23
26
from inventory import Inventory
24
27
from trace import mutter, note
25
from tree import Tree, EmptyTree, RevisionTree
28
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
26
29
from inventory import InventoryEntry, Inventory
27
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, \
30
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, chomp, \
28
31
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
29
joinpath, sha_string, file_kind, local_time_offset, appendpath
32
joinpath, sha_string, file_kind
30
33
from store import ImmutableStore
31
34
from revision import Revision
32
from errors import BzrError
35
from errors import bailout
33
36
from textui import show_status
37
from diff import diff_trees
35
39
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
36
40
## 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)
49
def with_writelock(method):
50
"""Method decorator for functions run with the branch locked."""
52
# called with self set to the branch
55
return method(self, *a, **k)
61
def with_readlock(method):
65
return method(self, *a, **k)
71
def find_branch_root(f=None):
72
"""Find the branch root enclosing f, or pwd.
74
f may be a filename or a URL.
76
It is not necessary that f exists.
78
Basically we keep looking up until we find the control directory or
82
elif hasattr(os.path, 'realpath'):
83
f = os.path.realpath(f)
85
f = os.path.abspath(f)
86
if not os.path.exists(f):
87
raise BzrError('%r does not exist' % f)
93
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
95
head, tail = os.path.split(f)
97
# reached the root, whatever that may be
98
raise BzrError('%r is not in a branch' % orig_f)
103
46
######################################################################
106
class Branch(object):
107
50
"""Branch holding a history of revisions.
110
Base directory of the branch.
116
If _lock_mode is true, a positive count of the number of times the
120
Open file used for locking.
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.
126
def __init__(self, base, init=False, find_root=True):
65
def __init__(self, base, init=False):
127
66
"""Create new branch object at a particular location.
129
base -- Base directory for the branch.
131
init -- If True, create new control files in a previously
68
:param base: Base directory for the branch.
70
:param init: If True, create new control files in a previously
132
71
unversioned directory. If False, the branch must already
135
find_root -- If true and init is false, find the root of the
136
existing branch containing base.
138
74
In the test suite, creation of new trees is tested using the
139
75
`ScratchBranch` class.
77
self.base = os.path.realpath(base)
142
self.base = os.path.realpath(base)
143
79
self._make_control()
145
self.base = find_branch_root(base)
147
self.base = os.path.realpath(base)
148
81
if not isdir(self.controlfilename('.')):
149
from errors import NotBranchError
150
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
151
['use "bzr init" to initialize a new working tree',
152
'current bzr can only operate from top-of-tree'])
154
self._lockfile = self.controlfile('branch-lock', 'wb')
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'])
156
87
self.text_store = ImmutableStore(self.controlfilename('text-store'))
157
88
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
165
96
__repr__ = __str__
170
from warnings import warn
171
warn("branch %r was not explicitly unlocked" % self)
175
def lock(self, mode):
177
if mode == 'w' and cur_lm == 'r':
178
raise BzrError("can't upgrade to a write lock")
180
assert self._lock_count >= 1
181
self._lock_count += 1
183
from bzrlib.lock import lock, LOCK_SH, LOCK_EX
189
raise ValueError('invalid lock mode %r' % mode)
191
lock(self._lockfile, m)
192
self._lock_mode = mode
197
if not self._lock_mode:
198
raise BzrError('branch %r is not locked' % (self))
200
if self._lock_count > 1:
201
self._lock_count -= 1
203
assert self._lock_count == 1
204
from bzrlib.lock import unlock
205
unlock(self._lockfile)
206
self._lock_mode = self._lock_count = None
209
def abspath(self, name):
210
"""Return absolute filename for something in the branch"""
100
"""Return filename relative to branch top"""
211
101
return os.path.join(self.base, name)
214
def relpath(self, path):
215
"""Return path relative to this branch of something inside it.
217
Raises an error if path is not in this branch."""
218
rp = os.path.realpath(path)
220
if not rp.startswith(self.base):
221
from errors import NotBranchError
222
raise NotBranchError("path %r is not within branch %r" % (rp, self.base))
223
rp = rp[len(self.base):]
224
rp = rp.lstrip(os.sep)
228
104
def controlfilename(self, file_or_path):
229
105
"""Return location relative to branch."""
283
136
In the future, we might need different in-memory Branch
284
137
classes to support downlevel branches. But not yet.
