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 bailout, 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.
92
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
93
"""Create new branch object at a particular location.
95
base -- Base directory for the branch.
97
init -- If True, create new control files in a previously
98
unversioned directory. If False, the branch must already
101
find_root -- If true and init is false, find the root of the
102
existing branch containing base.
104
In the test suite, creation of new trees is tested using the
105
`ScratchBranch` class.
108
self.base = os.path.realpath(base)
111
self.base = find_branch_root(base)
113
self.base = os.path.realpath(base)
114
if not isdir(self.controlfilename('.')):
115
bailout("not a bzr branch: %s" % quotefn(base),
116
['use "bzr init" to initialize a new working tree',
117
'current bzr can only operate from top-of-tree'])
121
self.text_store = ImmutableStore(self.controlfilename('text-store'))
122
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
123
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
127
return '%s(%r)' % (self.__class__.__name__, self.base)
134
def lock(self, mode='w'):
135
"""Lock the on-disk branch, excluding other processes."""
141
om = os.O_WRONLY | os.O_CREAT
146
raise BzrError("invalid locking mode %r" % mode)
149
lockfile = os.open(self.controlfilename('branch-lock'), om)
151
if e.errno == errno.ENOENT:
152
# might not exist on branches from <0.0.4
153
self.controlfile('branch-lock', 'w').close()
154
lockfile = os.open(self.controlfilename('branch-lock'), om)
158
fcntl.lockf(lockfile, lm)
160
fcntl.lockf(lockfile, fcntl.LOCK_UN)
162
self._lockmode = None
164
self._lockmode = mode
166
warning("please write a locking method for platform %r" % sys.platform)
168
self._lockmode = None
170
self._lockmode = mode
173
def _need_readlock(self):
174
if self._lockmode not in ['r', 'w']:
175
raise BzrError('need read lock on branch, only have %r' % self._lockmode)
177
def _need_writelock(self):
178
if self._lockmode not in ['w']:
179
raise BzrError('need write lock on branch, only have %r' % self._lockmode)
182
def abspath(self, name):
183
"""Return absolute filename for something in the branch"""
184
return os.path.join(self.base, name)
187
def relpath(self, path):
188
"""Return path relative to this branch of something inside it.
190
Raises an error if path is not in this branch."""
191
rp = os.path.realpath(path)
193
if not rp.startswith(self.base):
194
bailout("path %r is not within branch %r" % (rp, self.base))
195
rp = rp[len(self.base):]
196
rp = rp.lstrip(os.sep)
200
def controlfilename(self, file_or_path):
201
"""Return location relative to branch."""
202
if isinstance(file_or_path, types.StringTypes):
203
file_or_path = [file_or_path]
204
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
207
def controlfile(self, file_or_path, mode='r'):
208
"""Open a control file for this branch.
210
There are two classes of file in the control directory: text
211
and binary. binary files are untranslated byte streams. Text
212
control files are stored with Unix newlines and in UTF-8, even
213
if the platform or locale defaults are different.
215
Controlfiles should almost never be opened in write mode but
216
rather should be atomically copied and replaced using atomicfile.
219
fn = self.controlfilename(file_or_path)
221
if mode == 'rb' or mode == 'wb':
222
return file(fn, mode)
223
elif mode == 'r' or mode == 'w':
224
# open in binary mode anyhow so there's no newline translation;
225
# codecs uses line buffering by default; don't want that.
227
return codecs.open(fn, mode + 'b', 'utf-8',
230
raise BzrError("invalid controlfile mode %r" % mode)
234
def _make_control(self):
235
os.mkdir(self.controlfilename([]))
236
self.controlfile('README', 'w').write(
237
"This is a Bazaar-NG control directory.\n"
238
"Do not change any files in this directory.")
239
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
240
for d in ('text-store', 'inventory-store', 'revision-store'):
241
os.mkdir(self.controlfilename(d))
242
for f in ('revision-history', 'merged-patches',
243
'pending-merged-patches', 'branch-name',
245
self.controlfile(f, 'w').write('')
246
mutter('created control directory in ' + self.base)
247
Inventory().write_xml(self.controlfile('inventory','w'))
250
def _check_format(self):
251
"""Check this branch format is supported.
253
The current tool only supports the current unstable format.
255
In the future, we might need different in-memory Branch
256
classes to support downlevel branches. But not yet.
