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
37
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
38
## TODO: Maybe include checks for common corruption of newlines, etc?
42
def find_branch(f, **args):
43
if f and (f.startswith('http://') or f.startswith('https://')):
45
return remotebranch.RemoteBranch(f, **args)
47
return Branch(f, **args)
50
def find_branch_root(f=None):
51
"""Find the branch root enclosing f, or pwd.
53
f may be a filename or a URL.
55
It is not necessary that f exists.
57
Basically we keep looking up until we find the control directory or
61
elif hasattr(os.path, 'realpath'):
62
f = os.path.realpath(f)
64
f = os.path.abspath(f)
65
if not os.path.exists(f):
66
raise BzrError('%r does not exist' % f)
72
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
74
head, tail = os.path.split(f)
76
# reached the root, whatever that may be
77
raise BzrError('%r is not in a branch' % orig_f)
82
######################################################################
86
"""Branch holding a history of revisions.
89
Base directory of the branch.
94
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
95
"""Create new branch object at a particular location.
97
base -- Base directory for the branch.
99
init -- If True, create new control files in a previously
100
unversioned directory. If False, the branch must already
103
find_root -- If true and init is false, find the root of the
104
existing branch containing base.
106
In the test suite, creation of new trees is tested using the
107
`ScratchBranch` class.
110
self.base = os.path.realpath(base)
113
self.base = find_branch_root(base)
115
self.base = os.path.realpath(base)
116
if not isdir(self.controlfilename('.')):
117
from errors import NotBranchError
118
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
119
['use "bzr init" to initialize a new working tree',
120
'current bzr can only operate from top-of-tree'])
122
self._lockfile = None
125
self.text_store = ImmutableStore(self.controlfilename('text-store'))
126
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
127
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
131
return '%s(%r)' % (self.__class__.__name__, self.base)
138
def lock(self, mode='w'):
139
"""Lock the on-disk branch, excluding other processes."""
145
om = os.O_WRONLY | os.O_CREAT
150
raise BzrError("invalid locking mode %r" % mode)
153
lockfile = os.open(self.controlfilename('branch-lock'), om)
155
if e.errno == errno.ENOENT:
156
# might not exist on branches from <0.0.4
157
self.controlfile('branch-lock', 'w').close()
158
lockfile = os.open(self.controlfilename('branch-lock'), om)
162
fcntl.lockf(lockfile, lm)
164
fcntl.lockf(lockfile, fcntl.LOCK_UN)
166
self._lockmode = None
168
self._lockmode = mode
171
warnings.warning("please write a locking method for platform %r" % sys.platform)
173
self._lockmode = None
175
self._lockmode = mode
178
def _need_readlock(self):
179
if self._lockmode not in ['r', 'w']:
180
raise BzrError('need read lock on branch, only have %r' % self._lockmode)
182
def _need_writelock(self):
183
if self._lockmode not in ['w']:
184
raise BzrError('need write lock on branch, only have %r' % self._lockmode)
187
def abspath(self, name):
188
"""Return absolute filename for something in the branch"""
189
return os.path.join(self.base, name)
192
def relpath(self, path):
193
"""Return path relative to this branch of something inside it.
195
Raises an error if path is not in this branch."""
196
rp = os.path.realpath(path)
198
if not rp.startswith(self.base):
199
from errors import NotBranchError
200
raise NotBranchError("path %r is not within branch %r" % (rp, self.base))
201
rp = rp[len(self.base):]
202
rp = rp.lstrip(os.sep)
206
def controlfilename(self, file_or_path):
207
"""Return location relative to branch."""
208
if isinstance(file_or_path, types.StringTypes):
209
file_or_path = [file_or_path]
210
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
213
def controlfile(self, file_or_path, mode='r'):
214
"""Open a control file for this branch.
216
There are two classes of file in the control directory: text
217
and binary. binary files are untranslated byte streams. Text
218
control files are stored with Unix newlines and in UTF-8, even
219
if the platform or locale defaults are different.
221
Controlfiles should almost never be opened in write mode but
222
rather should be atomically copied and replaced using atomicfile.
225
fn = self.controlfilename(file_or_path)
227
if mode == 'rb' or mode == 'wb':
228
return file(fn, mode)
229
elif mode == 'r' or mode == 'w':
230
# open in binary mode anyhow so there's no newline translation;
231
# codecs uses line buffering by default; don't want that.
233
return codecs.open(fn, mode + 'b', 'utf-8',
236
raise BzrError("invalid controlfile mode %r" % mode)
240
def _make_control(self):
241
os.mkdir(self.controlfilename([]))
242
self.controlfile('README', 'w').write(
243
"This is a Bazaar-NG control directory.\n"
244
"Do not change any files in this directory.")
