26
23
from inventory import Inventory
27
24
from trace import mutter, note
28
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
25
from tree import Tree, EmptyTree, RevisionTree
29
26
from inventory import InventoryEntry, Inventory
30
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, chomp, \
27
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, \
31
28
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
32
joinpath, sha_string, file_kind, local_time_offset
29
joinpath, sha_string, file_kind, local_time_offset, appendpath
33
30
from store import ImmutableStore
34
31
from revision import Revision
35
from errors import bailout
32
from errors import BzrError
36
33
from textui import show_status
37
from diff import diff_trees
39
37
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
40
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)
46
82
######################################################################
50
86
"""Branch holding a history of revisions.
52
:todo: Perhaps use different stores for different classes of object,
53
so that we can keep track of how much space each one uses,
54
or garbage-collect them.
56
:todo: Add a RemoteBranch subclass. For the basic case of read-only
57
HTTP access this should be very easy by,
58
just redirecting controlfile access into HTTP requests.
59
We would need a RemoteStore working similarly.
61
:todo: Keep the on-disk branch locked while the object exists.
63
:todo: mkdir() method.
89
Base directory of the branch.
65
def __init__(self, base, init=False):
94
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
66
95
"""Create new branch object at a particular location.
68
:param base: Base directory for the branch.
70
:param init: If True, create new control files in a previously
97
base -- Base directory for the branch.
99
init -- If True, create new control files in a previously
71
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.
74
106
In the test suite, creation of new trees is tested using the
75
107
`ScratchBranch` class.
77
self.base = os.path.realpath(base)
110
self.base = os.path.realpath(base)
79
111
self._make_control()
113
self.base = find_branch_root(base)
115
self.base = os.path.realpath(base)
81
116
if not isdir(self.controlfilename('.')):
82
bailout("not a bzr branch: %s" % quotefn(base),
83
['use "bzr init" to initialize a new working tree',
84
'current bzr can only operate from top-of-tree'])
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
87
125
self.text_store = ImmutableStore(self.controlfilename('text-store'))
88
126
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
96
134
__repr__ = __str__
100
"""Return filename relative to branch top"""
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"""
101
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)
104
206
def controlfilename(self, file_or_path):
105
207
"""Return location relative to branch."""
136
261
In the future, we might need different in-memory Branch
137
262
classes to support downlevel branches. But not yet.
139
# read in binary mode to detect newline wierdness.
140
fmt = self.controlfile('branch-format', 'rb').read()
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', '')
141
269
if fmt != BZR_BRANCH_FORMAT:
142
bailout('sorry, branch format %r not supported' % fmt,
143
['use a different bzr version',
144
'or remove the .bzr directory and "bzr init" again'])
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'])
147
275
def read_working_inventory(self):
148
276
"""Read the working inventory."""
277
self._need_readlock()
149
278
before = time.time()
150
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
279
# ElementTree does its own conversion from UTF-8, so open in
281
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
151
282
mutter("loaded inventory of %d items in %f"
152
283
% (len(inv), time.time() - before))
172
308
"""Inventory for the working copy.""")
175
def add(self, files, verbose=False):
311
def add(self, files, verbose=False, ids=None):
176
312
"""Make files versioned.
314
Note that the command line normally calls smart_add instead.
178
316
This puts the files in the Added state, so that they will be
179
317
recorded by the next commit.
181
:todo: Perhaps have an option to add the ids even if the files do
319
TODO: Perhaps have an option to add the ids even if the files do
184
:todo: Perhaps return the ids of the files? But then again it
322
TODO: Perhaps return the ids of the files? But then again it
185
323
is easy to retrieve them if they're needed.
187
:todo: Option to specify file id.
325
TODO: Option to specify file id.
189
:todo: Adding a directory should optionally recurse down and
327
TODO: Adding a directory should optionally recurse down and
190
328
add all non-ignored children. Perhaps do that in a
191
329
higher-level method.
193
>>> b = ScratchBranch(files=['foo'])
194
>>> 'foo' in b.unknowns()
199
>>> 'foo' in b.unknowns()
201
>>> bool(b.inventory.path2id('foo'))
207
Traceback (most recent call last):
209
BzrError: ('foo is already versioned', [])
211
>>> b.add(['nothere'])
212
Traceback (most recent call last):
213
BzrError: ('cannot add: not a regular file or directory: nothere', [])
331
self._need_writelock()
