1
# Copyright (C) 2005 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
"""WorkingTree object and friends.
19
A WorkingTree represents the editable working copy of a branch.
20
Operations which represent the WorkingTree are also done here,
21
such as renaming or adding files. The WorkingTree has an inventory
22
which is updated by these operations. A commit produces a
23
new revision based on the workingtree and its inventory.
25
At the moment every WorkingTree has its own branch. Remote
26
WorkingTrees aren't supported.
28
To get a WorkingTree, call Branch.working_tree():
32
# TODO: Don't allow WorkingTrees to be constructed for remote branches if
35
# FIXME: I don't know if writing out the cache from the destructor is really a
36
# good idea, because destructors are considered poor taste in Python, and it's
37
# not predictable when it will be written out.
39
# TODO: Give the workingtree sole responsibility for the working inventory;
40
# remove the variable and references to it from the branch. This may require
41
# updating the commit code so as to update the inventory within the working
42
# copy, and making sure there's only one WorkingTree for any directory on disk.
43
# At the momenthey may alias the inventory and have old copies of it in memory.
45
from copy import deepcopy
50
from bzrlib.branch import (Branch,
53
from bzrlib.errors import (BzrCheckError,
56
WeaveRevisionNotPresent,
60
from bzrlib.inventory import InventoryEntry
61
from bzrlib.osutils import (appendpath,
75
from bzrlib.textui import show_status
77
from bzrlib.trace import mutter
79
from bzrlib.decorators import needs_read_lock, needs_write_lock
82
def gen_file_id(name):
83
"""Return new file id.
85
This should probably generate proper UUIDs, but for the moment we
86
cope with just randomness because running uuidgen every time is
89
from binascii import hexlify
96
idx = name.rfind('\\')
100
# make it not a hidden file
101
name = name.lstrip('.')
103
# remove any wierd characters; we don't escape them but rather
105
name = re.sub(r'[^\w.]', '', name)
107
s = hexlify(rand_bytes(8))
108
return '-'.join((name, compact_date(time()), s))
112
"""Return a new tree-root file id."""
113
return gen_file_id('TREE_ROOT')
116
class TreeEntry(object):
117
"""An entry that implements the minium interface used by commands.
119
This needs further inspection, it may be better to have
120
InventoryEntries without ids - though that seems wrong. For now,
121
this is a parallel hierarchy to InventoryEntry, and needs to become
122
one of several things: decorates to that hierarchy, children of, or
124
Another note is that these objects are currently only used when there is
125
no InventoryEntry available - i.e. for unversioned objects.
126
Perhaps they should be UnversionedEntry et al. ? - RBC 20051003
129
def __eq__(self, other):
130
# yes, this us ugly, TODO: best practice __eq__ style.
131
return (isinstance(other, TreeEntry)
132
and other.__class__ == self.__class__)
134
def kind_character(self):
138
class TreeDirectory(TreeEntry):
139
"""See TreeEntry. This is a directory in a working tree."""
141
def __eq__(self, other):
142
return (isinstance(other, TreeDirectory)
143
and other.__class__ == self.__class__)
145
def kind_character(self):
149
class TreeFile(TreeEntry):
150
"""See TreeEntry. This is a regular file in a working tree."""
152
def __eq__(self, other):
153
return (isinstance(other, TreeFile)
154
and other.__class__ == self.__class__)
156
def kind_character(self):
160
class TreeLink(TreeEntry):
161
"""See TreeEntry. This is a symlink in a working tree."""
163
def __eq__(self, other):
164
return (isinstance(other, TreeLink)
165
and other.__class__ == self.__class__)
167
def kind_character(self):
171
class WorkingTree(bzrlib.tree.Tree):
172
"""Working copy tree.
174
The inventory is held in the `Branch` working-inventory, and the
175
files are in a directory on disk.
177
It is possible for a `WorkingTree` to have a filename which is
178
not listed in the Inventory and vice versa.
181
def __init__(self, basedir=u'.', branch=None):
182
"""Construct a WorkingTree for basedir.
184
If the branch is not supplied, it is opened automatically.
