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
18
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
19
import traceback, socket, fnmatch, difflib, time
20
from binascii import hexlify
23
from inventory import Inventory
24
from trace import mutter, note
25
from tree import Tree, EmptyTree, RevisionTree
26
from inventory import InventoryEntry, Inventory
27
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, \
28
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
29
joinpath, sha_string, file_kind, local_time_offset, appendpath
30
from store import ImmutableStore
31
from revision import Revision
32
from errors import BzrError
33
from textui import show_status
35
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
36
## TODO: Maybe include checks for common corruption of newlines, etc?
40
def find_branch(f, **args):
41
if f and (f.startswith('http://') or f.startswith('https://')):
43
return remotebranch.RemoteBranch(f, **args)
45
return Branch(f, **args)
49
def with_writelock(method):
50
"""Method decorator for functions run with the branch locked."""
52
# called with self set to the branch
55
return method(self, *a, **k)
61
def with_readlock(method):
65
return method(self, *a, **k)
71
def _relpath(base, path):
72
"""Return path relative to base, or raise exception.
74
The path may be either an absolute path or a path relative to the
75
current working directory.
77
Lifted out of Branch.relpath for ease of testing.
79
os.path.commonprefix (python2.4) has a bad bug that it works just
80
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
81
avoids that problem."""
82
rp = os.path.abspath(path)
86
while len(head) >= len(base):
89
head, tail = os.path.split(head)
93
from errors import NotBranchError
94
raise NotBranchError("path %r is not within branch %r" % (rp, base))
99
def find_branch_root(f=None):
100
"""Find the branch root enclosing f, or pwd.
102
f may be a filename or a URL.
104
It is not necessary that f exists.
106
Basically we keep looking up until we find the control directory or
107
run into the root."""
110
elif hasattr(os.path, 'realpath'):
111
f = os.path.realpath(f)
113
f = os.path.abspath(f)
114
if not os.path.exists(f):
115
raise BzrError('%r does not exist' % f)
121
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
123
head, tail = os.path.split(f)
125
# reached the root, whatever that may be
126
raise BzrError('%r is not in a branch' % orig_f)
131
######################################################################
134
class Branch(object):
135
"""Branch holding a history of revisions.
138
Base directory of the branch.
144
If _lock_mode is true, a positive count of the number of times the
148
Open file used for locking.
154
def __init__(self, base, init=False, find_root=True):
155
"""Create new branch object at a particular location.
157
base -- Base directory for the branch.
159
init -- If True, create new control files in a previously
160
unversioned directory. If False, the branch must already
163
find_root -- If true and init is false, find the root of the
164
existing branch containing base.
166
In the test suite, creation of new trees is tested using the
167
`ScratchBranch` class.
170
self.base = os.path.realpath(base)
173
self.base = find_branch_root(base)
175
self.base = os.path.realpath(base)
176
if not isdir(self.controlfilename('.')):
177
from errors import NotBranchError
178
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
179
['use "bzr init" to initialize a new working tree',
180
'current bzr can only operate from top-of-tree'])
182
self._lockfile = self.controlfile('branch-lock', 'wb')
184
self.text_store = ImmutableStore(self.controlfilename('text-store'))
185
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
186
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
190
return '%s(%r)' % (self.__class__.__name__, self.base)
198
from warnings import warn
199
warn("branch %r was not explicitly unlocked" % self)
203
def lock(self, mode):
205
if mode == 'w' and cur_lm == 'r':
206
raise BzrError("can't upgrade to a write lock")
208
assert self._lock_count >= 1
209
self._lock_count += 1
211
from bzrlib.lock import lock, LOCK_SH, LOCK_EX
217
raise ValueError('invalid lock mode %r' % mode)
219
lock(self._lockfile, m)
220
self._lock_mode = mode
225
if not self._lock_mode:
226
raise BzrError('branch %r is not locked' % (self))
228
if self._lock_count > 1:
229
self._lock_count -= 1
231
assert self._lock_count == 1
232
from bzrlib.lock import unlock
233
unlock(self._lockfile)
234
self._lock_mode = self._lock_count = None
237
def abspath(self, name):
238
"""Return absolute filename for something in the branch"""
239
return os.path.join(self.base, name)
242
def relpath(self, path):
243
"""Return path relative to this branch of something inside it.
245
Raises an error if path is not in this branch."""
246
return _relpath(self.base, path)
249
def controlfilename(self, file_or_path):
250
"""Return location relative to branch."""
251
if isinstance(file_or_path, types.StringTypes):
252
file_or_path = [file_or_path]
253
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
256
def controlfile(self, file_or_path, mode='r'):
257
"""Open a control file for this branch.
259
There are two classes of file in the control directory: text
260
and binary. binary files are untranslated byte streams. Text
261
control files are stored with Unix newlines and in UTF-8, even
262
if the platform or locale defaults are different.