216
333
# TODO: Re-adding a file that is removed in the working copy
217
334
# should probably put it back with the previous ID.
218
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))
221
346
inv = self.read_working_inventory()
347
for f,file_id in zip(files, ids):
223
348
if is_control_file(f):
224
bailout("cannot add control file %s" % quotefn(f))
349
raise BzrError("cannot add control file %s" % quotefn(f))
226
351
fp = splitpath(f)
229
bailout("cannot add top-level %r" % f)
354
raise BzrError("cannot add top-level %r" % f)
231
fullpath = os.path.normpath(self._rel(f))
235
elif isdir(fullpath):
238
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
241
parent_name = joinpath(fp[:-1])
242
mutter("lookup parent %r" % parent_name)
243
parent_id = inv.path2id(parent_name)
244
if parent_id == None:
245
bailout("cannot add: parent %r is not versioned"
250
file_id = _gen_file_id(fp[-1])
251
inv.add(InventoryEntry(file_id, fp[-1], kind=kind, parent_id=parent_id))
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)
253
372
show_status('A', kind, quotefn(f))
255
mutter("add file %s file_id:{%s} kind=%r parent_id={%s}"
256
% (f, file_id, kind, parent_id))
374
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
257
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)
261
390
def remove(self, files, verbose=False):
262
391
"""Mark nominated files for removal from the inventory.
264
393
This does not remove their text. This does not run on
266
:todo: Refuse to remove modified files unless --force is given?
268
>>> b = ScratchBranch(files=['foo'])
270
>>> b.inventory.has_filename('foo')
273
>>> b.working_tree().has_filename('foo')
275
>>> b.inventory.has_filename('foo')
278
>>> b = ScratchBranch(files=['foo'])
283
>>> b.inventory.has_filename('foo')
285
>>> b.basis_tree().has_filename('foo')
287
>>> b.working_tree().has_filename('foo')
290
:todo: Do something useful with directories.
292
:todo: Should this remove the text or not? Tough call; not
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
293
400
removing may be useful and the user can just use use rm, and
294
401
is the opposite of add. Removing it is consistent with most
295
402
other tools. Maybe an option.
297
404
## TODO: Normalize names
298
405
## TODO: Remove nested loops; better scalability
406
self._need_writelock()
300
408
if isinstance(files, types.StringTypes):
303
inv = self.read_working_inventory()
411
tree = self.working_tree()
305
414
# do this before any modifications
307
416
fid = inv.path2id(f)
309
bailout("cannot remove unversioned file %s" % quotefn(f))
418
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
310
419
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
312
show_status('D', inv[fid].kind, quotefn(f))
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))
315
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)
318
441
def unknowns(self):
319
442
"""Return all unknown files.
334
457
return self.working_tree().unknowns()
337
def commit(self, message, timestamp=None, timezone=None,
340
"""Commit working copy as a new revision.
342
The basic approach is to add all the file texts into the
343
store, then the inventory, then make a new revision pointing
344
to that inventory and store that.
346
This is not quite safe if the working copy changes during the
347
commit; for the moment that is simply not allowed. A better
348
approach is to make a temporary copy of the files before
349
computing their hashes, and then add those hashes in turn to
350
the inventory. This should mean at least that there are no
351
broken hash pointers. There is no way we can get a snapshot
352
of the whole directory at an instant. This would also have to
353
be robust against files disappearing, moving, etc. So the
354
whole thing is a bit hard.
356
:param timestamp: if not None, seconds-since-epoch for a
357
postdated/predated commit.
360
## TODO: Show branch names
362
# TODO: Don't commit if there are no changes, unless forced?
364
# First walk over the working inventory; and both update that
365
# and also build a new revision inventory. The revision
366
# inventory needs to hold the text-id, sha1 and size of the
367
# actual file versions committed in the revision. (These are
368
# not present in the working inventory.) We also need to
369
# detect missing/deleted files, and remove them from the
372
work_inv = self.read_working_inventory()
374
basis = self.basis_tree()
375
basis_inv = basis.inventory
377
for path, entry in work_inv.iter_entries():
378
## TODO: Cope with files that have gone missing.
380
## TODO: Check that the file kind has not changed from the previous
381
## revision of this file (if any).
386
file_id = entry.file_id
387
mutter('commit prep file %s, id %r ' % (p, file_id))
389
if not os.path.exists(p):
390
mutter(" file is missing, removing from inventory")
392
show_status('D', entry.kind, quotefn(path))
393
missing_ids.append(file_id)
