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
20
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
21
import traceback, socket, fnmatch, difflib, time
22
from binascii import hexlify
25
from inventory import Inventory
26
from trace import mutter, note
27
from tree import Tree, EmptyTree, RevisionTree
28
from inventory import InventoryEntry, Inventory
29
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, \
30
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
31
joinpath, sha_string, file_kind, local_time_offset, appendpath
32
from store import ImmutableStore
33
from revision import Revision
34
from errors import bailout, BzrError
35
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.
93
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
94
"""Create new branch object at a particular location.
96
base -- Base directory for the branch.
98
init -- If True, create new control files in a previously
99
unversioned directory. If False, the branch must already
102
find_root -- If true and init is false, find the root of the
103
existing branch containing base.
105
In the test suite, creation of new trees is tested using the
106
`ScratchBranch` class.
109
self.base = os.path.realpath(base)
112
self.base = find_branch_root(base)
114
self.base = os.path.realpath(base)
115
if not isdir(self.controlfilename('.')):
116
bailout("not a bzr branch: %s" % quotefn(base),
117
['use "bzr init" to initialize a new working tree',
118
'current bzr can only operate from top-of-tree'])
122
self.text_store = ImmutableStore(self.controlfilename('text-store'))
123
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
124
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
128
return '%s(%r)' % (self.__class__.__name__, self.base)
135
def lock(self, mode='w'):
136
"""Lock the on-disk branch, excluding other processes."""
142
om = os.O_WRONLY | os.O_CREAT
147
raise BzrError("invalid locking mode %r" % mode)
150
lockfile = os.open(self.controlfilename('branch-lock'), om)
152
if e.errno == errno.ENOENT:
153
# might not exist on branches from <0.0.4
154
self.controlfile('branch-lock', 'w').close()
155
lockfile = os.open(self.controlfilename('branch-lock'), om)
159
fcntl.lockf(lockfile, lm)
161
fcntl.lockf(lockfile, fcntl.LOCK_UN)
163
self._lockmode = None
165
self._lockmode = mode
167
warning("please write a locking method for platform %r" % sys.platform)
169
self._lockmode = None
171
self._lockmode = mode
174
def _need_readlock(self):
175
if self._lockmode not in ['r', 'w']:
176
raise BzrError('need read lock on branch, only have %r' % self._lockmode)
178
def _need_writelock(self):
179
if self._lockmode not in ['w']:
180
raise BzrError('need write lock on branch, only have %r' % self._lockmode)
183
def abspath(self, name):
184
"""Return absolute filename for something in the branch"""
185
return os.path.join(self.base, name)
188
def relpath(self, path):
189
"""Return path relative to this branch of something inside it.
191
Raises an error if path is not in this branch."""
192
rp = os.path.realpath(path)
194
if not rp.startswith(self.base):
195
bailout("path %r is not within branch %r" % (rp, self.base))
196
rp = rp[len(self.base):]
197
rp = rp.lstrip(os.sep)
201
def controlfilename(self, file_or_path):
202
"""Return location relative to branch."""
203
if isinstance(file_or_path, types.StringTypes):
204
file_or_path = [file_or_path]
205
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
208
def controlfile(self, file_or_path, mode='r'):
209
"""Open a control file for this branch.
211
There are two classes of file in the control directory: text
212
and binary. binary files are untranslated byte streams. Text
213
control files are stored with Unix newlines and in UTF-8, even
214
if the platform or locale defaults are different.
216
Controlfiles should almost never be opened in write mode but
217
rather should be atomically copied and replaced using atomicfile.
220
fn = self.controlfilename(file_or_path)
222
if mode == 'rb' or mode == 'wb':
223
return file(fn, mode)
224
elif mode == 'r' or mode == 'w':
225
# open in binary mode anyhow so there's no newline translation;
226
# codecs uses line buffering by default; don't want that.
228
return codecs.open(fn, mode + 'b', 'utf-8',
231
raise BzrError("invalid controlfile mode %r" % mode)
235
def _make_control(self):
236
os.mkdir(self.controlfilename([]))
237
self.controlfile('README', 'w').write(
238
"This is a Bazaar-NG control directory.\n"
239
"Do not change any files in this directory.")
240
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
241
for d in ('text-store', 'inventory-store', 'revision-store'):
242
os.mkdir(self.controlfilename(d))
243
for f in ('revision-history', 'merged-patches',
244
'pending-merged-patches', 'branch-name',
246
self.controlfile(f, 'w').write('')
247
mutter('created control directory in ' + self.base)
248
Inventory().write_xml(self.controlfile('inventory','w'))
251
def _check_format(self):
252
"""Check this branch format is supported.
