(file_id))
1233
def get_file_lines(self, file_id):
1234
ie = self.inventory[file_id]
1235
return self._repository.weave_store.get_weave(file_id,
1236
self._repository.get_transaction()).get_lines(ie.revision)
1238
def get_file_size(self, file_id):
1239
return self.inventory[file_id].text_size
1241
def get_file_text(self, file_id):
1242
return ''.join(self.get_file_lines(file_id))
1244
def get_symlink_target(self, file_id):
1245
entry = self._get_entry(file_id=file_id)
1246
parent_index = self._get_parent_index()
1247
if entry[1][parent_index][0] != 'l':
1250
# At present, none of the tree implementations supports non-ascii
1251
# symlink targets. So we will just assume that the dirstate path is
1253
return entry[1][parent_index][1]
1255
def get_revision_id(self):
1256
"""Return the revision id for this tree."""
1257
return self._revision_id
1259
def _get_inventory(self):
1260
if self._inventory is not None:
1261
return self._inventory
1262
self._generate_inventory()
1263
return self._inventory
1265
inventory = property(_get_inventory,
1266
doc="Inventory of this Tree")
1268
def get_parent_ids(self):
1269
"""The parents of a tree in the dirstate are not cached."""
1270
return self._repository.get_revision(self._revision_id).parent_ids
1272
def has_filename(self, filename):
1273
return bool(self.path2id(filename))
1275
def kind(self, file_id):
1276
return self.inventory[file_id].kind
1278
def is_executable(self, file_id, path=None):
1279
ie = self.inventory[file_id]
1280
if ie.kind != "file":
1282
return ie.executable
1284
def list_files(self, include_root=False):
1285
# We use a standard implementation, because DirStateRevisionTree is
1286
# dealing with one of the parents of the current state
1287
inv = self._get_inventory()
1288
entries = inv.iter_entries()
1289
if self.inventory.root is not None and not include_root:
1291
for path, entry in entries:
1292
yield path, 'V', entry.kind, entry.file_id, entry
1294
def lock_read(self):
1295
"""Lock the tree for a set of operations."""
1296
if not self._locked:
1297
self._repository.lock_read()
1298
if self._dirstate._lock_token is None:
1299
self._dirstate.lock_read()
1300
self._dirstate_locked = True
1304
def path2id(self, path):
1305
"""Return the id for path in this tree."""
1306
# lookup by path: faster than splitting and walking the ivnentory.
1307
entry = self._get_entry(path=path)
1308
if entry == (None, None):
1313
"""Unlock, freeing any cache memory used during the lock."""
1314
# outside of a lock, the inventory is suspect: release it.
1316
if not self._locked:
1317
self._inventory = None
1319
if self._dirstate_locked:
1320
self._dirstate.unlock()
1321
self._dirstate_locked = False
1322
self._repository.unlock()
1324
def walkdirs(self, prefix=""):
1325
# TODO: jam 20070215 This is the cheap way by cheating and using the
1326
# RevisionTree implementation.
1327
# This should be cleaned up to use the much faster Dirstate code
1328
# This is a little tricky, though, because the dirstate is
1329
# indexed by current path, not by parent path.
1330
# So for now, we just build up the parent inventory, and extract
1331
# it the same way RevisionTree does.
1332
_directory = 'directory'
1333
inv = self._get_inventory()
1334
top_id = inv.path2id(prefix)
1338
pending = [(prefix, top_id)]
1341
relpath, file_id = pending.pop()
1342
# 0 - relpath, 1- file-id
1344
relroot = relpath + '/'
1347
# FIXME: stash the node in pending
1348
entry = inv[file_id]
1349
for name, child in entry.sorted_children():
1350
toppath = relroot + name
1351
dirblock.append((toppath, name, child.kind, None,
1352
child.file_id, child.kind
1354
yield (relpath, entry.file_id), dirblock
1355
# push the user specified dirs from dirblock
1356
for dir in reversed(dirblock):
1357
if dir[2] == _directory:
1358
pending.append((dir[0], dir[4]))
1361
class InterDirStateTree(InterTree):
1362
"""Fast path optimiser for changes_from with dirstate trees."""