185
If the branch is supplied, it must be the branch for this basedir.
186
(branch.base is not cross checked, because for remote branches that
187
would be meaningless).
189
from bzrlib.hashcache import HashCache
190
from bzrlib.trace import note, mutter
191
assert isinstance(basedir, basestring), \
192
"base directory %r is not a string" % basedir
194
branch = Branch.open(basedir)
195
assert isinstance(branch, Branch), \
196
"branch %r is not a Branch" % branch
198
self.basedir = realpath(basedir)
200
# update the whole cache up front and write to disk if anything changed;
201
# in the future we might want to do this more selectively
202
# two possible ways offer themselves : in self._unlock, write the cache
203
# if needed, or, when the cache sees a change, append it to the hash
204
# cache file, and have the parser take the most recent entry for a
206
hc = self._hashcache = HashCache(basedir)
214
self._set_inventory(self.read_working_inventory())
216
def _set_inventory(self, inv):
217
self._inventory = inv
218
self.path2id = self._inventory.path2id
221
def open_containing(path=None):
222
"""Open an existing working tree which has its root about path.
224
This probes for a working tree at path and searches upwards from there.
226
Basically we keep looking up until we find the control directory or
227
run into /. If there isn't one, raises NotBranchError.
228
TODO: give this a new exception.
229
If there is one, it is returned, along with the unused portion of path.
235
if path.find('://') != -1:
236
raise NotBranchError(path=path)
242
return WorkingTree(path), tail
243
except NotBranchError:
246
tail = pathjoin(os.path.basename(path), tail)
248
tail = os.path.basename(path)
250
path = os.path.dirname(path)
252
# reached the root, whatever that may be
253
raise NotBranchError(path=orig_path)
256
"""Iterate through file_ids for this tree.
258
file_ids are in a WorkingTree if they are in the working inventory
259
and the working file exists.
261
inv = self._inventory
262
for path, ie in inv.iter_entries():
263
if bzrlib.osutils.lexists(self.abspath(path)):
267
return "<%s of %s>" % (self.__class__.__name__,
268
getattr(self, 'basedir', None))
270
def abspath(self, filename):
271
return pathjoin(self.basedir, filename)
273
def relpath(self, abs):
274
"""Return the local path portion from a given absolute path."""
275
return relpath(self.basedir, abs)
277
def has_filename(self, filename):
278
return bzrlib.osutils.lexists(self.abspath(filename))
280
def get_file(self, file_id):
281
return self.get_file_byname(self.id2path(file_id))
283
def get_file_byname(self, filename):
284
return file(self.abspath(filename), 'rb')
286
def get_root_id(self):
287
"""Return the id of this trees root"""
288
inv = self.read_working_inventory()
289
return inv.root.file_id
291
def _get_store_filename(self, file_id):
292
## XXX: badly named; this is not in the store at all
293
return self.abspath(self.id2path(file_id))
296
def commit(self, *args, **kw):
297
from bzrlib.commit import Commit
298
Commit().commit(self.branch, *args, **kw)
299
self._set_inventory(self.read_working_inventory())
301
def id2abspath(self, file_id):
302
return self.abspath(self.id2path(file_id))
304
def has_id(self, file_id):
305
# files that have been deleted are excluded
306
inv = self._inventory
307
if not inv.has_id(file_id):
309
path = inv.id2path(file_id)
310
return bzrlib.osutils.lexists(self.abspath(path))
312
def has_or_had_id(self, file_id):
313
if file_id == self.inventory.root.file_id:
315
return self.inventory.has_id(file_id)
317
__contains__ = has_id
319
def get_file_size(self, file_id):
320
return os.path.getsize(self.id2abspath(file_id))
323
def get_file_sha1(self, file_id):
324
path = self._inventory.id2path(file_id)
325
return self._hashcache.get_sha1(path)
327
def is_executable(self, file_id):
329
return self._inventory[file_id].executable
331
path = self._inventory.id2path(file_id)
332
mode = os.lstat(self.abspath(path)).st_mode
333
return bool(stat.S_ISREG(mode) and stat.S_IEXEC&mode)
336
def add(self, files, ids=None):
337
"""Make files versioned.