258
# This ignores newlines so that we can open branches created
259
# on Windows from Linux and so on. I think it might be better
260
# to always make all internal files in unix format.
261
fmt = self.controlfile('branch-format', 'r').read()
262
fmt.replace('\r\n', '')
263
if fmt != BZR_BRANCH_FORMAT:
264
bailout('sorry, branch format %r not supported' % fmt,
265
['use a different bzr version',
266
'or remove the .bzr directory and "bzr init" again'])
269
def read_working_inventory(self):
270
"""Read the working inventory."""
271
self._need_readlock()
273
# ElementTree does its own conversion from UTF-8, so open in
275
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
276
mutter("loaded inventory of %d items in %f"
277
% (len(inv), time.time() - before))
281
def _write_inventory(self, inv):
282
"""Update the working inventory.
284
That is to say, the inventory describing changes underway, that
285
will be committed to the next revision.
287
self._need_writelock()
288
## TODO: factor out to atomicfile? is rename safe on windows?
289
## TODO: Maybe some kind of clean/dirty marker on inventory?
290
tmpfname = self.controlfilename('inventory.tmp')
291
tmpf = file(tmpfname, 'wb')
294
inv_fname = self.controlfilename('inventory')
295
if sys.platform == 'win32':
297
os.rename(tmpfname, inv_fname)
298
mutter('wrote working inventory')
301
inventory = property(read_working_inventory, _write_inventory, None,
302
"""Inventory for the working copy.""")
305
def add(self, files, verbose=False, ids=None):
306
"""Make files versioned.
308
Note that the command line normally calls smart_add instead.
310
This puts the files in the Added state, so that they will be
311
recorded by the next commit.
313
TODO: Perhaps have an option to add the ids even if the files do
316
TODO: Perhaps return the ids of the files? But then again it
317
is easy to retrieve them if they're needed.
319
TODO: Option to specify file id.
321
TODO: Adding a directory should optionally recurse down and
322
add all non-ignored children. Perhaps do that in a
325
self._need_writelock()
327
# TODO: Re-adding a file that is removed in the working copy
328
# should probably put it back with the previous ID.
329
if isinstance(files, types.StringTypes):
330
assert(ids is None or isinstance(ids, types.StringTypes))
336
ids = [None] * len(files)
338
assert(len(ids) == len(files))
340
inv = self.read_working_inventory()
341
for f,file_id in zip(files, ids):
342
if is_control_file(f):
343
bailout("cannot add control file %s" % quotefn(f))
348
bailout("cannot add top-level %r" % f)
350
fullpath = os.path.normpath(self.abspath(f))
353
kind = file_kind(fullpath)
355
# maybe something better?
356
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
358
if kind != 'file' and kind != 'directory':
359
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
362
file_id = gen_file_id(f)
363
inv.add_path(f, kind=kind, file_id=file_id)
366
show_status('A', kind, quotefn(f))
368
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
370
self._write_inventory(inv)
373
def print_file(self, file, revno):
374
"""Print `file` to stdout."""
375
self._need_readlock()
376
tree = self.revision_tree(self.lookup_revision(revno))
377
# use inventory as it was in that revision
378
file_id = tree.inventory.path2id(file)
380
bailout("%r is not present in revision %d" % (file, revno))
381
tree.print_file(file_id)
384
def remove(self, files, verbose=False):
385
"""Mark nominated files for removal from the inventory.
387
This does not remove their text. This does not run on
389
TODO: Refuse to remove modified files unless --force is given?
391
TODO: Do something useful with directories.
393
TODO: Should this remove the text or not? Tough call; not
394
removing may be useful and the user can just use use rm, and
395
is the opposite of add. Removing it is consistent with most
396
other tools. Maybe an option.
398
## TODO: Normalize names
399
## TODO: Remove nested loops; better scalability
400
self._need_writelock()
402
if isinstance(files, types.StringTypes):
405
tree = self.working_tree()
408
# do this before any modifications
412
bailout("cannot remove unversioned file %s" % quotefn(f))
413
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
415
# having remove it, it must be either ignored or unknown
416
if tree.is_ignored(f):
420
show_status(new_status, inv[fid].kind, quotefn(f))
423
self._write_inventory(inv)
425
def set_inventory(self, new_inventory_list):
427
for path, file_id, parent, kind in new_inventory_list:
428
name = os.path.basename(path)
431
inv.add(InventoryEntry(file_id, name, kind, parent))
432
self._write_inventory(inv)
436
"""Return all unknown files.
438
These are files in the working directory that are not versioned or
439
control files or ignored.
