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.
91
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
92
"""Create new branch object at a particular location.
94
base -- Base directory for the branch.
96
init -- If True, create new control files in a previously
97
unversioned directory. If False, the branch must already
100
find_root -- If true and init is false, find the root of the
101
existing branch containing base.
103
In the test suite, creation of new trees is tested using the
104
`ScratchBranch` class.
107
self.base = os.path.realpath(base)
110
self.base = find_branch_root(base)
112
self.base = os.path.realpath(base)
113
if not isdir(self.controlfilename('.')):
114
bailout("not a bzr branch: %s" % quotefn(base),
115
['use "bzr init" to initialize a new working tree',
116
'current bzr can only operate from top-of-tree'])
120
self.text_store = ImmutableStore(self.controlfilename('text-store'))
121
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
122
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
126
return '%s(%r)' % (self.__class__.__name__, self.base)
133
def lock(self, mode='w'):
134
"""Lock the on-disk branch, excluding other processes."""
140
om = os.O_WRONLY | os.O_CREAT
145
raise BzrError("invalid locking mode %r" % mode)
148
lockfile = os.open(self.controlfilename('branch-lock'), om)
150
if e.errno == errno.ENOENT:
151
# might not exist on branches from <0.0.4
152
self.controlfile('branch-lock', 'w').close()
153
lockfile = os.open(self.controlfilename('branch-lock'), om)
157
fcntl.lockf(lockfile, lm)
159
fcntl.lockf(lockfile, fcntl.LOCK_UN)
161
self._lockmode = None
163
self._lockmode = mode
165
warning("please write a locking method for platform %r" % sys.platform)
167
self._lockmode = None
169
self._lockmode = mode
172
def _need_readlock(self):
173
if self._lockmode not in ['r', 'w']:
174
raise BzrError('need read lock on branch, only have %r' % self._lockmode)
176
def _need_writelock(self):
177
if self._lockmode not in ['w']:
178
raise BzrError('need write lock on branch, only have %r' % self._lockmode)
181
def abspath(self, name):
182
"""Return absolute filename for something in the branch"""
183
return os.path.join(self.base, name)
186
def relpath(self, path):
187
"""Return path relative to this branch of something inside it.
189
Raises an error if path is not in this branch."""
190
rp = os.path.realpath(path)
192
if not rp.startswith(self.base):
193
bailout("path %r is not within branch %r" % (rp, self.base))
194
rp = rp[len(self.base):]
195
rp = rp.lstrip(os.sep)
199
def controlfilename(self, file_or_path):
200
"""Return location relative to branch."""
201
if isinstance(file_or_path, types.StringTypes):
202
file_or_path = [file_or_path]
203
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
206
def controlfile(self, file_or_path, mode='r'):
207
"""Open a control file for this branch.
209
There are two classes of file in the control directory: text
210
and binary. binary files are untranslated byte streams. Text
211
control files are stored with Unix newlines and in UTF-8, even
212
if the platform or locale defaults are different.
214
Controlfiles should almost never be opened in write mode but
215
rather should be atomically copied and replaced using atomicfile.
218
fn = self.controlfilename(file_or_path)
220
if mode == 'rb' or mode == 'wb':
221
return file(fn, mode)
222
elif mode == 'r' or mode == 'w':
223
# open in binary mode anyhow so there's no newline translation;
224
# codecs uses line buffering by default; don't want that.
226
return codecs.open(fn, mode + 'b', 'utf-8',
229
raise BzrError("invalid controlfile mode %r" % mode)
233
def _make_control(self):
234
os.mkdir(self.controlfilename([]))
235
self.controlfile('README', 'w').write(
236
"This is a Bazaar-NG control directory.\n"
237
"Do not change any files in this directory.")
238
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
239
for d in ('text-store', 'inventory-store', 'revision-store'):
240
os.mkdir(self.controlfilename(d))
241
for f in ('revision-history', 'merged-patches',
242
'pending-merged-patches', 'branch-name',
244
self.controlfile(f, 'w').write('')
245
mutter('created control directory in ' + self.base)
246
Inventory().write_xml(self.controlfile('inventory','w'))
249
def _check_format(self):
250
"""Check this branch format is supported.
252
The current tool only supports the current unstable format.
254
In the future, we might need different in-memory Branch
255
classes to support downlevel branches. But not yet.