264
Controlfiles should almost never be opened in write mode but
265
rather should be atomically copied and replaced using atomicfile.
268
fn = self.controlfilename(file_or_path)
270
if mode == 'rb' or mode == 'wb':
271
return file(fn, mode)
272
elif mode == 'r' or mode == 'w':
273
# open in binary mode anyhow so there's no newline translation;
274
# codecs uses line buffering by default; don't want that.
276
return codecs.open(fn, mode + 'b', 'utf-8',
279
raise BzrError("invalid controlfile mode %r" % mode)
283
def _make_control(self):
284
os.mkdir(self.controlfilename([]))
285
self.controlfile('README', 'w').write(
286
"This is a Bazaar-NG control directory.\n"
287
"Do not change any files in this directory.")
288
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
289
for d in ('text-store', 'inventory-store', 'revision-store'):
290
os.mkdir(self.controlfilename(d))
291
for f in ('revision-history', 'merged-patches',
292
'pending-merged-patches', 'branch-name',
294
self.controlfile(f, 'w').write('')
295
mutter('created control directory in ' + self.base)
296
Inventory().write_xml(self.controlfile('inventory','w'))
299
def _check_format(self):
300
"""Check this branch format is supported.
302
The current tool only supports the current unstable format.
304
In the future, we might need different in-memory Branch
305
classes to support downlevel branches. But not yet.
307
# This ignores newlines so that we can open branches created
308
# on Windows from Linux and so on. I think it might be better
309
# to always make all internal files in unix format.
310
fmt = self.controlfile('branch-format', 'r').read()
311
fmt.replace('\r\n', '')
312
if fmt != BZR_BRANCH_FORMAT:
313
raise BzrError('sorry, branch format %r not supported' % fmt,
314
['use a different bzr version',
315
'or remove the .bzr directory and "bzr init" again'])
320
def read_working_inventory(self):
321
"""Read the working inventory."""
323
# ElementTree does its own conversion from UTF-8, so open in
325
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
326
mutter("loaded inventory of %d items in %f"
327
% (len(inv), time.time() - before))
331
def _write_inventory(self, inv):
332
"""Update the working inventory.
334
That is to say, the inventory describing changes underway, that
335
will be committed to the next revision.
337
## TODO: factor out to atomicfile? is rename safe on windows?
338
## TODO: Maybe some kind of clean/dirty marker on inventory?
339
tmpfname = self.controlfilename('inventory.tmp')
340
tmpf = file(tmpfname, 'wb')
343
inv_fname = self.controlfilename('inventory')
344
if sys.platform == 'win32':
346
os.rename(tmpfname, inv_fname)
347
mutter('wrote working inventory')
350
inventory = property(read_working_inventory, _write_inventory, None,
351
"""Inventory for the working copy.""")
355
def add(self, files, verbose=False, ids=None):
356
"""Make files versioned.
358
Note that the command line normally calls smart_add instead.
360
This puts the files in the Added state, so that they will be
361
recorded by the next commit.
364
List of paths to add, relative to the base of the tree.
367
If set, use these instead of automatically generated ids.
368
Must be the same length as the list of files, but may
369
contain None for ids that are to be autogenerated.
371
TODO: Perhaps have an option to add the ids even if the files do
374
TODO: Perhaps return the ids of the files? But then again it
375
is easy to retrieve them if they're needed.
377
TODO: Adding a directory should optionally recurse down and
378
add all non-ignored children. Perhaps do that in a
381
# TODO: Re-adding a file that is removed in the working copy
382
# should probably put it back with the previous ID.
383
if isinstance(files, types.StringTypes):
384
assert(ids is None or isinstance(ids, types.StringTypes))
390
ids = [None] * len(files)
392
assert(len(ids) == len(files))
394
inv = self.read_working_inventory()
395
for f,file_id in zip(files, ids):
396
if is_control_file(f):
397
raise BzrError("cannot add control file %s" % quotefn(f))
402
raise BzrError("cannot add top-level %r" % f)
404
fullpath = os.path.normpath(self.abspath(f))
407
kind = file_kind(fullpath)
409
# maybe something better?
410
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
412
if kind != 'file' and kind != 'directory':
413
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
416
file_id = gen_file_id(f)
417
inv.add_path(f, kind=kind, file_id=file_id)
420
show_status('A', kind, quotefn(f))
422
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
424
self._write_inventory(inv)
427
def print_file(self, file, revno):
428
"""Print `file` to stdout."""
429
tree = self.revision_tree(self.lookup_revision(revno))
430
# use inventory as it was in that revision
431
file_id = tree.inventory.path2id(file)
433
raise BzrError("%r is not present in revision %d" % (file, revno))
434
tree.print_file(file_id)
438
def remove(self, files, verbose=False):
439
"""Mark nominated files for removal from the inventory.
