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)
48
def find_branch_root(f=None):
49
"""Find the branch root enclosing f, or pwd.
51
f may be a filename or a URL.
53
It is not necessary that f exists.
55
Basically we keep looking up until we find the control directory or
59
elif hasattr(os.path, 'realpath'):
60
f = os.path.realpath(f)
62
f = os.path.abspath(f)
63
if not os.path.exists(f):
64
raise BzrError('%r does not exist' % f)
70
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
72
head, tail = os.path.split(f)
74
# reached the root, whatever that may be
75
raise BzrError('%r is not in a branch' % orig_f)
80
######################################################################
84
"""Branch holding a history of revisions.
87
Base directory of the branch.
90
None, or a duple with 'r' or 'w' for the first element and a positive
94
Open file used for locking.
99
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
100
"""Create new branch object at a particular location.
102
base -- Base directory for the branch.
104
init -- If True, create new control files in a previously
105
unversioned directory. If False, the branch must already
108
find_root -- If true and init is false, find the root of the
109
existing branch containing base.
111
In the test suite, creation of new trees is tested using the
112
`ScratchBranch` class.
115
self.base = os.path.realpath(base)
118
self.base = find_branch_root(base)
120
self.base = os.path.realpath(base)
121
if not isdir(self.controlfilename('.')):
122
from errors import NotBranchError
123
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
124
['use "bzr init" to initialize a new working tree',
125
'current bzr can only operate from top-of-tree'])
127
self._lockfile = self.controlfile('branch-lock', 'wb')
130
self.text_store = ImmutableStore(self.controlfilename('text-store'))
131
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
132
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
136
return '%s(%r)' % (self.__class__.__name__, self.base)
144
from warnings import warn
145
warn("branch %r was not explicitly unlocked" % self)
149
def lock(self, mode):
151
raise BzrError('branch %r is already locked: %r' % (self, self._lock_mode))
153
from bzrlib.lock import lock, LOCK_SH, LOCK_EX
159
raise ValueError('invalid lock mode %r' % mode)
161
lock(self._lockfile, m)
162
self._lock_mode = (mode, 1)
166
if not self._lock_mode:
167
raise BzrError('branch %r is not locked' % (self))
168
from bzrlib.lock import unlock
169
unlock(self._lockfile)
170
self._lock_mode = None
173
def _need_readlock(self):
174
if not self._lock_mode:
175
raise BzrError('need read lock on branch, only have %r' % self._lockmode)
178
def _need_writelock(self):
179
if (self._lock_mode == None) or (self._lock_mode[0] != '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
from errors import NotBranchError
196
raise NotBranchError("path %r is not within branch %r" % (rp, self.base))
197
rp = rp[len(self.base):]
198
rp = rp.lstrip(os.sep)
202
def controlfilename(self, file_or_path):
203
"""Return location relative to branch."""
204
if isinstance(file_or_path, types.StringTypes):
205
file_or_path = [file_or_path]
206
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
209
def controlfile(self, file_or_path, mode='r'):
210
"""Open a control file for this branch.
212
There are two classes of file in the control directory: text
213
and binary. binary files are untranslated byte streams. Text
214
control files are stored with Unix newlines and in UTF-8, even
215
if the platform or locale defaults are different.
217
Controlfiles should almost never be opened in write mode but
218
rather should be atomically copied and replaced using atomicfile.
221
fn = self.controlfilename(file_or_path)
223
if mode == 'rb' or mode == 'wb':
224
return file(fn, mode)
225
elif mode == 'r' or mode == 'w':
226
# open in binary mode anyhow so there's no newline translation;
227
# codecs uses line buffering by default; don't want that.
229
return codecs.open(fn, mode + 'b', 'utf-8',
232
raise BzrError("invalid controlfile mode %r" % mode)
236
def _make_control(self):
237
os.mkdir(self.controlfilename([]))
238
self.controlfile('README', 'w').write(
239
"This is a Bazaar-NG control directory.\n"
240
"Do not change any files in this directory.")
241
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
242
for d in ('text-store', 'inventory-store', 'revision-store'):
243
os.mkdir(self.controlfilename(d))
244
for f in ('revision-history', 'merged-patches',
245
'pending-merged-patches', 'branch-name',
247
self.controlfile(f, 'w').write('')
248
mutter('created control directory in ' + self.base)
249
Inventory().write_xml(self.controlfile('inventory','w'))
252
def _check_format(self):
253
"""Check this branch format is supported.