254
The current tool only supports the current unstable format.
256
In the future, we might need different in-memory Branch
257
classes to support downlevel branches. But not yet.
259
# This ignores newlines so that we can open branches created
260
# on Windows from Linux and so on. I think it might be better
261
# to always make all internal files in unix format.
262
fmt = self.controlfile('branch-format', 'r').read()
263
fmt.replace('\r\n', '')
264
if fmt != BZR_BRANCH_FORMAT:
265
bailout('sorry, branch format %r not supported' % fmt,
266
['use a different bzr version',
267
'or remove the .bzr directory and "bzr init" again'])
270
def read_working_inventory(self):
271
"""Read the working inventory."""
272
self._need_readlock()
274
# ElementTree does its own conversion from UTF-8, so open in
276
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
277
mutter("loaded inventory of %d items in %f"
278
% (len(inv), time.time() - before))
282
def _write_inventory(self, inv):
283
"""Update the working inventory.
285
That is to say, the inventory describing changes underway, that
286
will be committed to the next revision.
288
self._need_writelock()
289
## TODO: factor out to atomicfile? is rename safe on windows?
290
## TODO: Maybe some kind of clean/dirty marker on inventory?
291
tmpfname = self.controlfilename('inventory.tmp')
292
tmpf = file(tmpfname, 'wb')
295
inv_fname = self.controlfilename('inventory')
296
if sys.platform == 'win32':
298
os.rename(tmpfname, inv_fname)
299
mutter('wrote working inventory')
302
inventory = property(read_working_inventory, _write_inventory, None,
303
"""Inventory for the working copy.""")
306
def add(self, files, verbose=False, ids=None):
307
"""Make files versioned.
309
Note that the command line normally calls smart_add instead.
311
This puts the files in the Added state, so that they will be
312
recorded by the next commit.
314
TODO: Perhaps have an option to add the ids even if the files do
317
TODO: Perhaps return the ids of the files? But then again it
318
is easy to retrieve them if they're needed.
320
TODO: Option to specify file id.
322
TODO: Adding a directory should optionally recurse down and
323
add all non-ignored children. Perhaps do that in a
326
self._need_writelock()
328
# TODO: Re-adding a file that is removed in the working copy
329
# should probably put it back with the previous ID.
330
if isinstance(files, types.StringTypes):
331
assert(ids is None or isinstance(ids, types.StringTypes))
337
ids = [None] * len(files)
339
assert(len(ids) == len(files))
341
inv = self.read_working_inventory()
342
for f,file_id in zip(files, ids):
343
if is_control_file(f):
344
bailout("cannot add control file %s" % quotefn(f))
349
bailout("cannot add top-level %r" % f)
351
fullpath = os.path.normpath(self.abspath(f))
354
kind = file_kind(fullpath)
356
# maybe something better?
357
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
359
if kind != 'file' and kind != 'directory':
360
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
363
file_id = gen_file_id(f)
364
inv.add_path(f, kind=kind, file_id=file_id)
367
show_status('A', kind, quotefn(f))
369
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
371
self._write_inventory(inv)
374
def print_file(self, file, revno):
375
"""Print `file` to stdout."""
376
self._need_readlock()
377
tree = self.revision_tree(self.lookup_revision(revno))
378
# use inventory as it was in that revision
379
file_id = tree.inventory.path2id(file)
381
bailout("%r is not present in revision %d" % (file, revno))
382
tree.print_file(file_id)
385
def remove(self, files, verbose=False):
386
"""Mark nominated files for removal from the inventory.
388
This does not remove their text. This does not run on
390
TODO: Refuse to remove modified files unless --force is given?
392
TODO: Do something useful with directories.
394
TODO: Should this remove the text or not? Tough call; not
395
removing may be useful and the user can just use use rm, and
396
is the opposite of add. Removing it is consistent with most
397
other tools. Maybe an option.
399
## TODO: Normalize names
400
## TODO: Remove nested loops; better scalability
401
self._need_writelock()
403
if isinstance(files, types.StringTypes):
406
tree = self.working_tree()
409
# do this before any modifications
413
bailout("cannot remove unversioned file %s" % quotefn(f))
414
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
416
# having remove it, it must be either ignored or unknown
417
if tree.is_ignored(f):
421
show_status(new_status, inv[fid].kind, quotefn(f))
424
self._write_inventory(inv)
426
def set_inventory(self, new_inventory_list):
428
for path, file_id, parent, kind in new_inventory_list:
429
name = os.path.basename(path)
432
inv.add(InventoryEntry(file_id, name, kind, parent))
433
self._write_inventory(inv)
437
"""Return all unknown files.