339
Note that the command line normally calls smart_add instead,
340
which can automatically recurse.
342
This adds the files to the inventory, so that they will be
343
recorded by the next commit.
346
List of paths to add, relative to the base of the tree.
349
If set, use these instead of automatically generated ids.
350
Must be the same length as the list of files, but may
351
contain None for ids that are to be autogenerated.
353
TODO: Perhaps have an option to add the ids even if the files do
356
TODO: Perhaps callback with the ids and paths as they're added.
358
# TODO: Re-adding a file that is removed in the working copy
359
# should probably put it back with the previous ID.
360
if isinstance(files, basestring):
361
assert(ids is None or isinstance(ids, basestring))
367
ids = [None] * len(files)
369
assert(len(ids) == len(files))
371
inv = self.read_working_inventory()
372
for f,file_id in zip(files, ids):
373
if is_control_file(f):
374
raise BzrError("cannot add control file %s" % quotefn(f))
379
raise BzrError("cannot add top-level %r" % f)
381
fullpath = normpath(self.abspath(f))
384
kind = file_kind(fullpath)
386
# maybe something better?
387
raise BzrError('cannot add: not a regular file, symlink or directory: %s' % quotefn(f))
389
if not InventoryEntry.versionable_kind(kind):
390
raise BzrError('cannot add: not a versionable file ('
391
'i.e. regular file, symlink or directory): %s' % quotefn(f))
394
file_id = gen_file_id(f)
395
inv.add_path(f, kind=kind, file_id=file_id)
397
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
398
self._write_inventory(inv)
401
def add_pending_merge(self, *revision_ids):
402
# TODO: Perhaps should check at this point that the
403
# history of the revision is actually present?
404
p = self.pending_merges()
406
for rev_id in revision_ids:
412
self.set_pending_merges(p)
415
def pending_merges(self):
416
"""Return a list of pending merges.
418
These are revisions that have been merged into the working
419
directory but not yet committed.
422
f = self.branch.control_files.get_utf8('pending-merges')
426
for l in f.readlines():
427
p.append(l.rstrip('\n'))
431
def set_pending_merges(self, rev_list):
432
self.branch.control_files.put_utf8('pending-merges', '\n'.join(rev_list))
434
def get_symlink_target(self, file_id):
435
return os.readlink(self.id2abspath(file_id))
437
def file_class(self, filename):
438
if self.path2id(filename):
440
elif self.is_ignored(filename):
446
def list_files(self):
447
"""Recursively list all files as (path, class, kind, id).
449
Lists, but does not descend into unversioned directories.
451
This does not include files that have been deleted in this
454
Skips the control directory.
456
inv = self._inventory
458
def descend(from_dir_relpath, from_dir_id, dp):
462
## TODO: If we find a subdirectory with its own .bzr
463
## directory, then that is a separate tree and we
464
## should exclude it.
465
if bzrlib.BZRDIR == f:
469
fp = appendpath(from_dir_relpath, f)
472
fap = appendpath(dp, f)
474
f_ie = inv.get_child(from_dir_id, f)
477
elif self.is_ignored(fp):
486
raise BzrCheckError("file %r entered as kind %r id %r, "
488
% (fap, f_ie.kind, f_ie.file_id, fk))
490
# make a last minute entry
494
if fk == 'directory':
495
entry = TreeDirectory()
498
elif fk == 'symlink':
503
yield fp, c, fk, (f_ie and f_ie.file_id), entry
505
if fk != 'directory':
509
# don't descend unversioned directories
512
for ff in descend(fp, f_ie.file_id, fap):
515
for f in descend(u'', inv.root.file_id, self.basedir):
519
def move(self, from_paths, to_name):
522
to_name must exist in the inventory.
524
If to_name exists and is a directory, the files are moved into
525
it, keeping their old names.
527
Note that to_name is only the last component of the new name;
528
this doesn't change the directory.