441
This does not remove their text. This does not run on
443
TODO: Refuse to remove modified files unless --force is given?
445
TODO: Do something useful with directories.
447
TODO: Should this remove the text or not? Tough call; not
448
removing may be useful and the user can just use use rm, and
449
is the opposite of add. Removing it is consistent with most
450
other tools. Maybe an option.
452
## TODO: Normalize names
453
## TODO: Remove nested loops; better scalability
454
if isinstance(files, types.StringTypes):
457
tree = self.working_tree()
460
# do this before any modifications
464
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
465
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
467
# having remove it, it must be either ignored or unknown
468
if tree.is_ignored(f):
472
show_status(new_status, inv[fid].kind, quotefn(f))
475
self._write_inventory(inv)
478
def set_inventory(self, new_inventory_list):
480
for path, file_id, parent, kind in new_inventory_list:
481
name = os.path.basename(path)
484
inv.add(InventoryEntry(file_id, name, kind, parent))
485
self._write_inventory(inv)
489
"""Return all unknown files.
491
These are files in the working directory that are not versioned or
492
control files or ignored.
494
>>> b = ScratchBranch(files=['foo', 'foo~'])
495
>>> list(b.unknowns())
498
>>> list(b.unknowns())
501
>>> list(b.unknowns())
504
return self.working_tree().unknowns()
507
def append_revision(self, revision_id):
508
mutter("add {%s} to revision-history" % revision_id)
509
rev_history = self.revision_history()
511
tmprhname = self.controlfilename('revision-history.tmp')
512
rhname = self.controlfilename('revision-history')
514
f = file(tmprhname, 'wt')
515
rev_history.append(revision_id)
516
f.write('\n'.join(rev_history))
520
if sys.platform == 'win32':
522
os.rename(tmprhname, rhname)
526
def get_revision(self, revision_id):
527
"""Return the Revision object for a named revision"""
528
r = Revision.read_xml(self.revision_store[revision_id])
529
assert r.revision_id == revision_id
533
def get_inventory(self, inventory_id):
534
"""Get Inventory object by hash.
536
TODO: Perhaps for this and similar methods, take a revision
537
parameter which can be either an integer revno or a
539
i = Inventory.read_xml(self.inventory_store[inventory_id])
543
def get_revision_inventory(self, revision_id):
544
"""Return inventory of a past revision."""
545
if revision_id == None:
548
return self.get_inventory(self.get_revision(revision_id).inventory_id)
552
def revision_history(self):
553
"""Return sequence of revision hashes on to this branch.
555
>>> ScratchBranch().revision_history()
558
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
561
def enum_history(self, direction):
562
"""Return (revno, revision_id) for history of branch.
565
'forward' is from earliest to latest
566
'reverse' is from latest to earliest
568
rh = self.revision_history()
569
if direction == 'forward':
574
elif direction == 'reverse':
580
raise ValueError('invalid history direction', direction)
584
"""Return current revision number for this branch.
586
That is equivalent to the number of revisions committed to
589
return len(self.revision_history())
592
def last_patch(self):
593
"""Return last patch hash, or None if no history.
595
ph = self.revision_history()
602
def commit(self, *args, **kw):
604
from bzrlib.commit import commit
605
commit(self, *args, **kw)
608
def lookup_revision(self, revno):
609
"""Return revision hash for revision number."""
614
# list is 0-based; revisions are 1-based
615
return self.revision_history()[revno-1]
617
raise BzrError("no such revision %s" % revno)
620
def revision_tree(self, revision_id):
621
"""Return Tree for a revision on this branch.
623
`revision_id` may be None for the null revision, in which case
624
an `EmptyTree` is returned."""
625
# TODO: refactor this to use an existing revision object
626
# so we don't need to read it in twice.
627
if revision_id == None:
630
inv = self.get_revision_inventory(revision_id)
631
return RevisionTree(self.text_store, inv)
634
def working_tree(self):
635
"""Return a `Tree` for the working copy."""
636
from workingtree import WorkingTree
637
return WorkingTree(self.base, self.read_working_inventory())
640
def basis_tree(self):
641
"""Return `Tree` object for last revision.
643
If there are no revisions yet, return an `EmptyTree`.
645
r = self.last_patch()
649
return RevisionTree(self.text_store, self.get_revision_inventory(r))
654
def rename_one(self, from_rel, to_rel):
657
This can change the directory or the filename or both.