439
These are files in the working directory that are not versioned or
440
control files or ignored.
442
>>> b = ScratchBranch(files=['foo', 'foo~'])
443
>>> list(b.unknowns())
446
>>> list(b.unknowns())
449
>>> list(b.unknowns())
452
return self.working_tree().unknowns()
455
def append_revision(self, revision_id):
456
mutter("add {%s} to revision-history" % revision_id)
457
rev_history = self.revision_history()
459
tmprhname = self.controlfilename('revision-history.tmp')
460
rhname = self.controlfilename('revision-history')
462
f = file(tmprhname, 'wt')
463
rev_history.append(revision_id)
464
f.write('\n'.join(rev_history))
468
if sys.platform == 'win32':
470
os.rename(tmprhname, rhname)
474
def get_revision(self, revision_id):
475
"""Return the Revision object for a named revision"""
476
self._need_readlock()
477
r = Revision.read_xml(self.revision_store[revision_id])
478
assert r.revision_id == revision_id
482
def get_inventory(self, inventory_id):
483
"""Get Inventory object by hash.
485
TODO: Perhaps for this and similar methods, take a revision
486
parameter which can be either an integer revno or a
488
self._need_readlock()
489
i = Inventory.read_xml(self.inventory_store[inventory_id])
493
def get_revision_inventory(self, revision_id):
494
"""Return inventory of a past revision."""
495
self._need_readlock()
496
if revision_id == None:
499
return self.get_inventory(self.get_revision(revision_id).inventory_id)
502
def revision_history(self):
503
"""Return sequence of revision hashes on to this branch.
505
>>> ScratchBranch().revision_history()
508
self._need_readlock()
509
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
512
def enum_history(self, direction):
513
"""Return (revno, revision_id) for history of branch.
516
'forward' is from earliest to latest
517
'reverse' is from latest to earliest
519
rh = self.revision_history()
520
if direction == 'forward':
525
elif direction == 'reverse':
531
raise ValueError('invalid history direction', direction)
535
"""Return current revision number for this branch.
537
That is equivalent to the number of revisions committed to
540
return len(self.revision_history())
543
def last_patch(self):
544
"""Return last patch hash, or None if no history.
546
ph = self.revision_history()
553
def commit(self, *args, **kw):
555
from bzrlib.commit import commit
556
commit(self, *args, **kw)
559
def lookup_revision(self, revno):
560
"""Return revision hash for revision number."""
565
# list is 0-based; revisions are 1-based
566
return self.revision_history()[revno-1]
568
raise BzrError("no such revision %s" % revno)
571
def revision_tree(self, revision_id):
572
"""Return Tree for a revision on this branch.
574
`revision_id` may be None for the null revision, in which case
575
an `EmptyTree` is returned."""
576
# TODO: refactor this to use an existing revision object
577
# so we don't need to read it in twice.
578
self._need_readlock()
579
if revision_id == None:
582
inv = self.get_revision_inventory(revision_id)
583
return RevisionTree(self.text_store, inv)
586
def working_tree(self):
587
"""Return a `Tree` for the working copy."""
588
from workingtree import WorkingTree
589
return WorkingTree(self.base, self.read_working_inventory())
592
def basis_tree(self):
593
"""Return `Tree` object for last revision.
595
If there are no revisions yet, return an `EmptyTree`.
597
r = self.last_patch()
601
return RevisionTree(self.text_store, self.get_revision_inventory(r))
605
def rename_one(self, from_rel, to_rel):
608
This can change the directory or the filename or both.
610
self._need_writelock()
611
tree = self.working_tree()
613
if not tree.has_filename(from_rel):
614
bailout("can't rename: old working file %r does not exist" % from_rel)
615
if tree.has_filename(to_rel):
616
bailout("can't rename: new working file %r already exists" % to_rel)
618
file_id = inv.path2id(from_rel)
620
bailout("can't rename: old name %r is not versioned" % from_rel)
622
if inv.path2id(to_rel):
623
bailout("can't rename: new name %r is already versioned" % to_rel)
625
to_dir, to_tail = os.path.split(to_rel)
626
to_dir_id = inv.path2id(to_dir)
627
if to_dir_id == None and to_dir != '':
628
bailout("can't determine destination directory id for %r" % to_dir)
630
mutter("rename_one:")
631
mutter(" file_id {%s}" % file_id)
632
mutter(" from_rel %r" % from_rel)
633
mutter(" to_rel %r" % to_rel)
634
mutter(" to_dir %r" % to_dir)
635
mutter(" to_dir_id {%s}" % to_dir_id)
637
inv.rename(file_id, to_dir_id, to_tail)
639
print "%s => %s" % (from_rel, to_rel)
641
from_abs = self.abspath(from_rel)
642
to_abs = self.abspath(to_rel)
644
os.rename(from_abs, to_abs)
646
bailout("failed to rename %r to %r: %s"
647
% (from_abs, to_abs, e[1]),
648
["rename rolled back"])
650
self._write_inventory(inv)
654
def move(self, from_paths, to_name):
657
to_name must exist as a versioned directory.