257
# This ignores newlines so that we can open branches created
258
# on Windows from Linux and so on. I think it might be better
259
# to always make all internal files in unix format.
260
fmt = self.controlfile('branch-format', 'r').read()
261
fmt.replace('\r\n', '')
262
if fmt != BZR_BRANCH_FORMAT:
263
bailout('sorry, branch format %r not supported' % fmt,
264
['use a different bzr version',
265
'or remove the .bzr directory and "bzr init" again'])
268
def read_working_inventory(self):
269
"""Read the working inventory."""
270
self._need_readlock()
272
# ElementTree does its own conversion from UTF-8, so open in
274
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
275
mutter("loaded inventory of %d items in %f"
276
% (len(inv), time.time() - before))
280
def _write_inventory(self, inv):
281
"""Update the working inventory.
283
That is to say, the inventory describing changes underway, that
284
will be committed to the next revision.
286
self._need_writelock()
287
## TODO: factor out to atomicfile? is rename safe on windows?
288
## TODO: Maybe some kind of clean/dirty marker on inventory?
289
tmpfname = self.controlfilename('inventory.tmp')
290
tmpf = file(tmpfname, 'wb')
293
inv_fname = self.controlfilename('inventory')
294
if sys.platform == 'win32':
296
os.rename(tmpfname, inv_fname)
297
mutter('wrote working inventory')
300
inventory = property(read_working_inventory, _write_inventory, None,
301
"""Inventory for the working copy.""")
304
def add(self, files, verbose=False, ids=None):
305
"""Make files versioned.
307
Note that the command line normally calls smart_add instead.
309
This puts the files in the Added state, so that they will be
310
recorded by the next commit.
312
TODO: Perhaps have an option to add the ids even if the files do
315
TODO: Perhaps return the ids of the files? But then again it
316
is easy to retrieve them if they're needed.
318
TODO: Option to specify file id.
320
TODO: Adding a directory should optionally recurse down and
321
add all non-ignored children. Perhaps do that in a
324
self._need_writelock()
326
# TODO: Re-adding a file that is removed in the working copy
327
# should probably put it back with the previous ID.
328
if isinstance(files, types.StringTypes):
329
assert(ids is None or isinstance(ids, types.StringTypes))
335
ids = [None] * len(files)
337
assert(len(ids) == len(files))
339
inv = self.read_working_inventory()
340
for f,file_id in zip(files, ids):
341
if is_control_file(f):
342
bailout("cannot add control file %s" % quotefn(f))
347
bailout("cannot add top-level %r" % f)
349
fullpath = os.path.normpath(self.abspath(f))
352
kind = file_kind(fullpath)
354
# maybe something better?
355
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
357
if kind != 'file' and kind != 'directory':
358
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
361
file_id = gen_file_id(f)
362
inv.add_path(f, kind=kind, file_id=file_id)
365
show_status('A', kind, quotefn(f))
367
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
369
self._write_inventory(inv)
372
def print_file(self, file, revno):
373
"""Print `file` to stdout."""
374
self._need_readlock()
375
tree = self.revision_tree(self.lookup_revision(revno))
376
# use inventory as it was in that revision
377
file_id = tree.inventory.path2id(file)
379
bailout("%r is not present in revision %d" % (file, revno))
380
tree.print_file(file_id)
383
def remove(self, files, verbose=False):
384
"""Mark nominated files for removal from the inventory.
386
This does not remove their text. This does not run on
388
TODO: Refuse to remove modified files unless --force is given?
390
TODO: Do something useful with directories.
392
TODO: Should this remove the text or not? Tough call; not
393
removing may be useful and the user can just use use rm, and
394
is the opposite of add. Removing it is consistent with most
395
other tools. Maybe an option.
397
## TODO: Normalize names
398
## TODO: Remove nested loops; better scalability
399
self._need_writelock()
401
if isinstance(files, types.StringTypes):
404
tree = self.working_tree()
407
# do this before any modifications
411
bailout("cannot remove unversioned file %s" % quotefn(f))
412
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
414
# having remove it, it must be either ignored or unknown
415
if tree.is_ignored(f):
419
show_status(new_status, inv[fid].kind, quotefn(f))
422
self._write_inventory(inv)
424
def set_inventory(self, new_inventory_list):
426
for path, file_id, parent, kind in new_inventory_list:
427
name = os.path.basename(path)
430
inv.add(InventoryEntry(file_id, name, kind, parent))
431
self._write_inventory(inv)
435
"""Return all unknown files.
437
These are files in the working directory that are not versioned or
438
control files or ignored.