1364
def __init__(self, source, target):
1365
super(InterDirStateTree, self).__init__(source, target)
1366
if not InterDirStateTree.is_compatible(source, target):
1367
raise Exception, "invalid source %r and target %r" % (source, target)
1370
def make_source_parent_tree(source, target):
1371
"""Change the source tree into a parent of the target."""
1372
revid = source.commit('record tree')
1373
target.branch.repository.fetch(source.branch.repository, revid)
1374
target.set_parent_ids([revid])
1375
return target.basis_tree(), target
1376
_matching_from_tree_format = WorkingTreeFormat4()
1377
_matching_to_tree_format = WorkingTreeFormat4()
1378
_test_mutable_trees_to_test_trees = make_source_parent_tree
1380
def _iter_changes(self, include_unchanged=False,
1381
specific_files=None, pb=None, extra_trees=[],
1382
require_versioned=True):
1383
"""Return the changes from source to target.
1385
:return: An iterator that yields tuples. See InterTree._iter_changes
1387
:param specific_files: An optional list of file paths to restrict the
1388
comparison to. When mapping filenames to ids, all matches in all
1389
trees (including optional extra_trees) are used, and all children of
1390
matched directories are included.
1391
:param include_unchanged: An optional boolean requesting the inclusion of
1392
unchanged entries in the result.
1393
:param extra_trees: An optional list of additional trees to use when
1394
mapping the contents of specific_files (paths) to file_ids.
1395
:param require_versioned: If True, all files in specific_files must be
1396
versioned in one of source, target, extra_trees or
1397
PathsNotVersionedError is raised.
1399
# NB: show_status depends on being able to pass in non-versioned files
1400
# and report them as unknown
1401
# TODO: handle extra trees in the dirstate.
1402
# TODO: handle comparisons as an empty tree as a different special
1403
# case? mbp 20070226
1404
if extra_trees or (self.source._revision_id == NULL_REVISION):
1405
# we can't fast-path these cases (yet)
1406
for f in super(InterDirStateTree, self)._iter_changes(
1407
include_unchanged, specific_files, pb, extra_trees,
1411
assert (self.source._revision_id in self.target.get_parent_ids()), \
1412
"revision {%s} is not stored in {%s}, but %s " \
1413
"can only be used for trees stored in the dirstate" \
1414
% (self.source._revision_id, self.target, self._iter_changes)
1415
parents = self.target.get_parent_ids()
1417
source_index = 1 + parents.index(self.source._revision_id)
1418
# -- make all specific_files utf8 --
1420
specific_files_utf8 = set()
1421
for path in specific_files:
1422
specific_files_utf8.add(path.encode('utf8'))
1423
specific_files = specific_files_utf8
1425
specific_files = set([''])
1426
# -- specific_files is now a utf8 path set --
1427
# -- get the state object and prepare it.
1428
state = self.target.current_dirstate()
1429
state._read_dirblocks_if_needed()
1430
def _entries_for_path(path):
1431
"""Return a list with all the entries that match path for all ids.
1433
dirname, basename = os.path.split(path)
1434
key = (dirname, basename, '')
1435
block_index, present = state._find_block_index_from_key(key)
1437
# the block which should contain path is absent.
1440
block = state._dirblocks[block_index][1]
1441
entry_index, _ = state._find_entry_index(key, block)