530
This returns a list of (from_path, to_path) pairs for each
534
## TODO: Option to move IDs only
535
assert not isinstance(from_paths, basestring)
537
to_abs = self.abspath(to_name)
538
if not isdir(to_abs):
539
raise BzrError("destination %r is not a directory" % to_abs)
540
if not self.has_filename(to_name):
541
raise BzrError("destination %r not in working directory" % to_abs)
542
to_dir_id = inv.path2id(to_name)
543
if to_dir_id == None and to_name != '':
544
raise BzrError("destination %r is not a versioned directory" % to_name)
545
to_dir_ie = inv[to_dir_id]
546
if to_dir_ie.kind not in ('directory', 'root_directory'):
547
raise BzrError("destination %r is not a directory" % to_abs)
549
to_idpath = inv.get_idpath(to_dir_id)
552
if not self.has_filename(f):
553
raise BzrError("%r does not exist in working tree" % f)
554
f_id = inv.path2id(f)
556
raise BzrError("%r is not versioned" % f)
557
name_tail = splitpath(f)[-1]
558
dest_path = appendpath(to_name, name_tail)
559
if self.has_filename(dest_path):
560
raise BzrError("destination %r already exists" % dest_path)
561
if f_id in to_idpath:
562
raise BzrError("can't move %r to a subdirectory of itself" % f)
564
# OK, so there's a race here, it's possible that someone will
565
# create a file in this interval and then the rename might be
566
# left half-done. But we should have caught most problems.
567
orig_inv = deepcopy(self.inventory)
570
name_tail = splitpath(f)[-1]
571
dest_path = appendpath(to_name, name_tail)
572
result.append((f, dest_path))
573
inv.rename(inv.path2id(f), to_dir_id, name_tail)
575
rename(self.abspath(f), self.abspath(dest_path))
577
raise BzrError("failed to rename %r to %r: %s" %
578
(f, dest_path, e[1]),
579
["rename rolled back"])
581
# restore the inventory on error
582
self._set_inventory(orig_inv)
584
self._write_inventory(inv)
588
def rename_one(self, from_rel, to_rel):
591
This can change the directory or the filename or both.
594
if not self.has_filename(from_rel):
595
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
596
if self.has_filename(to_rel):
597
raise BzrError("can't rename: new working file %r already exists" % to_rel)
599
file_id = inv.path2id(from_rel)
601
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
604
from_parent = entry.parent_id
605
from_name = entry.name
607
if inv.path2id(to_rel):
608
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
610
to_dir, to_tail = os.path.split(to_rel)
611
to_dir_id = inv.path2id(to_dir)
612
if to_dir_id == None and to_dir != '':
613
raise BzrError("can't determine destination directory id for %r" % to_dir)
615
mutter("rename_one:")
616
mutter(" file_id {%s}" % file_id)
617
mutter(" from_rel %r" % from_rel)
618
mutter(" to_rel %r" % to_rel)
619
mutter(" to_dir %r" % to_dir)
620
mutter(" to_dir_id {%s}" % to_dir_id)
622
inv.rename(file_id, to_dir_id, to_tail)
624
from_abs = self.abspath(from_rel)
625
to_abs = self.abspath(to_rel)
627
rename(from_abs, to_abs)
629
inv.rename(file_id, from_parent, from_name)
630
raise BzrError("failed to rename %r to %r: %s"
631
% (from_abs, to_abs, e[1]),
632
["rename rolled back"])
633
self._write_inventory(inv)
637
"""Return all unknown files.
639
These are files in the working directory that are not versioned or
640
control files or ignored.
642
>>> from bzrlib.branch import ScratchBranch
643
>>> b = ScratchBranch(files=['foo', 'foo~'])
644
>>> tree = WorkingTree(b.base, b)
645
>>> map(str, tree.unknowns())
648
>>> list(b.unknowns())
650
>>> tree.remove('foo')
651
>>> list(b.unknowns())
654
for subp in self.extras():
655
if not self.is_ignored(subp):
658
def iter_conflicts(self):
660
for path in (s[0] for s in self.list_files()):
661
stem = get_conflicted_stem(path)
664
if stem not in conflicted:
669
def pull(self, source, overwrite=False):
670
from bzrlib.merge import merge_inner
673
old_revision_history = self.branch.revision_history()
674
count = self.branch.pull(source, overwrite)
675
new_revision_history = self.branch.revision_history()
676
if new_revision_history != old_revision_history:
677
if len(old_revision_history):
678
other_revision = old_revision_history[-1]
680
other_revision = None
681
repository = self.branch.repository
682
merge_inner(self.branch,
683
self.branch.basis_tree(),
684
repository.revision_tree(other_revision))
690
"""Yield all unknown files in this WorkingTree.