286
# This ignores newlines so that we can open branches created
287
# on Windows from Linux and so on. I think it might be better
288
# to always make all internal files in unix format.
289
fmt = self.controlfile('branch-format', 'r').read()
290
fmt.replace('\r\n', '')
139
# read in binary mode to detect newline wierdness.
140
fmt = self.controlfile('branch-format', 'rb').read()
291
141
if fmt != BZR_BRANCH_FORMAT:
292
raise BzrError('sorry, branch format %r not supported' % fmt,
293
['use a different bzr version',
294
'or remove the .bzr directory and "bzr init" again'])
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'])
299
147
def read_working_inventory(self):
300
148
"""Read the working inventory."""
301
149
before = time.time()
302
# ElementTree does its own conversion from UTF-8, so open in
304
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
150
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
305
151
mutter("loaded inventory of %d items in %f"
306
152
% (len(inv), time.time() - before))
310
156
def _write_inventory(self, inv):
311
157
"""Update the working inventory.
313
159
That is to say, the inventory describing changes underway, that
314
160
will be committed to the next revision.
316
## TODO: factor out to atomicfile? is rename safe on windows?
317
## TODO: Maybe some kind of clean/dirty marker on inventory?
318
tmpfname = self.controlfilename('inventory.tmp')
319
tmpf = file(tmpfname, 'wb')
322
inv_fname = self.controlfilename('inventory')
323
if sys.platform == 'win32':
325
os.rename(tmpfname, inv_fname)
326
mutter('wrote working inventory')
162
inv.write_xml(self.controlfile('inventory', 'w'))
163
mutter('wrote inventory to %s' % quotefn(self.controlfilename('inventory')))
329
166
inventory = property(read_working_inventory, _write_inventory, None,
330
167
"""Inventory for the working copy.""")
334
def add(self, files, verbose=False, ids=None):
170
def add(self, files, verbose=False):
335
171
"""Make files versioned.
337
Note that the command line normally calls smart_add instead.
339
173
This puts the files in the Added state, so that they will be
340
174
recorded by the next commit.
342
TODO: Perhaps have an option to add the ids even if the files do
176
:todo: Perhaps have an option to add the ids even if the files do
345
TODO: Perhaps return the ids of the files? But then again it
179
:todo: Perhaps return the ids of the files? But then again it
346
180
is easy to retrieve them if they're needed.
348
TODO: Option to specify file id.
182
:todo: Option to specify file id.
350
TODO: Adding a directory should optionally recurse down and
184
:todo: Adding a directory should optionally recurse down and
351
185
add all non-ignored children. Perhaps do that in a
352
186
higher-level method.
188
>>> b = ScratchBranch(files=['foo'])
189
>>> 'foo' in b.unknowns()
194
>>> 'foo' in b.unknowns()
196
>>> bool(b.inventory.path2id('foo'))
202
Traceback (most recent call last):
204
BzrError: ('foo is already versioned', [])
206
>>> b.add(['nothere'])
207
Traceback (most recent call last):
208
BzrError: ('cannot add: not a regular file or directory: nothere', [])
354
211
# TODO: Re-adding a file that is removed in the working copy
355
212
# should probably put it back with the previous ID.
356
213
if isinstance(files, types.StringTypes):
357
assert(ids is None or isinstance(ids, types.StringTypes))
363
ids = [None] * len(files)
365
assert(len(ids) == len(files))
367
216
inv = self.read_working_inventory()
368
for f,file_id in zip(files, ids):
369
218
if is_control_file(f):
370
raise BzrError("cannot add control file %s" % quotefn(f))
219
bailout("cannot add control file %s" % quotefn(f))
372
221
fp = splitpath(f)
375
raise BzrError("cannot add top-level %r" % f)
377
fullpath = os.path.normpath(self.abspath(f))
380
kind = file_kind(fullpath)
382
# maybe something better?
383
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
385
if kind != 'file' and kind != 'directory':
386
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
389
file_id = gen_file_id(f)
390
inv.add_path(f, kind=kind, file_id=file_id)
224
bailout("cannot add top-level %r" % f)
226
fullpath = os.path.normpath(self._rel(f))
230
elif isdir(fullpath):
233
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
236
parent_name = joinpath(fp[:-1])
237
mutter("lookup parent %r" % parent_name)
238
parent_id = inv.path2id(parent_name)
239
if parent_id == None:
240
bailout("cannot add: parent %r is not versioned"
245
file_id = _gen_file_id(fp[-1])
246
inv.add(InventoryEntry(file_id, fp[-1], kind=kind, parent_id=parent_id))
393
248
show_status('A', kind, quotefn(f))
395
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
250
mutter("add file %s file_id:{%s} kind=%r parent_id={%s}"
251
% (f, file_id, kind, parent_id))
397
252
self._write_inventory(inv)
400
def print_file(self, file, revno):
401
"""Print `file` to stdout."""