659
If to_name exists and is a directory, the files are moved into
660
it, keeping their old names. If it is a directory,
662
Note that to_name is only the last component of the new name;
663
this doesn't change the directory.
665
self._need_writelock()
666
## TODO: Option to move IDs only
667
assert not isinstance(from_paths, basestring)
668
tree = self.working_tree()
670
to_abs = self.abspath(to_name)
671
if not isdir(to_abs):
672
bailout("destination %r is not a directory" % to_abs)
673
if not tree.has_filename(to_name):
674
bailout("destination %r not in working directory" % to_abs)
675
to_dir_id = inv.path2id(to_name)
676
if to_dir_id == None and to_name != '':
677
bailout("destination %r is not a versioned directory" % to_name)
678
to_dir_ie = inv[to_dir_id]
679
if to_dir_ie.kind not in ('directory', 'root_directory'):
680
bailout("destination %r is not a directory" % to_abs)
682
to_idpath = Set(inv.get_idpath(to_dir_id))
685
if not tree.has_filename(f):
686
bailout("%r does not exist in working tree" % f)
687
f_id = inv.path2id(f)
689
bailout("%r is not versioned" % f)
690
name_tail = splitpath(f)[-1]
691
dest_path = appendpath(to_name, name_tail)
692
if tree.has_filename(dest_path):
693
bailout("destination %r already exists" % dest_path)
694
if f_id in to_idpath:
695
bailout("can't move %r to a subdirectory of itself" % f)
697
# OK, so there's a race here, it's possible that someone will
698
# create a file in this interval and then the rename might be
699
# left half-done. But we should have caught most problems.
702
name_tail = splitpath(f)[-1]
703
dest_path = appendpath(to_name, name_tail)
704
print "%s => %s" % (f, dest_path)
705
inv.rename(inv.path2id(f), to_dir_id, name_tail)
707
os.rename(self.abspath(f), self.abspath(dest_path))
709
bailout("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
710
["rename rolled back"])
712
self._write_inventory(inv)
717
class ScratchBranch(Branch):
718
"""Special test class: a branch that cleans up after itself.
720
>>> b = ScratchBranch()
728
def __init__(self, files=[], dirs=[]):
729
"""Make a test branch.
731
This creates a temporary directory and runs init-tree in it.
733
If any files are listed, they are created in the working copy.
735
Branch.__init__(self, tempfile.mkdtemp(), init=True)
737
os.mkdir(self.abspath(d))
740
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
747
"""Destroy the test branch, removing the scratch directory."""
749
mutter("delete ScratchBranch %s" % self.base)
750
shutil.rmtree(self.base)
752
# Work around for shutil.rmtree failing on Windows when
753
# readonly files are encountered
754
mutter("hit exception in destroying ScratchBranch: %s" % e)
755
for root, dirs, files in os.walk(self.base, topdown=False):
757
os.chmod(os.path.join(root, name), 0700)
758
shutil.rmtree(self.base)
763
######################################################################
767
def is_control_file(filename):
768
## FIXME: better check
769
filename = os.path.normpath(filename)
770
while filename != '':
771
head, tail = os.path.split(filename)
772
## mutter('check %r for control file' % ((head, tail), ))
773
if tail == bzrlib.BZRDIR:
782
def gen_file_id(name):
783
"""Return new file id.
785
This should probably generate proper UUIDs, but for the moment we
786
cope with just randomness because running uuidgen every time is
791
idx = name.rfind('/')
793
name = name[idx+1 : ]
794
idx = name.rfind('\\')
796
name = name[idx+1 : ]
798
# make it not a hidden file
799
name = name.lstrip('.')
801
# remove any wierd characters; we don't escape them but rather
803
name = re.sub(r'[^\w.]', '', name)
805
s = hexlify(rand_bytes(8))
806
return '-'.join((name, compact_date(time.time()), s))