1442
# we may need to look at multiple entries at this path: walk while the specific_files match.
1443
while (entry_index < len(block) and
1444
block[entry_index][0][0:2] == key[0:2]):
1445
result.append(block[entry_index])
1448
if require_versioned:
1449
# -- check all supplied paths are versioned in a search tree. --
1450
all_versioned = True
1451
for path in specific_files:
1452
path = path.encode('utf8')
1453
path_entries = _entries_for_path(path)
1454
if not path_entries:
1455
# this specified path is not present at all: error
1456
all_versioned = False
1458
found_versioned = False
1459
# for each id at this path
1460
for entry in path_entries:
1462
for index in source_index, target_index:
1463
if entry[1][index][0] != 'a': # absent
1464
found_versioned = True
1465
# all good: found a versioned cell
1467
if not found_versioned:
1468
# none of the indexes was not 'absent' at all ids for this
1470
all_versioned = False
1472
if not all_versioned:
1473
raise errors.PathsNotVersionedError(paths)
1474
# -- remove redundancy in supplied specific_files to prevent over-scanning --
1475
search_specific_files = set()
1476
for path in specific_files:
1477
other_specific_files = specific_files.difference(set([path]))
1478
if not osutils.is_inside_any(other_specific_files, path):
1479
# this is a top level path, we must check it.
1480
search_specific_files.add(path)
1482
# compare source_index and target_index at or under each element of search_specific_files.
1483
# follow the following comparison table. Note that we only want to do diff operations when
1484
# the target is fdl because thats when the walkdirs logic will have exposed the pathinfo
1488
# Source | Target | disk | action
1489
# r | fdl | | add source to search, add id path move and perform
1490
# | | | diff check on source-target
1491
# r | fdl | a | dangling file that was present in the basis.
1493
# r | a | | add source to search
1495
# r | r | | this path is present in a non-examined tree, skip.
1496
# r | r | a | this path is present in a non-examined tree, skip.
1497
# a | fdl | | add new id
1498
# a | fdl | a | dangling locally added file, skip
1499
# a | a | | not present in either tree, skip
1500
# a | a | a | not present in any tree, skip
1501
# a | r | | not present in either tree at this path, skip as it
1502
# | | | may not be selected by the users list of paths.
1503
# a | r | a | not present in either tree at this path, skip as it
1504
# | | | may not be selected by the users list of paths.
1505
# fdl | fdl | | content in both: diff them
1506
# fdl | fdl | a | deleted locally, but not unversioned - show as deleted ?
1507
# fdl | a | | unversioned: output deleted id for now
1508
# fdl | a | a | unversioned and deleted: output deleted id
1509
# fdl | r | | relocated in this tree, so add target to search.
1510
# | | | Dont diff, we will see an r,fd; pair when we reach
1511
# | | | this id at the other path.
1512
# fdl | r | a | relocated in this tree, so add target to search.
1513
# | | | Dont diff, we will see an r,fd; pair when we reach
1514
# | | | this id at the other path.
1516
# for all search_indexs in each path at or under each element of
1517
# search_specific_files, if the detail is relocated: add the id, and add the
1518
# relocated path as one to search if its not searched already. If the
1519
# detail is not relocated, add the id.
1520
searched_specific_files = set()
1521
def _process_entry(entry, path_info):
1522
"""Compare an entry and real disk to generate delta information.
1524
:param path_info: top_relpath, basename, kind, lstat, abspath for
1525
the path of entry. If None, then the path is considered absent.
1526
(Perhaps we should pass in a concrete entry for this ?)
1528
# TODO: when a parent has been renamed, dont emit path renames for children,
1529
source_details = entry[1][source_index]
1530
target_details = entry[1][target_index]
1531
if source_details[0] in 'rfdl' and target_details[0] in 'fdl':
1532
# claimed content in both: diff
1533
# r | fdl | | add source to search, add id path move and perform
1534
# | | | diff check on source-target
1535
# r | fdl | a | dangling file that was present in the basis.
1537
if source_details[0] in 'r':