255
The current tool only supports the current unstable format.
257
In the future, we might need different in-memory Branch
258
classes to support downlevel branches. But not yet.
260
# This ignores newlines so that we can open branches created
261
# on Windows from Linux and so on. I think it might be better
262
# to always make all internal files in unix format.
263
fmt = self.controlfile('branch-format', 'r').read()
264
fmt.replace('\r\n', '')
265
if fmt != BZR_BRANCH_FORMAT:
266
raise BzrError('sorry, branch format %r not supported' % fmt,
267
['use a different bzr version',
268
'or remove the .bzr directory and "bzr init" again'])
271
def read_working_inventory(self):
272
"""Read the working inventory."""
273
self._need_readlock()
275
# ElementTree does its own conversion from UTF-8, so open in
277
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
278
mutter("loaded inventory of %d items in %f"
279
% (len(inv), time.time() - before))
283
def _write_inventory(self, inv):
284
"""Update the working inventory.
286
That is to say, the inventory describing changes underway, that
287
will be committed to the next revision.
289
self._need_writelock()
290
## TODO: factor out to atomicfile? is rename safe on windows?
291
## TODO: Maybe some kind of clean/dirty marker on inventory?
292
tmpfname = self.controlfilename('inventory.tmp')
293
tmpf = file(tmpfname, 'wb')
296
inv_fname = self.controlfilename('inventory')
297
if sys.platform == 'win32':
299
os.rename(tmpfname, inv_fname)
300
mutter('wrote working inventory')
303
inventory = property(read_working_inventory, _write_inventory, None,
304
"""Inventory for the working copy.""")
307
def add(self, files, verbose=False, ids=None):
308
"""Make files versioned.
310
Note that the command line normally calls smart_add instead.
312
This puts the files in the Added state, so that they will be
313
recorded by the next commit.
315
TODO: Perhaps have an option to add the ids even if the files do
318
TODO: Perhaps return the ids of the files? But then again it
319
is easy to retrieve them if they're needed.
321
TODO: Option to specify file id.
323
TODO: Adding a directory should optionally recurse down and
324
add all non-ignored children. Perhaps do that in a
327
self._need_writelock()
329
# TODO: Re-adding a file that is removed in the working copy
330
# should probably put it back with the previous ID.
331
if isinstance(files, types.StringTypes):
332
assert(ids is None or isinstance(ids, types.StringTypes))
338
ids = [None] * len(files)
340
assert(len(ids) == len(files))
342
inv = self.read_working_inventory()
343
for f,file_id in zip(files, ids):
344
if is_control_file(f):
345
raise BzrError("cannot add control file %s" % quotefn(f))
350
raise BzrError("cannot add top-level %r" % f)
352
fullpath = os.path.normpath(self.abspath(f))
355
kind = file_kind(fullpath)
357
# maybe something better?
358
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
360
if kind != 'file' and kind != 'directory':
361
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
364
file_id = gen_file_id(f)
365
inv.add_path(f, kind=kind, file_id=file_id)
368
show_status('A', kind, quotefn(f))
370
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
372
self._write_inventory(inv)
375
def print_file(self, file, revno):
376
"""Print `file` to stdout."""
377
self._need_readlock()
378
tree = self.revision_tree(self.lookup_revision(revno))
379
# use inventory as it was in that revision
380
file_id = tree.inventory.path2id(file)
382
raise BzrError("%r is not present in revision %d" % (file, revno))
383
tree.print_file(file_id)
386
def remove(self, files, verbose=False):
387
"""Mark nominated files for removal from the inventory.
389
This does not remove their text. This does not run on
391
TODO: Refuse to remove modified files unless --force is given?
393
TODO: Do something useful with directories.
395
TODO: Should this remove the text or not? Tough call; not
396
removing may be useful and the user can just use use rm, and
397
is the opposite of add. Removing it is consistent with most
398
other tools. Maybe an option.
400
## TODO: Normalize names
401
## TODO: Remove nested loops; better scalability
402
self._need_writelock()
404
if isinstance(files, types.StringTypes):
407
tree = self.working_tree()
410
# do this before any modifications
414
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
415
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
417
# having remove it, it must be either ignored or unknown
418
if tree.is_ignored(f):
422
show_status(new_status, inv[fid].kind, quotefn(f))
425
self._write_inventory(inv)
427
def set_inventory(self, new_inventory_list):
429
for path, file_id, parent, kind in new_inventory_list:
430
name = os.path.basename(path)
433
inv.add(InventoryEntry(file_id, name, kind, parent))
434
self._write_inventory(inv)
438
"""Return all unknown files.