440
>>> b = ScratchBranch(files=['foo', 'foo~'])
441
>>> list(b.unknowns())
444
>>> list(b.unknowns())
447
>>> list(b.unknowns())
450
return self.working_tree().unknowns()
453
def append_revision(self, revision_id):
454
mutter("add {%s} to revision-history" % revision_id)
455
rev_history = self.revision_history()
457
tmprhname = self.controlfilename('revision-history.tmp')
458
rhname = self.controlfilename('revision-history')
460
f = file(tmprhname, 'wt')
461
rev_history.append(revision_id)
462
f.write('\n'.join(rev_history))
466
if sys.platform == 'win32':
468
os.rename(tmprhname, rhname)
472
def get_revision(self, revision_id):
473
"""Return the Revision object for a named revision"""
474
self._need_readlock()
475
r = Revision.read_xml(self.revision_store[revision_id])
476
assert r.revision_id == revision_id
480
def get_inventory(self, inventory_id):
481
"""Get Inventory object by hash.
483
TODO: Perhaps for this and similar methods, take a revision
484
parameter which can be either an integer revno or a
486
self._need_readlock()
487
i = Inventory.read_xml(self.inventory_store[inventory_id])
491
def get_revision_inventory(self, revision_id):
492
"""Return inventory of a past revision."""
493
self._need_readlock()
494
if revision_id == None:
497
return self.get_inventory(self.get_revision(revision_id).inventory_id)
500
def revision_history(self):
501
"""Return sequence of revision hashes on to this branch.
503
>>> ScratchBranch().revision_history()
506
self._need_readlock()
507
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
510
def enum_history(self, direction):
511
"""Return (revno, revision_id) for history of branch.
514
'forward' is from earliest to latest
515
'reverse' is from latest to earliest
517
rh = self.revision_history()
518
if direction == 'forward':
523
elif direction == 'reverse':
529
raise ValueError('invalid history direction', direction)
533
"""Return current revision number for this branch.
535
That is equivalent to the number of revisions committed to
538
return len(self.revision_history())
541
def last_patch(self):
542
"""Return last patch hash, or None if no history.
544
ph = self.revision_history()
551
def commit(self, *args, **kw):
553
from bzrlib.commit import commit
554
commit(self, *args, **kw)
557
def lookup_revision(self, revno):
558
"""Return revision hash for revision number."""
563
# list is 0-based; revisions are 1-based
564
return self.revision_history()[revno-1]
566
raise BzrError("no such revision %s" % revno)
569
def revision_tree(self, revision_id):
570
"""Return Tree for a revision on this branch.
572
`revision_id` may be None for the null revision, in which case
573
an `EmptyTree` is returned."""
574
# TODO: refactor this to use an existing revision object
575
# so we don't need to read it in twice.
576
self._need_readlock()
577
if revision_id == None:
580
inv = self.get_revision_inventory(revision_id)
581
return RevisionTree(self.text_store, inv)
584
def working_tree(self):
585
"""Return a `Tree` for the working copy."""
586
from workingtree import WorkingTree
587
return WorkingTree(self.base, self.read_working_inventory())
590
def basis_tree(self):
591
"""Return `Tree` object for last revision.
593
If there are no revisions yet, return an `EmptyTree`.
595
r = self.last_patch()
599
return RevisionTree(self.text_store, self.get_revision_inventory(r))
603
def rename_one(self, from_rel, to_rel):
606
This can change the directory or the filename or both.
608
self._need_writelock()
609
tree = self.working_tree()
611
if not tree.has_filename(from_rel):
612
bailout("can't rename: old working file %r does not exist" % from_rel)
613
if tree.has_filename(to_rel):
614
bailout("can't rename: new working file %r already exists" % to_rel)
616
file_id = inv.path2id(from_rel)
618
bailout("can't rename: old name %r is not versioned" % from_rel)
620
if inv.path2id(to_rel):
621
bailout("can't rename: new name %r is already versioned" % to_rel)
623
to_dir, to_tail = os.path.split(to_rel)
624
to_dir_id = inv.path2id(to_dir)
625
if to_dir_id == None and to_dir != '':
626
bailout("can't determine destination directory id for %r" % to_dir)
628
mutter("rename_one:")
629
mutter(" file_id {%s}" % file_id)
630
mutter(" from_rel %r" % from_rel)
631
mutter(" to_rel %r" % to_rel)
632
mutter(" to_dir %r" % to_dir)
633
mutter(" to_dir_id {%s}" % to_dir_id)
635
inv.rename(file_id, to_dir_id, to_tail)
637
print "%s => %s" % (from_rel, to_rel)
639
from_abs = self.abspath(from_rel)
640
to_abs = self.abspath(to_rel)
642
os.rename(from_abs, to_abs)
644
bailout("failed to rename %r to %r: %s"
645
% (from_abs, to_abs, e[1]),
646
["rename rolled back"])
648
self._write_inventory(inv)
652
def move(self, from_paths, to_name):
655
to_name must exist as a versioned directory.