1538
# add the source to the search path to find any children it
1539
# has. TODO ? : only add if it is a container ?
1540
if not osutils.is_inside_any(searched_specific_files, source_details[1]):
1541
search_specific_files.add(source_details[1])
1542
# generate the old path; this is needed for stating later
1544
old_path = source_details[1]
1545
old_dirname, old_basename = os.path.split(old_path)
1546
path = os.path.join(*entry[0][0:2])
1547
old_entry = state._get_entry(source_index, path_utf8=old_path)
1548
# update the source details variable to be the real
1550
source_details = old_entry[1][source_index]
1552
old_path = path = os.path.join(*entry[0][0:2])
1553
old_dirname, old_basename = entry[0][0:2]
1554
if path_info is None:
1555
# the file is missing on disk, show as removed.
1556
print "missing file"
1557
old_path = os.path.join(*entry[0][0:2])
1558
result.removed.append((old_path, entry[0][2], dirstate.DirState._minikind_to_kind[source_details[0]]))
1559
# use the kind from disk.
1560
elif source_details[0] != path_info[2][0]:
1562
import pdb;pdb.set_trace()
1566
if path_info[2][0] == 'd':
1567
# directories have no fingerprint
1568
content_change = False
1569
executable_change = False
1570
elif path_info[2][0] == 'f':
1571
# has it changed? fast path: size, slow path: sha1.
1572
executable_change = source_details[3] != bool(
1573
stat.S_ISREG(path_info[3].st_mode)
1574
and stat.S_IEXEC & path_info[3].st_mode)
1575
if source_details[2] != path_info[3].st_size:
1576
content_change = True
1578
# maybe the same. Get the hash
1579
new_hash = self.target._hashcache.get_sha1(path, path_info[3])
1580
content_change = (new_hash != source_details[1])
1581
elif path_info[2][0] == 'l':
1582
import pdb;pdb.set_trace()
1585
raise Exception, "unknown minikind"
1586
# parent id is the entry for the path in the target tree
1587
# TODO: the target is the same for an entire directory: cache em.
1588
source_parent_id = state._get_entry(source_index, path_utf8=old_dirname)[0][2]
1589
if source_parent_id == entry[0][2]:
1590
source_parent_id = None
1591
target_parent_id = state._get_entry(target_index, path_utf8=entry[0][0])[0][2]
1592
if target_parent_id == entry[0][2]:
1593
target_parent_id = None
1594
source_exec = source_details[3]
1596
stat.S_ISREG(path_info[3].st_mode)
1597
and stat.S_IEXEC & path_info[3].st_mode)
1598
return ((entry[0][2], path, content_change, (True, True), (source_parent_id, target_parent_id), (old_basename, entry[0][1]), (dirstate.DirState._minikind_to_kind[source_details[0]], path_info[2]), (source_exec, target_exec)),)
1599
elif source_details[0] in 'a' and target_details[0] in 'fdl':
1600
# looks like a new file
1601
if path_info is not None:
1602
path = os.path.join(*entry[0][0:2])
1603
# parent id is the entry for the path in the target tree
1604
# TODO: these are the same for an entire directory: cache em.
1605
parent_id = state._get_entry(target_index, path_utf8=entry[0][0])[0][2]
1606
if parent_id == entry[0][2]:
1609
new_executable = bool(
1610
stat.S_ISREG(path_info[3].st_mode)
1611
and stat.S_IEXEC & path_info[3].st_mode)
1612
return ((entry[0][2], path, True, (False, True), (None, parent_id), (None, entry[0][1]), (None, path_info[2]), (None, new_executable)),)
1614
# but its not on disk: we deliberately treat this as just
1615
# never-present. (Why ?! - RBC 20070224)
1617
elif source_details[0] in 'fdl' and target_details[0] in 'a':