245
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
246
for d in ('text-store', 'inventory-store', 'revision-store'):
247
os.mkdir(self.controlfilename(d))
248
for f in ('revision-history', 'merged-patches',
249
'pending-merged-patches', 'branch-name',
251
self.controlfile(f, 'w').write('')
252
mutter('created control directory in ' + self.base)
253
Inventory().write_xml(self.controlfile('inventory','w'))
256
def _check_format(self):
257
"""Check this branch format is supported.
259
The current tool only supports the current unstable format.
261
In the future, we might need different in-memory Branch
262
classes to support downlevel branches. But not yet.
264
# This ignores newlines so that we can open branches created
265
# on Windows from Linux and so on. I think it might be better
266
# to always make all internal files in unix format.
267
fmt = self.controlfile('branch-format', 'r').read()
268
fmt.replace('\r\n', '')
269
if fmt != BZR_BRANCH_FORMAT:
270
raise BzrError('sorry, branch format %r not supported' % fmt,
271
['use a different bzr version',
272
'or remove the .bzr directory and "bzr init" again'])
275
def read_working_inventory(self):
276
"""Read the working inventory."""
277
self._need_readlock()
279
# ElementTree does its own conversion from UTF-8, so open in
281
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
282
mutter("loaded inventory of %d items in %f"
283
% (len(inv), time.time() - before))
287
def _write_inventory(self, inv):
288
"""Update the working inventory.
290
That is to say, the inventory describing changes underway, that
291
will be committed to the next revision.
293
self._need_writelock()
294
## TODO: factor out to atomicfile? is rename safe on windows?
295
## TODO: Maybe some kind of clean/dirty marker on inventory?
296
tmpfname = self.controlfilename('inventory.tmp')
297
tmpf = file(tmpfname, 'wb')
300
inv_fname = self.controlfilename('inventory')
301
if sys.platform == 'win32':
303
os.rename(tmpfname, inv_fname)
304
mutter('wrote working inventory')
307
inventory = property(read_working_inventory, _write_inventory, None,
308
"""Inventory for the working copy.""")
311
def add(self, files, verbose=False, ids=None):
312
"""Make files versioned.
314
Note that the command line normally calls smart_add instead.
316
This puts the files in the Added state, so that they will be
317
recorded by the next commit.
319
TODO: Perhaps have an option to add the ids even if the files do
322
TODO: Perhaps return the ids of the files? But then again it
323
is easy to retrieve them if they're needed.
325
TODO: Option to specify file id.
327
TODO: Adding a directory should optionally recurse down and
328
add all non-ignored children. Perhaps do that in a
331
self._need_writelock()
333
# TODO: Re-adding a file that is removed in the working copy
334
# should probably put it back with the previous ID.
335
if isinstance(files, types.StringTypes):
336
assert(ids is None or isinstance(ids, types.StringTypes))
342
ids = [None] * len(files)
344
assert(len(ids) == len(files))
346
inv = self.read_working_inventory()
347
for f,file_id in zip(files, ids):
348
if is_control_file(f):
349
raise BzrError("cannot add control file %s" % quotefn(f))
354
raise BzrError("cannot add top-level %r" % f)
356
fullpath = os.path.normpath(self.abspath(f))
359
kind = file_kind(fullpath)
361
# maybe something better?
362
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
364
if kind != 'file' and kind != 'directory':
365
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
368
file_id = gen_file_id(f)
369
inv.add_path(f, kind=kind, file_id=file_id)
372
show_status('A', kind, quotefn(f))
374
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
376
self._write_inventory(inv)
379
def print_file(self, file, revno):
380
"""Print `file` to stdout."""
381
self._need_readlock()
382
tree = self.revision_tree(self.lookup_revision(revno))
383
# use inventory as it was in that revision
384
file_id = tree.inventory.path2id(file)
386
raise BzrError("%r is not present in revision %d" % (file, revno))
387
tree.print_file(file_id)
390
def remove(self, files, verbose=False):
391
"""Mark nominated files for removal from the inventory.
393
This does not remove their text. This does not run on
395
TODO: Refuse to remove modified files unless --force is given?
397
TODO: Do something useful with directories.
399
TODO: Should this remove the text or not? Tough call; not
400
removing may be useful and the user can just use use rm, and
401
is the opposite of add. Removing it is consistent with most
402
other tools. Maybe an option.