657
If to_name exists and is a directory, the files are moved into
658
it, keeping their old names. If it is a directory,
660
Note that to_name is only the last component of the new name;
661
this doesn't change the directory.
663
self._need_writelock()
664
## TODO: Option to move IDs only
665
assert not isinstance(from_paths, basestring)
666
tree = self.working_tree()
668
to_abs = self.abspath(to_name)
669
if not isdir(to_abs):
670
bailout("destination %r is not a directory" % to_abs)
671
if not tree.has_filename(to_name):
672
bailout("destination %r not in working directory" % to_abs)
673
to_dir_id = inv.path2id(to_name)
674
if to_dir_id == None and to_name != '':
675
bailout("destination %r is not a versioned directory" % to_name)
676
to_dir_ie = inv[to_dir_id]
677
if to_dir_ie.kind not in ('directory', 'root_directory'):
678
bailout("destination %r is not a directory" % to_abs)
680
to_idpath = inv.get_idpath(to_dir_id)
683
if not tree.has_filename(f):
684
bailout("%r does not exist in working tree" % f)
685
f_id = inv.path2id(f)
687
bailout("%r is not versioned" % f)
688
name_tail = splitpath(f)[-1]
689
dest_path = appendpath(to_name, name_tail)
690
if tree.has_filename(dest_path):
691
bailout("destination %r already exists" % dest_path)
692
if f_id in to_idpath:
693
bailout("can't move %r to a subdirectory of itself" % f)
695
# OK, so there's a race here, it's possible that someone will
696
# create a file in this interval and then the rename might be
697
# left half-done. But we should have caught most problems.
700
name_tail = splitpath(f)[-1]
701
dest_path = appendpath(to_name, name_tail)
702
print "%s => %s" % (f, dest_path)
703
inv.rename(inv.path2id(f), to_dir_id, name_tail)
705
os.rename(self.abspath(f), self.abspath(dest_path))
707
bailout("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
708
["rename rolled back"])
710
self._write_inventory(inv)
715
class ScratchBranch(Branch):
716
"""Special test class: a branch that cleans up after itself.
718
>>> b = ScratchBranch()
726
def __init__(self, files=[], dirs=[]):
727
"""Make a test branch.
729
This creates a temporary directory and runs init-tree in it.
731
If any files are listed, they are created in the working copy.
733
Branch.__init__(self, tempfile.mkdtemp(), init=True)
735
os.mkdir(self.abspath(d))
738
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
745
"""Destroy the test branch, removing the scratch directory."""
747
mutter("delete ScratchBranch %s" % self.base)
748
shutil.rmtree(self.base)
750
# Work around for shutil.rmtree failing on Windows when
751
# readonly files are encountered
752
mutter("hit exception in destroying ScratchBranch: %s" % e)
753
for root, dirs, files in os.walk(self.base, topdown=False):
755
os.chmod(os.path.join(root, name), 0700)
756
shutil.rmtree(self.base)
761
######################################################################
765
def is_control_file(filename):
766
## FIXME: better check
767
filename = os.path.normpath(filename)
768
while filename != '':
769
head, tail = os.path.split(filename)
770
## mutter('check %r for control file' % ((head, tail), ))
771
if tail == bzrlib.BZRDIR:
780
def gen_file_id(name):
781
"""Return new file id.
783
This should probably generate proper UUIDs, but for the moment we
784
cope with just randomness because running uuidgen every time is
789
idx = name.rfind('/')
791
name = name[idx+1 : ]
792
idx = name.rfind('\\')
794
name = name[idx+1 : ]
796
# make it not a hidden file
797
name = name.lstrip('.')
799
# remove any wierd characters; we don't escape them but rather
801
name = re.sub(r'[^\w.]', '', name)
803
s = hexlify(rand_bytes(8))
804
return '-'.join((name, compact_date(time.time()), s))