1618
# unversioned, possibly, or possibly not deleted: we dont care.
1619
# if its still on disk, *and* theres no other entry at this
1620
# path [we dont know this in this routine at the moment -
1621
# perhaps we should change this - then it would be an unknown.
1622
old_path = os.path.join(*entry[0][0:2])
1623
# parent id is the entry for the path in the target tree
1624
parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
1625
if parent_id == entry[0][2]:
1627
return ((entry[0][2], old_path, True, (True, False), (parent_id, None), (entry[0][1], None), (dirstate.DirState._minikind_to_kind[source_details[0]], None), (source_details[3], None)),)
1628
elif source_details[0] in 'fdl' and target_details[0] in 'r':
1629
# a rename; could be a true rename, or a rename inherited from
1630
# a renamed parent. TODO: handle this efficiently. Its not
1631
# common case to rename dirs though, so a correct but slow
1632
# implementation will do.
1633
if not osutils.is_inside_any(searched_specific_files, target_details[1]):
1634
search_specific_files.add(target_details[1])
1636
import pdb;pdb.set_trace()
1638
while search_specific_files:
1639
# TODO: the pending list should be lexically sorted?
1640
current_root = search_specific_files.pop()
1641
searched_specific_files.add(current_root)
1642
# process the entries for this containing directory: the rest will be
1643
# found by their parents recursively.
1644
root_entries = _entries_for_path(current_root)
1645
root_abspath = self.target.abspath(current_root)
1647
root_stat = os.lstat(root_abspath)
1649
if e.errno == errno.ENOENT:
1650
# the path does not exist: let _process_entry know that.
1651
root_dir_info = None
1653
# some other random error: hand it up.
1656
root_dir_info = ('', current_root,
1657
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1659
if not root_entries and not root_dir_info:
1660
# this specified path is not present at all, skip it.
1662
for entry in root_entries:
1663
for result in _process_entry(entry, root_dir_info):
1664
# this check should probably be outside the loop: one
1665
# 'iterate two trees' api, and then _iter_changes filters
1666
# unchanged pairs. - RBC 20070226
1667
if include_unchanged or result[2] or True in map(lambda x:x[0]!=x[1], result[3:8]):
1669
dir_iterator = osutils.walkdirs(root_abspath, prefix=current_root)
1670
initial_key = (current_root, '', '')
1671
block_index, _ = state._find_block_index_from_key(initial_key)
1672
if block_index == 0:
1673
# we have processed the total root already, but because the
1674
# initial key matched it we sould skip it here.
1677
current_dir_info = dir_iterator.next()
1679
if e.errno in (errno.ENOENT, errno.ENOTDIR):
1680
# there may be directories in the inventory even though
1681
# this path is not a file on disk: so mark it as end of
1683
current_dir_info = None
1687
if current_dir_info[0][0] == '':
1688
# remove .bzr from iteration
1689
bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
1690
assert current_dir_info[1][bzr_index][0] == '.bzr'
1691
del current_dir_info[1][bzr_index]
1692
# convert the unicode relpaths in the dir index to uf8 for
1693
# comparison with dirstate data.
1694
# TODO: keep the utf8 version around for giving to the caller.
1695
current_dir_info = ((current_dir_info[0][0].encode('utf8'), current_dir_info[0][1]),
1696
[(line[0].encode('utf8'), line[1].encode('utf8')) + line[2:] for line in current_dir_info[1]])
1697
# walk until both the directory listing and the versioned metadata
1698
# are exhausted. TODO: reevaluate this, perhaps we should stop when
1699
# the versioned data runs out.
1700
if (block_index < len(state._dirblocks) and
1701
osutils.is_inside(current_root, state._dirblocks[block_index][0])):
1702
current_block = state._dirblocks[block_index]
1704
current_block = None
1705
while (current_dir_info is not None or
1706
current_block is not None):
1707
if current_dir_info and current_block and current_dir_info[0][0] != current_block[0]:
1708
if current_block[0] < current_dir_info[0][0]:
1709
# extra dir on disk: pass for now? should del from info ?
1710
import pdb;pdb.set_trace()
1711
print 'unversioned dir'