692
If there are any unknown directories then only the directory is
693
returned, not all its children. But if there are unknown files
694
under a versioned subdirectory, they are returned.
696
Currently returned depth-first, sorted by name within directories.
698
## TODO: Work from given directory downwards
699
for path, dir_entry in self.inventory.directories():
700
mutter("search for unknowns in %r", path)
701
dirabs = self.abspath(path)
702
if not isdir(dirabs):
703
# e.g. directory deleted
707
for subf in os.listdir(dirabs):
709
and (subf not in dir_entry.children)):
714
subp = appendpath(path, subf)
718
def ignored_files(self):
719
"""Yield list of PATH, IGNORE_PATTERN"""
720
for subp in self.extras():
721
pat = self.is_ignored(subp)
726
def get_ignore_list(self):
727
"""Return list of ignore patterns.
729
Cached in the Tree object after the first call.
731
if hasattr(self, '_ignorelist'):
732
return self._ignorelist
734
l = bzrlib.DEFAULT_IGNORE[:]
735
if self.has_filename(bzrlib.IGNORE_FILENAME):
736
f = self.get_file_byname(bzrlib.IGNORE_FILENAME)
737
l.extend([line.rstrip("\n\r") for line in f.readlines()])
742
def is_ignored(self, filename):
743
r"""Check whether the filename matches an ignore pattern.
745
Patterns containing '/' or '\' need to match the whole path;
746
others match against only the last component.
748
If the file is ignored, returns the pattern which caused it to
749
be ignored, otherwise None. So this can simply be used as a
750
boolean if desired."""
752
# TODO: Use '**' to match directories, and other extended
753
# globbing stuff from cvs/rsync.
755
# XXX: fnmatch is actually not quite what we want: it's only
756
# approximately the same as real Unix fnmatch, and doesn't
757
# treat dotfiles correctly and allows * to match /.
758
# Eventually it should be replaced with something more
761
for pat in self.get_ignore_list():
762
if '/' in pat or '\\' in pat:
764
# as a special case, you can put ./ at the start of a
765
# pattern; this is good to match in the top-level
768
if (pat[:2] == './') or (pat[:2] == '.\\'):
772
if fnmatch.fnmatchcase(filename, newpat):
775
if fnmatch.fnmatchcase(splitpath(filename)[-1], pat):
780
def kind(self, file_id):
781
return file_kind(self.id2abspath(file_id))
784
"""See Branch.lock_read, and WorkingTree.unlock."""
785
return self.branch.lock_read()
787
def lock_write(self):
788
"""See Branch.lock_write, and WorkingTree.unlock."""
789
return self.branch.lock_write()
791
def _basis_inventory_name(self, revision_id):
792
return 'basis-inventory.%s' % revision_id
794
def set_last_revision(self, new_revision, old_revision=None):
795
if old_revision is not None:
797
path = self._basis_inventory_name(old_revision)
798
path = self.branch.control_files._escape(path)
799
self.branch.control_files._transport.delete(path)
803
xml = self.branch.repository.get_inventory_xml(new_revision)
804
path = self._basis_inventory_name(new_revision)
805
self.branch.control_files.put_utf8(path, xml)
806
except WeaveRevisionNotPresent:
809
def read_basis_inventory(self, revision_id):
810
"""Read the cached basis inventory."""
811
path = self._basis_inventory_name(revision_id)
812
return self.branch.control_files.get_utf8(path).read()
815
def read_working_inventory(self):
816
"""Read the working inventory."""