404
## TODO: Normalize names
405
## TODO: Remove nested loops; better scalability
406
self._need_writelock()
408
if isinstance(files, types.StringTypes):
411
tree = self.working_tree()
414
# do this before any modifications
418
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
419
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
421
# having remove it, it must be either ignored or unknown
422
if tree.is_ignored(f):
426
show_status(new_status, inv[fid].kind, quotefn(f))
429
self._write_inventory(inv)
431
def set_inventory(self, new_inventory_list):
433
for path, file_id, parent, kind in new_inventory_list:
434
name = os.path.basename(path)
437
inv.add(InventoryEntry(file_id, name, kind, parent))
438
self._write_inventory(inv)
442
"""Return all unknown files.
444
These are files in the working directory that are not versioned or
445
control files or ignored.
447
>>> b = ScratchBranch(files=['foo', 'foo~'])
448
>>> list(b.unknowns())
451
>>> list(b.unknowns())
454
>>> list(b.unknowns())
457
return self.working_tree().unknowns()
460
def append_revision(self, revision_id):
461
mutter("add {%s} to revision-history" % revision_id)
462
rev_history = self.revision_history()
464
tmprhname = self.controlfilename('revision-history.tmp')
465
rhname = self.controlfilename('revision-history')
467
f = file(tmprhname, 'wt')
468
rev_history.append(revision_id)
469
f.write('\n'.join(rev_history))
473
if sys.platform == 'win32':
475
os.rename(tmprhname, rhname)
479
def get_revision(self, revision_id):
480
"""Return the Revision object for a named revision"""
481
self._need_readlock()
482
r = Revision.read_xml(self.revision_store[revision_id])
483
assert r.revision_id == revision_id
487
def get_inventory(self, inventory_id):
488
"""Get Inventory object by hash.
490
TODO: Perhaps for this and similar methods, take a revision
491
parameter which can be either an integer revno or a
493
self._need_readlock()
494
i = Inventory.read_xml(self.inventory_store[inventory_id])
498
def get_revision_inventory(self, revision_id):
499
"""Return inventory of a past revision."""
500
self._need_readlock()
501
if revision_id == None:
504
return self.get_inventory(self.get_revision(revision_id).inventory_id)
507
def revision_history(self):
508
"""Return sequence of revision hashes on to this branch.
510
>>> ScratchBranch().revision_history()
513
self._need_readlock()
514
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
517
def enum_history(self, direction):
518
"""Return (revno, revision_id) for history of branch.
521
'forward' is from earliest to latest
522
'reverse' is from latest to earliest
524
rh = self.revision_history()
525
if direction == 'forward':
530
elif direction == 'reverse':
536
raise ValueError('invalid history direction', direction)
540
"""Return current revision number for this branch.
542
That is equivalent to the number of revisions committed to
545
return len(self.revision_history())
548
def last_patch(self):
549
"""Return last patch hash, or None if no history.
551
ph = self.revision_history()
558
def commit(self, *args, **kw):
560
from bzrlib.commit import commit
561
commit(self, *args, **kw)
564
def lookup_revision(self, revno):
565
"""Return revision hash for revision number."""
570
# list is 0-based; revisions are 1-based
571
return self.revision_history()[revno-1]
573
raise BzrError("no such revision %s" % revno)
576
def revision_tree(self, revision_id):
577
"""Return Tree for a revision on this branch.
579
`revision_id` may be None for the null revision, in which case
580
an `EmptyTree` is returned."""
581
# TODO: refactor this to use an existing revision object
582
# so we don't need to read it in twice.
583
self._need_readlock()
584
if revision_id == None:
587
inv = self.get_revision_inventory(revision_id)
588
return RevisionTree(self.text_store, inv)
591
def working_tree(self):
592
"""Return a `Tree` for the working copy."""
593
from workingtree import WorkingTree
594
return WorkingTree(self.base, self.read_working_inventory())
597
def basis_tree(self):
598
"""Return `Tree` object for last revision.
600
If there are no revisions yet, return an `EmptyTree`.
602
r = self.last_patch()
606
return RevisionTree(self.text_store, self.get_revision_inventory(r))
610
def rename_one(self, from_rel, to_rel):
613
This can change the directory or the filename or both.