659
tree = self.working_tree()
661
if not tree.has_filename(from_rel):
662
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
663
if tree.has_filename(to_rel):
664
raise BzrError("can't rename: new working file %r already exists" % to_rel)
666
file_id = inv.path2id(from_rel)
668
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
670
if inv.path2id(to_rel):
671
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
673
to_dir, to_tail = os.path.split(to_rel)
674
to_dir_id = inv.path2id(to_dir)
675
if to_dir_id == None and to_dir != '':
676
raise BzrError("can't determine destination directory id for %r" % to_dir)
678
mutter("rename_one:")
679
mutter(" file_id {%s}" % file_id)
680
mutter(" from_rel %r" % from_rel)
681
mutter(" to_rel %r" % to_rel)
682
mutter(" to_dir %r" % to_dir)
683
mutter(" to_dir_id {%s}" % to_dir_id)
685
inv.rename(file_id, to_dir_id, to_tail)
687
print "%s => %s" % (from_rel, to_rel)
689
from_abs = self.abspath(from_rel)
690
to_abs = self.abspath(to_rel)
692
os.rename(from_abs, to_abs)
694
raise BzrError("failed to rename %r to %r: %s"
695
% (from_abs, to_abs, e[1]),
696
["rename rolled back"])
698
self._write_inventory(inv)
703
def move(self, from_paths, to_name):
706
to_name must exist as a versioned directory.
708
If to_name exists and is a directory, the files are moved into
709
it, keeping their old names. If it is a directory,
711
Note that to_name is only the last component of the new name;
712
this doesn't change the directory.
714
## TODO: Option to move IDs only
715
assert not isinstance(from_paths, basestring)
716
tree = self.working_tree()
718
to_abs = self.abspath(to_name)
719
if not isdir(to_abs):
720
raise BzrError("destination %r is not a directory" % to_abs)
721
if not tree.has_filename(to_name):
722
raise BzrError("destination %r not in working directory" % to_abs)
723
to_dir_id = inv.path2id(to_name)
724
if to_dir_id == None and to_name != '':
725
raise BzrError("destination %r is not a versioned directory" % to_name)
726
to_dir_ie = inv[to_dir_id]
727
if to_dir_ie.kind not in ('directory', 'root_directory'):
728
raise BzrError("destination %r is not a directory" % to_abs)
730
to_idpath = inv.get_idpath(to_dir_id)
733
if not tree.has_filename(f):
734
raise BzrError("%r does not exist in working tree" % f)
735
f_id = inv.path2id(f)
737
raise BzrError("%r is not versioned" % f)
738
name_tail = splitpath(f)[-1]
739
dest_path = appendpath(to_name, name_tail)
740
if tree.has_filename(dest_path):
741
raise BzrError("destination %r already exists" % dest_path)
742
if f_id in to_idpath:
743
raise BzrError("can't move %r to a subdirectory of itself" % f)
745
# OK, so there's a race here, it's possible that someone will
746
# create a file in this interval and then the rename might be
747
# left half-done. But we should have caught most problems.
750
name_tail = splitpath(f)[-1]
751
dest_path = appendpath(to_name, name_tail)
752
print "%s => %s" % (f, dest_path)
753
inv.rename(inv.path2id(f), to_dir_id, name_tail)
755
os.rename(self.abspath(f), self.abspath(dest_path))
757
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
758
["rename rolled back"])
760
self._write_inventory(inv)
765
class ScratchBranch(Branch):
766
"""Special test class: a branch that cleans up after itself.
768
>>> b = ScratchBranch()
776
def __init__(self, files=[], dirs=[]):
777
"""Make a test branch.
779
This creates a temporary directory and runs init-tree in it.
781
If any files are listed, they are created in the working copy.
783
Branch.__init__(self, tempfile.mkdtemp(), init=True)
785
os.mkdir(self.abspath(d))
788
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
795
"""Destroy the test branch, removing the scratch directory."""
797
mutter("delete ScratchBranch %s" % self.base)
798
shutil.rmtree(self.base)
800
# Work around for shutil.rmtree failing on Windows when
801
# readonly files are encountered
802
mutter("hit exception in destroying ScratchBranch: %s" % e)
803
for root, dirs, files in os.walk(self.base, topdown=False):
805
os.chmod(os.path.join(root, name), 0700)
806
shutil.rmtree(self.base)
811
######################################################################
815
def is_control_file(filename):
816
## FIXME: better check
817
filename = os.path.normpath(filename)
818
while filename != '':
819
head, tail = os.path.split(filename)
820
## mutter('check %r for control file' % ((head, tail), ))
821
if tail == bzrlib.BZRDIR:
830
def gen_file_id(name):
831
"""Return new file id.
833
This should probably generate proper UUIDs, but for the moment we
834
cope with just randomness because running uuidgen every time is
839
idx = name.rfind('/')
841
name = name[idx+1 : ]
842
idx = name.rfind('\\')
844
name = name[idx+1 : ]
846
# make it not a hidden file
847
name = name.lstrip('.')
849
# remove any wierd characters; we don't escape them but rather
851
name = re.sub(r'[^\w.]', '', name)
853
s = hexlify(rand_bytes(8))
854
return '-'.join((name, compact_date(time.time()), s))