396
# TODO: Handle files that have been deleted
398
# TODO: Maybe a special case for empty files? Seems a
399
# waste to store them many times.
403
if basis_inv.has_id(file_id):
404
old_kind = basis_inv[file_id].kind
405
if old_kind != entry.kind:
406
bailout("entry %r changed kind from %r to %r"
407
% (file_id, old_kind, entry.kind))
409
if entry.kind == 'directory':
411
bailout("%s is entered as directory but not a directory" % quotefn(p))
412
elif entry.kind == 'file':
414
bailout("%s is entered as file but is not a file" % quotefn(p))
416
content = file(p, 'rb').read()
418
entry.text_sha1 = sha_string(content)
419
entry.text_size = len(content)
421
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
423
and (old_ie.text_size == entry.text_size)
424
and (old_ie.text_sha1 == entry.text_sha1)):
425
## assert content == basis.get_file(file_id).read()
426
entry.text_id = basis_inv[file_id].text_id
427
mutter(' unchanged from previous text_id {%s}' %
431
entry.text_id = _gen_file_id(entry.name)
432
self.text_store.add(content, entry.text_id)
433
mutter(' stored with text_id {%s}' % entry.text_id)
437
elif (old_ie.name == entry.name
438
and old_ie.parent_id == entry.parent_id):
443
show_status(state, entry.kind, quotefn(path))
445
for file_id in missing_ids:
446
# have to do this later so we don't mess up the iterator.
447
# since parents may be removed before their children we
450
# FIXME: There's probably a better way to do this; perhaps
451
# the workingtree should know how to filter itself.
452
if work_inv.has_id(file_id):
453
del work_inv[file_id]
456
inv_id = rev_id = _gen_revision_id(time.time())
458
inv_tmp = tempfile.TemporaryFile()
459
inv.write_xml(inv_tmp)
461
self.inventory_store.add(inv_tmp, inv_id)
462
mutter('new inventory_id is {%s}' % inv_id)
464
self._write_inventory(work_inv)
466
if timestamp == None:
467
timestamp = time.time()
469
if committer == None:
470
committer = username()
473
timezone = local_time_offset()
475
mutter("building commit log message")
476
rev = Revision(timestamp=timestamp,
479
precursor = self.last_patch(),
484
rev_tmp = tempfile.TemporaryFile()
485
rev.write_xml(rev_tmp)
487
self.revision_store.add(rev_tmp, rev_id)
488
mutter("new revision_id is {%s}" % rev_id)
490
## XXX: Everything up to here can simply be orphaned if we abort
491
## the commit; it will leave junk files behind but that doesn't
494
## TODO: Read back the just-generated changeset, and make sure it
495
## applies and recreates the right state.
497
## TODO: Also calculate and store the inventory SHA1
498
mutter("committing patch r%d" % (self.revno() + 1))
500
mutter("append to revision-history")
501
self.controlfile('revision-history', 'at').write(rev_id + '\n')
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)
506
479
def get_revision(self, revision_id):
507
480
"""Return the Revision object for a named revision"""
481
self._need_readlock()
508
482
r = Revision.read_xml(self.revision_store[revision_id])
509
483
assert r.revision_id == revision_id
620
def write_log(self, show_timezone='original'):
621
"""Write out human-readable log of commits to this branch
623
:param utc: If true, show dates in universal time, not local time."""
624
## TODO: Option to choose either original, utc or local timezone
627
for p in self.revision_history():
629
print 'revno:', revno
630
## TODO: Show hash if --id is given.
631
##print 'revision-hash:', p
632
rev = self.get_revision(p)
633
print 'committer:', rev.committer
634
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
637
## opportunistic consistency check, same as check_patch_chaining
638
if rev.precursor != precursor:
639
bailout("mismatched precursor!")
643
print ' (no message)'
645
for l in rev.message.split('\n'):
653
def show_status(branch, show_all=False):
654
"""Display single-line status for non-ignored working files.
656
The list is show sorted in order by file name.
658
>>> b = ScratchBranch(files=['foo', 'foo~'])
664
>>> b.commit("add foo")
666
>>> os.unlink(b._rel('foo'))
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)
671
:todo: Get state for single files.
673
:todo: Perhaps show a slash at the end of directory names.
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.
677
# We have to build everything into a list first so that it can
678
# sorted by name, incorporating all the different sources.
680
# FIXME: Rather than getting things in random order and then sorting,
681
# just step through in order.
683
# Interesting case: the old ID for a file has been removed,
684
# but a new file has been created under that name.
686
old = branch.basis_tree()
687
old_inv = old.inventory
688
new = branch.working_tree()
689
new_inv = new.inventory
691
for fs, fid, oldname, newname, kind in diff_trees(old, new):
693
show_status(fs, kind,
694
oldname + ' => ' + newname)
695
elif fs == 'A' or fs == 'M':
696
show_status(fs, kind, newname)
698
show_status(fs, kind, oldname)
701
show_status(fs, kind, newname)
704
show_status(fs, kind, newname)
706
show_status(fs, kind, newname)
708
bailout("wierd file state %r" % ((fs, fid),))
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)
712
722
class ScratchBranch(Branch):