1713
# entry referring to missing dir.
1714
import pdb;pdb.set_trace()
1717
if current_block and entry_index < len(current_block[1]):
1718
current_entry = current_block[1][entry_index]
1720
current_entry = None
1721
advance_entry = True
1723
if current_dir_info and path_index < len(current_dir_info[1]):
1724
current_path_info = current_dir_info[1][path_index]
1726
current_path_info = None
1728
while (current_entry is not None or
1729
current_path_info is not None):
1730
if current_entry is None:
1731
# no more entries: yield current_pathinfo as an
1732
# unversioned file: its not the same as a path in any
1733
# tree in the dirstate.
1734
new_executable = bool(
1735
stat.S_ISREG(current_path_info[3].st_mode)
1736
and stat.S_IEXEC & current_path_info[3].st_mode)
1737
yield (None, current_path_info[0], True, (False, False), (None, None), (None, current_path_info[1]), (None, current_path_info[2]), (None, new_executable))
1738
elif current_path_info is None:
1739
# no path is fine: the per entry code will handle it.
1740
for result in _process_entry(current_entry, current_path_info):
1741
# this check should probably be outside the loop: one
1742
# 'iterate two trees' api, and then _iter_changes filters
1743
# unchanged pairs. - RBC 20070226
1744
if include_unchanged or result[2] or True in map(lambda x:x[0]!=x[1], result[3:8]):
1746
elif current_entry[0][1] != current_path_info[1]:
1747
if current_path_info[1] < current_entry[0][1]:
1748
# extra file on disk: pass for now
1749
import pdb;pdb.set_trace()
1750
print 'unversioned file'
1752
# entry referring to file not present on disk.
1753
# advance the entry only, after processing.
1754
for result in _process_entry(current_entry, None):
1755
# this check should probably be outside the loop: one
1756
# 'iterate two trees' api, and then _iter_changes filters
1757
# unchanged pairs. - RBC 20070226
1758
if include_unchanged or result[2] or True in map(lambda x:x[0]!=x[1], result[3:8]):
1760
advance_path = False
1762
for result in _process_entry(current_entry, current_path_info):
1763
# this check should probably be outside the loop: one
1764
# 'iterate two trees' api, and then _iter_changes filters
1765
# unchanged pairs. - RBC 20070226
1766
if include_unchanged or result[2] or True in map(lambda x:x[0]!=x[1], result[3:8]):
1768
if advance_entry and current_entry is not None:
1770
if entry_index < len(current_block[1]):
1771
current_entry = current_block[1][entry_index]
1773
current_entry = None
1775
advance_entry = True # reset the advance flaga
1776
if advance_path and current_path_info is not None:
1778
if path_index < len(current_dir_info[1]):
1779
current_path_info = current_dir_info[1][path_index]
1781
current_path_info = None
1783
advance_path = True # reset the advance flagg.
1784
if current_block is not None:
1786
if (block_index < len(state._dirblocks) and
1787
osutils.is_inside(current_root, state._dirblocks[block_index][0])):
1788
current_block = state._dirblocks[block_index]
1790
current_block = None
1791
if current_dir_info is not None:
1793
current_dir_info = dir_iterator.next()
1794
# convert the unicode relpaths in the dir index to uf8 for
1795
# comparison with dirstate data.
1796
# TODO: keep the utf8 version around for giving to the caller.
1797
current_dir_info = ((current_dir_info[0][0].encode('utf8'), current_dir_info[0][1]),
1798
[(line[0].encode('utf8'), line[1].encode('utf8')) + line[2:] for line in current_dir_info[1]])
1799
except StopIteration:
1800
current_dir_info = None
1804
def is_compatible(source, target):
1805
# the target must be a dirstate working tree
1806
if not isinstance(target, WorkingTree4):
1808
# the source must be a revtreee or dirstate rev tree.
1809
if not isinstance(source,
1810
(revisiontree.RevisionTree, DirStateRevisionTree)):
1812
# the source revid must be in the target dirstate
1813
if not (source._revision_id == NULL_REVISION or
1814
source._revision_id in target.get_parent_ids()):
1815
# TODO: what about ghosts? it may well need to
1816
# check for them explicitly.
1820
InterTree.register_optimiser(InterDirStateTree)