441
>>> b = ScratchBranch(files=['foo', 'foo~'])
442
>>> list(b.unknowns())
445
>>> list(b.unknowns())
448
>>> list(b.unknowns())
451
return self.working_tree().unknowns()
454
def append_revision(self, revision_id):
455
mutter("add {%s} to revision-history" % revision_id)
456
rev_history = self.revision_history()
458
tmprhname = self.controlfilename('revision-history.tmp')
459
rhname = self.controlfilename('revision-history')
461
f = file(tmprhname, 'wt')
462
rev_history.append(revision_id)
463
f.write('\n'.join(rev_history))
467
if sys.platform == 'win32':
469
os.rename(tmprhname, rhname)
473
def get_revision(self, revision_id):
474
"""Return the Revision object for a named revision"""
475
self._need_readlock()
476
r = Revision.read_xml(self.revision_store[revision_id])
477
assert r.revision_id == revision_id
481
def get_inventory(self, inventory_id):
482
"""Get Inventory object by hash.
484
TODO: Perhaps for this and similar methods, take a revision
485
parameter which can be either an integer revno or a
487
self._need_readlock()
488
i = Inventory.read_xml(self.inventory_store[inventory_id])
492
def get_revision_inventory(self, revision_id):
493
"""Return inventory of a past revision."""
494
self._need_readlock()
495
if revision_id == None:
498
return self.get_inventory(self.get_revision(revision_id).inventory_id)
501
def revision_history(self):
502
"""Return sequence of revision hashes on to this branch.
504
>>> ScratchBranch().revision_history()
507
self._need_readlock()
508
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
511
def enum_history(self, direction):
512
"""Return (revno, revision_id) for history of branch.
515
'forward' is from earliest to latest
516
'reverse' is from latest to earliest
518
rh = self.revision_history()
519
if direction == 'forward':
524
elif direction == 'reverse':
530
raise ValueError('invalid history direction', direction)
534
"""Return current revision number for this branch.
536
That is equivalent to the number of revisions committed to
539
return len(self.revision_history())
542
def last_patch(self):
543
"""Return last patch hash, or None if no history.
545
ph = self.revision_history()
552
def commit(self, *args, **kw):
554
from bzrlib.commit import commit
555
commit(self, *args, **kw)
558
def lookup_revision(self, revno):
559
"""Return revision hash for revision number."""
564
# list is 0-based; revisions are 1-based
565
return self.revision_history()[revno-1]
567
raise BzrError("no such revision %s" % revno)
570
def revision_tree(self, revision_id):
571
"""Return Tree for a revision on this branch.
573
`revision_id` may be None for the null revision, in which case
574
an `EmptyTree` is returned."""
575
# TODO: refactor this to use an existing revision object
576
# so we don't need to read it in twice.
577
self._need_readlock()
578
if revision_id == None:
581
inv = self.get_revision_inventory(revision_id)
582
return RevisionTree(self.text_store, inv)
585
def working_tree(self):
586
"""Return a `Tree` for the working copy."""
587
from workingtree import WorkingTree
588
return WorkingTree(self.base, self.read_working_inventory())
591
def basis_tree(self):
592
"""Return `Tree` object for last revision.
594
If there are no revisions yet, return an `EmptyTree`.
596
r = self.last_patch()
600
return RevisionTree(self.text_store, self.get_revision_inventory(r))
604
def rename_one(self, from_rel, to_rel):
607
This can change the directory or the filename or both.
609
self._need_writelock()
610
tree = self.working_tree()
612
if not tree.has_filename(from_rel):
613
bailout("can't rename: old working file %r does not exist" % from_rel)
614
if tree.has_filename(to_rel):
615
bailout("can't rename: new working file %r already exists" % to_rel)
617
file_id = inv.path2id(from_rel)
619
bailout("can't rename: old name %r is not versioned" % from_rel)
621
if inv.path2id(to_rel):
622
bailout("can't rename: new name %r is already versioned" % to_rel)
624
to_dir, to_tail = os.path.split(to_rel)
625
to_dir_id = inv.path2id(to_dir)
626
if to_dir_id == None and to_dir != '':
627
bailout("can't determine destination directory id for %r" % to_dir)
629
mutter("rename_one:")
630
mutter(" file_id {%s}" % file_id)
631
mutter(" from_rel %r" % from_rel)
632
mutter(" to_rel %r" % to_rel)
633
mutter(" to_dir %r" % to_dir)
634
mutter(" to_dir_id {%s}" % to_dir_id)
636
inv.rename(file_id, to_dir_id, to_tail)
638
print "%s => %s" % (from_rel, to_rel)
640
from_abs = self.abspath(from_rel)
641
to_abs = self.abspath(to_rel)
643
os.rename(from_abs, to_abs)
645
bailout("failed to rename %r to %r: %s"
646
% (from_abs, to_abs, e[1]),
647
["rename rolled back"])
649
self._write_inventory(inv)
653
def move(self, from_paths, to_name):
656
to_name must exist as a versioned directory.