402
tree = self.revision_tree(self.lookup_revision(revno))
403
# use inventory as it was in that revision
404
file_id = tree.inventory.path2id(file)
406
raise BzrError("%r is not present in revision %d" % (file, revno))
407
tree.print_file(file_id)
411
256
def remove(self, files, verbose=False):
412
257
"""Mark nominated files for removal from the inventory.
414
259
This does not remove their text. This does not run on
416
TODO: Refuse to remove modified files unless --force is given?
418
TODO: Do something useful with directories.
420
TODO: Should this remove the text or not? Tough call; not
261
:todo: Refuse to remove modified files unless --force is given?
263
>>> b = ScratchBranch(files=['foo'])
265
>>> b.inventory.has_filename('foo')
268
>>> b.working_tree().has_filename('foo')
270
>>> b.inventory.has_filename('foo')
273
>>> b = ScratchBranch(files=['foo'])
278
>>> b.inventory.has_filename('foo')
280
>>> b.basis_tree().has_filename('foo')
282
>>> b.working_tree().has_filename('foo')
285
:todo: Do something useful with directories.
287
:todo: Should this remove the text or not? Tough call; not
421
288
removing may be useful and the user can just use use rm, and
422
289
is the opposite of add. Removing it is consistent with most
423
290
other tools. Maybe an option.
425
292
## TODO: Normalize names
426
293
## TODO: Remove nested loops; better scalability
427
295
if isinstance(files, types.StringTypes):
430
tree = self.working_tree()
298
inv = self.read_working_inventory()
433
300
# do this before any modifications
435
302
fid = inv.path2id(f)
437
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
304
bailout("cannot remove unversioned file %s" % quotefn(f))
438
305
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
440
# having remove it, it must be either ignored or unknown
441
if tree.is_ignored(f):
445
show_status(new_status, inv[fid].kind, quotefn(f))
307
show_status('D', inv[fid].kind, quotefn(f))
448
310
self._write_inventory(inv)
451
def set_inventory(self, new_inventory_list):
453
for path, file_id, parent, kind in new_inventory_list:
454
name = os.path.basename(path)
457
inv.add(InventoryEntry(file_id, name, kind, parent))
458
self._write_inventory(inv)
461
313
def unknowns(self):
462
314
"""Return all unknown files.
477
329
return self.working_tree().unknowns()
480
def append_revision(self, revision_id):
481
mutter("add {%s} to revision-history" % revision_id)
482
rev_history = self.revision_history()
484
tmprhname = self.controlfilename('revision-history.tmp')
485
rhname = self.controlfilename('revision-history')
487
f = file(tmprhname, 'wt')
488
rev_history.append(revision_id)
489
f.write('\n'.join(rev_history))
493
if sys.platform == 'win32':
495
os.rename(tmprhname, rhname)
332
def commit(self, message, timestamp=None, committer=None,
334
"""Commit working copy as a new revision.
336
The basic approach is to add all the file texts into the
337
store, then the inventory, then make a new revision pointing
338
to that inventory and store that.
340
This is not quite safe if the working copy changes during the
341
commit; for the moment that is simply not allowed. A better
342
approach is to make a temporary copy of the files before
343
computing their hashes, and then add those hashes in turn to
344
the inventory. This should mean at least that there are no
345
broken hash pointers. There is no way we can get a snapshot
346
of the whole directory at an instant. This would also have to
347
be robust against files disappearing, moving, etc. So the
348
whole thing is a bit hard.
350
:param timestamp: if not None, seconds-since-epoch for a
351
postdated/predated commit.
354
## TODO: Show branch names
356
# TODO: Don't commit if there are no changes, unless forced?
358
# First walk over the working inventory; and both update that
359
# and also build a new revision inventory. The revision
360
# inventory needs to hold the text-id, sha1 and size of the
361
# actual file versions committed in the revision. (These are
362
# not present in the working inventory.) We also need to
363
# detect missing/deleted files, and remove them from the
366
work_inv = self.read_working_inventory()
368
basis = self.basis_tree()
369
basis_inv = basis.inventory
371
for path, entry in work_inv.iter_entries():
372
## TODO: Cope with files that have gone missing.
374
## TODO: Check that the file kind has not changed from the previous
375
## revision of this file (if any).
380
file_id = entry.file_id
381
mutter('commit prep file %s, id %r ' % (p, file_id))
383
if not os.path.exists(p):
384
mutter(" file is missing, removing from inventory")
386
show_status('D', entry.kind, quotefn(path))
387
missing_ids.append(file_id)