440
These are files in the working directory that are not versioned or
441
control files or ignored.
443
>>> b = ScratchBranch(files=['foo', 'foo~'])
444
>>> list(b.unknowns())
447
>>> list(b.unknowns())
450
>>> list(b.unknowns())
453
return self.working_tree().unknowns()
456
def append_revision(self, revision_id):
457
mutter("add {%s} to revision-history" % revision_id)
458
rev_history = self.revision_history()
460
tmprhname = self.controlfilename('revision-history.tmp')
461
rhname = self.controlfilename('revision-history')
463
f = file(tmprhname, 'wt')
464
rev_history.append(revision_id)
465
f.write('\n'.join(rev_history))
469
if sys.platform == 'win32':
471
os.rename(tmprhname, rhname)
475
def get_revision(self, revision_id):
476
"""Return the Revision object for a named revision"""
477
self._need_readlock()
478
r = Revision.read_xml(self.revision_store[revision_id])
479
assert r.revision_id == revision_id
483
def get_inventory(self, inventory_id):
484
"""Get Inventory object by hash.
486
TODO: Perhaps for this and similar methods, take a revision
487
parameter which can be either an integer revno or a
489
self._need_readlock()
490
i = Inventory.read_xml(self.inventory_store[inventory_id])
494
def get_revision_inventory(self, revision_id):
495
"""Return inventory of a past revision."""
496
self._need_readlock()
497
if revision_id == None:
500
return self.get_inventory(self.get_revision(revision_id).inventory_id)
503
def revision_history(self):
504
"""Return sequence of revision hashes on to this branch.
506
>>> ScratchBranch().revision_history()
509
self._need_readlock()
510
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
513
def enum_history(self, direction):
514
"""Return (revno, revision_id) for history of branch.
517
'forward' is from earliest to latest
518
'reverse' is from latest to earliest
520
rh = self.revision_history()
521
if direction == 'forward':
526
elif direction == 'reverse':
532
raise ValueError('invalid history direction', direction)
536
"""Return current revision number for this branch.
538
That is equivalent to the number of revisions committed to
541
return len(self.revision_history())
544
def last_patch(self):
545
"""Return last patch hash, or None if no history.
547
ph = self.revision_history()
554
def commit(self, *args, **kw):
556
from bzrlib.commit import commit
557
commit(self, *args, **kw)
560
def lookup_revision(self, revno):
561
"""Return revision hash for revision number."""
566
# list is 0-based; revisions are 1-based
567
return self.revision_history()[revno-1]
569
raise BzrError("no such revision %s" % revno)
572
def revision_tree(self, revision_id):
573
"""Return Tree for a revision on this branch.
575
`revision_id` may be None for the null revision, in which case
576
an `EmptyTree` is returned."""
577
# TODO: refactor this to use an existing revision object
578
# so we don't need to read it in twice.
579
self._need_readlock()
580
if revision_id == None:
583
inv = self.get_revision_inventory(revision_id)
584
return RevisionTree(self.text_store, inv)
587
def working_tree(self):
588
"""Return a `Tree` for the working copy."""
589
from workingtree import WorkingTree
590
return WorkingTree(self.base, self.read_working_inventory())
593
def basis_tree(self):
594
"""Return `Tree` object for last revision.
596
If there are no revisions yet, return an `EmptyTree`.
598
r = self.last_patch()
602
return RevisionTree(self.text_store, self.get_revision_inventory(r))
606
def rename_one(self, from_rel, to_rel):
609
This can change the directory or the filename or both.
611
self._need_writelock()
612
tree = self.working_tree()
614
if not tree.has_filename(from_rel):
615
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
616
if tree.has_filename(to_rel):
617
raise BzrError("can't rename: new working file %r already exists" % to_rel)
619
file_id = inv.path2id(from_rel)
621
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
623
if inv.path2id(to_rel):
624
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
626
to_dir, to_tail = os.path.split(to_rel)
627
to_dir_id = inv.path2id(to_dir)
628
if to_dir_id == None and to_dir != '':
629
raise BzrError("can't determine destination directory id for %r" % to_dir)
631
mutter("rename_one:")
632
mutter(" file_id {%s}" % file_id)
633
mutter(" from_rel %r" % from_rel)
634
mutter(" to_rel %r" % to_rel)
635
mutter(" to_dir %r" % to_dir)
636
mutter(" to_dir_id {%s}" % to_dir_id)
638
inv.rename(file_id, to_dir_id, to_tail)
640
print "%s => %s" % (from_rel, to_rel)
642
from_abs = self.abspath(from_rel)
643
to_abs = self.abspath(to_rel)
645
os.rename(from_abs, to_abs)
647
raise BzrError("failed to rename %r to %r: %s"
648
% (from_abs, to_abs, e[1]),
649
["rename rolled back"])
651
self._write_inventory(inv)
655
def move(self, from_paths, to_name):
658
to_name must exist as a versioned directory.