817
# ElementTree does its own conversion from UTF-8, so open in
819
f = self.branch.control_files.get('inventory')
820
return bzrlib.xml5.serializer_v5.read_inventory(f)
823
def remove(self, files, verbose=False):
824
"""Remove nominated files from the working inventory..
826
This does not remove their text. This does not run on XXX on what? RBC
828
TODO: Refuse to remove modified files unless --force is given?
830
TODO: Do something useful with directories.
832
TODO: Should this remove the text or not? Tough call; not
833
removing may be useful and the user can just use use rm, and
834
is the opposite of add. Removing it is consistent with most
835
other tools. Maybe an option.
837
## TODO: Normalize names
838
## TODO: Remove nested loops; better scalability
839
if isinstance(files, basestring):
844
# do this before any modifications
848
# TODO: Perhaps make this just a warning, and continue?
849
# This tends to happen when
850
raise NotVersionedError(path=f)
851
mutter("remove inventory entry %s {%s}", quotefn(f), fid)
853
# having remove it, it must be either ignored or unknown
854
if self.is_ignored(f):
858
show_status(new_status, inv[fid].kind, quotefn(f))
861
self._write_inventory(inv)
864
def revert(self, filenames, old_tree=None, backups=True):
865
from bzrlib.merge import merge_inner
867
old_tree = self.branch.basis_tree()
868
merge_inner(self.branch, old_tree,
869
self, ignore_zero=True,
870
backup_files=backups,
871
interesting_files=filenames)
872
if not len(filenames):
873
self.set_pending_merges([])
876
def set_inventory(self, new_inventory_list):
877
from bzrlib.inventory import (Inventory,
882
inv = Inventory(self.get_root_id())
883
for path, file_id, parent, kind in new_inventory_list:
884
name = os.path.basename(path)
887
# fixme, there should be a factory function inv,add_??
888
if kind == 'directory':
889
inv.add(InventoryDirectory(file_id, name, parent))
891
inv.add(InventoryFile(file_id, name, parent))
892
elif kind == 'symlink':
893
inv.add(InventoryLink(file_id, name, parent))
895
raise BzrError("unknown kind %r" % kind)
896
self._write_inventory(inv)
899
def set_root_id(self, file_id):
900
"""Set the root id for this tree."""
901
inv = self.read_working_inventory()
902
orig_root_id = inv.root.file_id
903
del inv._byid[inv.root.file_id]
904
inv.root.file_id = file_id
905
inv._byid[inv.root.file_id] = inv.root
908
if entry.parent_id in (None, orig_root_id):
909
entry.parent_id = inv.root.file_id
910
self._write_inventory(inv)
913
"""See Branch.unlock.
915
WorkingTree locking just uses the Branch locking facilities.
916
This is current because all working trees have an embedded branch
917
within them. IF in the future, we were to make branch data shareable
918
between multiple working trees, i.e. via shared storage, then we
919
would probably want to lock both the local tree, and the branch.
921
# FIXME: We want to write out the hashcache only when the last lock on
922
# this working copy is released. Peeking at the lock count is a bit
923
# of a nasty hack; probably it's better to have a transaction object,
924
# which can do some finalization when it's either successfully or
925
# unsuccessfully completed. (Denys's original patch did that.)
926
if self._hashcache.needs_write and self.branch.control_files._lock_count==1:
927
self._hashcache.write()
928
return self.branch.unlock()
931
def _write_inventory(self, inv):
932
"""Write inventory as the current inventory."""
933
from cStringIO import StringIO
934
from bzrlib.atomicfile import AtomicFile
936
bzrlib.xml5.serializer_v5.write_inventory(inv, sio)
938
f = AtomicFile(self.branch.control_files.controlfilename('inventory'))
944
self._set_inventory(inv)
945
mutter('wrote working inventory')
948
CONFLICT_SUFFIXES = ('.THIS', '.BASE', '.OTHER')
949
def get_conflicted_stem(path):
950
for suffix in CONFLICT_SUFFIXES:
951
if path.endswith(suffix):
952
return path[:-len(suffix)]
added
removed
bzrlib/workingtree.py