615
self._need_writelock()
616
tree = self.working_tree()
618
if not tree.has_filename(from_rel):
619
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
620
if tree.has_filename(to_rel):
621
raise BzrError("can't rename: new working file %r already exists" % to_rel)
623
file_id = inv.path2id(from_rel)
625
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
627
if inv.path2id(to_rel):
628
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
630
to_dir, to_tail = os.path.split(to_rel)
631
to_dir_id = inv.path2id(to_dir)
632
if to_dir_id == None and to_dir != '':
633
raise BzrError("can't determine destination directory id for %r" % to_dir)
635
mutter("rename_one:")
636
mutter(" file_id {%s}" % file_id)
637
mutter(" from_rel %r" % from_rel)
638
mutter(" to_rel %r" % to_rel)
639
mutter(" to_dir %r" % to_dir)
640
mutter(" to_dir_id {%s}" % to_dir_id)
642
inv.rename(file_id, to_dir_id, to_tail)
644
print "%s => %s" % (from_rel, to_rel)
646
from_abs = self.abspath(from_rel)
647
to_abs = self.abspath(to_rel)
649
os.rename(from_abs, to_abs)
651
raise BzrError("failed to rename %r to %r: %s"
652
% (from_abs, to_abs, e[1]),
653
["rename rolled back"])
655
self._write_inventory(inv)
659
def move(self, from_paths, to_name):
662
to_name must exist as a versioned directory.
664
If to_name exists and is a directory, the files are moved into
665
it, keeping their old names. If it is a directory,
667
Note that to_name is only the last component of the new name;
668
this doesn't change the directory.
670
self._need_writelock()
671
## TODO: Option to move IDs only
672
assert not isinstance(from_paths, basestring)
673
tree = self.working_tree()
675
to_abs = self.abspath(to_name)
676
if not isdir(to_abs):
677
raise BzrError("destination %r is not a directory" % to_abs)
678
if not tree.has_filename(to_name):
679
raise BzrError("destination %r not in working directory" % to_abs)
680
to_dir_id = inv.path2id(to_name)
681
if to_dir_id == None and to_name != '':
682
raise BzrError("destination %r is not a versioned directory" % to_name)
683
to_dir_ie = inv[to_dir_id]
684
if to_dir_ie.kind not in ('directory', 'root_directory'):
685
raise BzrError("destination %r is not a directory" % to_abs)
687
to_idpath = inv.get_idpath(to_dir_id)
690
if not tree.has_filename(f):
691
raise BzrError("%r does not exist in working tree" % f)
692
f_id = inv.path2id(f)
694
raise BzrError("%r is not versioned" % f)
695
name_tail = splitpath(f)[-1]
696
dest_path = appendpath(to_name, name_tail)
697
if tree.has_filename(dest_path):
698
raise BzrError("destination %r already exists" % dest_path)
699
if f_id in to_idpath:
700
raise BzrError("can't move %r to a subdirectory of itself" % f)
702
# OK, so there's a race here, it's possible that someone will
703
# create a file in this interval and then the rename might be
704
# left half-done. But we should have caught most problems.
707
name_tail = splitpath(f)[-1]
708
dest_path = appendpath(to_name, name_tail)
709
print "%s => %s" % (f, dest_path)
710
inv.rename(inv.path2id(f), to_dir_id, name_tail)
712
os.rename(self.abspath(f), self.abspath(dest_path))
714
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
715
["rename rolled back"])
717
self._write_inventory(inv)
722
class ScratchBranch(Branch):
723
"""Special test class: a branch that cleans up after itself.
725
>>> b = ScratchBranch()
733
def __init__(self, files=[], dirs=[]):
734
"""Make a test branch.
736
This creates a temporary directory and runs init-tree in it.
738
If any files are listed, they are created in the working copy.
740
Branch.__init__(self, tempfile.mkdtemp(), init=True)
742
os.mkdir(self.abspath(d))
745
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
752
"""Destroy the test branch, removing the scratch directory."""
754
mutter("delete ScratchBranch %s" % self.base)
755
shutil.rmtree(self.base)
757
# Work around for shutil.rmtree failing on Windows when
758
# readonly files are encountered
759
mutter("hit exception in destroying ScratchBranch: %s" % e)
760
for root, dirs, files in os.walk(self.base, topdown=False):
762
os.chmod(os.path.join(root, name), 0700)
763
shutil.rmtree(self.base)
768
######################################################################
772
def is_control_file(filename):
773
## FIXME: better check
774
filename = os.path.normpath(filename)
775
while filename != '':
776
head, tail = os.path.split(filename)
777
## mutter('check %r for control file' % ((head, tail), ))
778
if tail == bzrlib.BZRDIR:
787
def gen_file_id(name):
788
"""Return new file id.
790
This should probably generate proper UUIDs, but for the moment we
791
cope with just randomness because running uuidgen every time is
796
idx = name.rfind('/')
798
name = name[idx+1 : ]
799
idx = name.rfind('\\')
801
name = name[idx+1 : ]
803
# make it not a hidden file
804
name = name.lstrip('.')
806
# remove any wierd characters; we don't escape them but rather
808
name = re.sub(r'[^\w.]', '', name)
810
s = hexlify(rand_bytes(8))
811
return '-'.join((name, compact_date(time.time()), s))