660
If to_name exists and is a directory, the files are moved into
661
it, keeping their old names. If it is a directory,
663
Note that to_name is only the last component of the new name;
664
this doesn't change the directory.
666
self._need_writelock()
667
## TODO: Option to move IDs only
668
assert not isinstance(from_paths, basestring)
669
tree = self.working_tree()
671
to_abs = self.abspath(to_name)
672
if not isdir(to_abs):
673
raise BzrError("destination %r is not a directory" % to_abs)
674
if not tree.has_filename(to_name):
675
raise BzrError("destination %r not in working directory" % to_abs)
676
to_dir_id = inv.path2id(to_name)
677
if to_dir_id == None and to_name != '':
678
raise BzrError("destination %r is not a versioned directory" % to_name)
679
to_dir_ie = inv[to_dir_id]
680
if to_dir_ie.kind not in ('directory', 'root_directory'):
681
raise BzrError("destination %r is not a directory" % to_abs)
683
to_idpath = inv.get_idpath(to_dir_id)
686
if not tree.has_filename(f):
687
raise BzrError("%r does not exist in working tree" % f)
688
f_id = inv.path2id(f)
690
raise BzrError("%r is not versioned" % f)
691
name_tail = splitpath(f)[-1]
692
dest_path = appendpath(to_name, name_tail)
693
if tree.has_filename(dest_path):
694
raise BzrError("destination %r already exists" % dest_path)
695
if f_id in to_idpath:
696
raise BzrError("can't move %r to a subdirectory of itself" % f)
698
# OK, so there's a race here, it's possible that someone will
699
# create a file in this interval and then the rename might be
700
# left half-done. But we should have caught most problems.
703
name_tail = splitpath(f)[-1]
704
dest_path = appendpath(to_name, name_tail)
705
print "%s => %s" % (f, dest_path)
706
inv.rename(inv.path2id(f), to_dir_id, name_tail)
708
os.rename(self.abspath(f), self.abspath(dest_path))
710
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
711
["rename rolled back"])
713
self._write_inventory(inv)
718
class ScratchBranch(Branch):
719
"""Special test class: a branch that cleans up after itself.
721
>>> b = ScratchBranch()
729
def __init__(self, files=[], dirs=[]):
730
"""Make a test branch.
732
This creates a temporary directory and runs init-tree in it.
734
If any files are listed, they are created in the working copy.
736
Branch.__init__(self, tempfile.mkdtemp(), init=True)
738
os.mkdir(self.abspath(d))
741
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
748
"""Destroy the test branch, removing the scratch directory."""
750
mutter("delete ScratchBranch %s" % self.base)
751
shutil.rmtree(self.base)
753
# Work around for shutil.rmtree failing on Windows when
754
# readonly files are encountered
755
mutter("hit exception in destroying ScratchBranch: %s" % e)
756
for root, dirs, files in os.walk(self.base, topdown=False):
758
os.chmod(os.path.join(root, name), 0700)
759
shutil.rmtree(self.base)
764
######################################################################
768
def is_control_file(filename):
769
## FIXME: better check
770
filename = os.path.normpath(filename)
771
while filename != '':
772
head, tail = os.path.split(filename)
773
## mutter('check %r for control file' % ((head, tail), ))
774
if tail == bzrlib.BZRDIR:
783
def gen_file_id(name):
784
"""Return new file id.
786
This should probably generate proper UUIDs, but for the moment we
787
cope with just randomness because running uuidgen every time is
792
idx = name.rfind('/')
794
name = name[idx+1 : ]
795
idx = name.rfind('\\')
797
name = name[idx+1 : ]
799
# make it not a hidden file
800
name = name.lstrip('.')
802
# remove any wierd characters; we don't escape them but rather
804
name = re.sub(r'[^\w.]', '', name)
806
s = hexlify(rand_bytes(8))
807
return '-'.join((name, compact_date(time.time()), s))