658
If to_name exists and is a directory, the files are moved into
659
it, keeping their old names. If it is a directory,
661
Note that to_name is only the last component of the new name;
662
this doesn't change the directory.
664
self._need_writelock()
665
## TODO: Option to move IDs only
666
assert not isinstance(from_paths, basestring)
667
tree = self.working_tree()
669
to_abs = self.abspath(to_name)
670
if not isdir(to_abs):
671
bailout("destination %r is not a directory" % to_abs)
672
if not tree.has_filename(to_name):
673
bailout("destination %r not in working directory" % to_abs)
674
to_dir_id = inv.path2id(to_name)
675
if to_dir_id == None and to_name != '':
676
bailout("destination %r is not a versioned directory" % to_name)
677
to_dir_ie = inv[to_dir_id]
678
if to_dir_ie.kind not in ('directory', 'root_directory'):
679
bailout("destination %r is not a directory" % to_abs)
681
to_idpath = inv.get_idpath(to_dir_id)
684
if not tree.has_filename(f):
685
bailout("%r does not exist in working tree" % f)
686
f_id = inv.path2id(f)
688
bailout("%r is not versioned" % f)
689
name_tail = splitpath(f)[-1]
690
dest_path = appendpath(to_name, name_tail)
691
if tree.has_filename(dest_path):
692
bailout("destination %r already exists" % dest_path)
693
if f_id in to_idpath:
694
bailout("can't move %r to a subdirectory of itself" % f)
696
# OK, so there's a race here, it's possible that someone will
697
# create a file in this interval and then the rename might be
698
# left half-done. But we should have caught most problems.
701
name_tail = splitpath(f)[-1]
702
dest_path = appendpath(to_name, name_tail)
703
print "%s => %s" % (f, dest_path)
704
inv.rename(inv.path2id(f), to_dir_id, name_tail)
706
os.rename(self.abspath(f), self.abspath(dest_path))
708
bailout("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
709
["rename rolled back"])
711
self._write_inventory(inv)
716
class ScratchBranch(Branch):
717
"""Special test class: a branch that cleans up after itself.
719
>>> b = ScratchBranch()
727
def __init__(self, files=[], dirs=[]):
728
"""Make a test branch.
730
This creates a temporary directory and runs init-tree in it.
732
If any files are listed, they are created in the working copy.
734
Branch.__init__(self, tempfile.mkdtemp(), init=True)
736
os.mkdir(self.abspath(d))
739
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
746
"""Destroy the test branch, removing the scratch directory."""
748
mutter("delete ScratchBranch %s" % self.base)
749
shutil.rmtree(self.base)
751
# Work around for shutil.rmtree failing on Windows when
752
# readonly files are encountered
753
mutter("hit exception in destroying ScratchBranch: %s" % e)
754
for root, dirs, files in os.walk(self.base, topdown=False):
756
os.chmod(os.path.join(root, name), 0700)
757
shutil.rmtree(self.base)
762
######################################################################
766
def is_control_file(filename):
767
## FIXME: better check
768
filename = os.path.normpath(filename)
769
while filename != '':
770
head, tail = os.path.split(filename)
771
## mutter('check %r for control file' % ((head, tail), ))
772
if tail == bzrlib.BZRDIR:
781
def gen_file_id(name):
782
"""Return new file id.
784
This should probably generate proper UUIDs, but for the moment we
785
cope with just randomness because running uuidgen every time is
790
idx = name.rfind('/')
792
name = name[idx+1 : ]
793
idx = name.rfind('\\')
795
name = name[idx+1 : ]
797
# make it not a hidden file
798
name = name.lstrip('.')
800
# remove any wierd characters; we don't escape them but rather
802
name = re.sub(r'[^\w.]', '', name)
804
s = hexlify(rand_bytes(8))
805
return '-'.join((name, compact_date(time.time()), s))