390
# TODO: Handle files that have been deleted
392
# TODO: Maybe a special case for empty files? Seems a
393
# waste to store them many times.
397
if basis_inv.has_id(file_id):
398
old_kind = basis_inv[file_id].kind
399
if old_kind != entry.kind:
400
bailout("entry %r changed kind from %r to %r"
401
% (file_id, old_kind, entry.kind))
403
if entry.kind == 'directory':
405
bailout("%s is entered as directory but not a directory" % quotefn(p))
406
elif entry.kind == 'file':
408
bailout("%s is entered as file but is not a file" % quotefn(p))
410
content = file(p, 'rb').read()
412
entry.text_sha1 = sha_string(content)
413
entry.text_size = len(content)
415
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
417
and (old_ie.text_size == entry.text_size)
418
and (old_ie.text_sha1 == entry.text_sha1)):
419
## assert content == basis.get_file(file_id).read()
420
entry.text_id = basis_inv[file_id].text_id
421
mutter(' unchanged from previous text_id {%s}' %
425
entry.text_id = _gen_file_id(entry.name)
426
self.text_store.add(content, entry.text_id)
427
mutter(' stored with text_id {%s}' % entry.text_id)
431
elif (old_ie.name == entry.name
432
and old_ie.parent_id == entry.parent_id):
437
show_status(state, entry.kind, quotefn(path))
439
for file_id in missing_ids:
440
# have to do this later so we don't mess up the iterator.
441
# since parents may be removed before their children we
444
# FIXME: There's probably a better way to do this; perhaps
445
# the workingtree should know how to filter itself.
446
if work_inv.has_id(file_id):
447
del work_inv[file_id]
450
inv_id = rev_id = _gen_revision_id(time.time())
452
inv_tmp = tempfile.TemporaryFile()
453
inv.write_xml(inv_tmp)
455
self.inventory_store.add(inv_tmp, inv_id)
456
mutter('new inventory_id is {%s}' % inv_id)
458
self._write_inventory(work_inv)
460
if timestamp == None:
461
timestamp = time.time()
463
if committer == None:
464
committer = username()
466
mutter("building commit log message")
467
rev = Revision(timestamp=timestamp,
469
precursor = self.last_patch(),
474
rev_tmp = tempfile.TemporaryFile()
475
rev.write_xml(rev_tmp)
477
self.revision_store.add(rev_tmp, rev_id)
478
mutter("new revision_id is {%s}" % rev_id)
480
## XXX: Everything up to here can simply be orphaned if we abort
481
## the commit; it will leave junk files behind but that doesn't
484
## TODO: Read back the just-generated changeset, and make sure it
485
## applies and recreates the right state.
487
## TODO: Also calculate and store the inventory SHA1
488
mutter("committing patch r%d" % (self.revno() + 1))
490
mutter("append to revision-history")
491
self.controlfile('revision-history', 'at').write(rev_id + '\n')
499
496
def get_revision(self, revision_id):
627
def rename_one(self, from_rel, to_rel):
630
This can change the directory or the filename or both.
632
tree = self.working_tree()
634
if not tree.has_filename(from_rel):
635
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
636
if tree.has_filename(to_rel):
637
raise BzrError("can't rename: new working file %r already exists" % to_rel)
639
file_id = inv.path2id(from_rel)
641
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
643
if inv.path2id(to_rel):
644
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
646
to_dir, to_tail = os.path.split(to_rel)
647
to_dir_id = inv.path2id(to_dir)
648
if to_dir_id == None and to_dir != '':
649
raise BzrError("can't determine destination directory id for %r" % to_dir)
651
mutter("rename_one:")
652
mutter(" file_id {%s}" % file_id)
653
mutter(" from_rel %r" % from_rel)
654
mutter(" to_rel %r" % to_rel)
655
mutter(" to_dir %r" % to_dir)
656
mutter(" to_dir_id {%s}" % to_dir_id)
658
inv.rename(file_id, to_dir_id, to_tail)
660
print "%s => %s" % (from_rel, to_rel)
662
from_abs = self.abspath(from_rel)
663
to_abs = self.abspath(to_rel)
665
os.rename(from_abs, to_abs)
667
raise BzrError("failed to rename %r to %r: %s"
668
% (from_abs, to_abs, e[1]),
669
["rename rolled back"])
671
self._write_inventory(inv)
676
def move(self, from_paths, to_name):
679
to_name must exist as a versioned directory.
681
If to_name exists and is a directory, the files are moved into
682
it, keeping their old names. If it is a directory,
684
Note that to_name is only the last component of the new name;
685
this doesn't change the directory.
687
## TODO: Option to move IDs only
688
assert not isinstance(from_paths, basestring)
689
tree = self.working_tree()
691
to_abs = self.abspath(to_name)
692
if not isdir(to_abs):
693
raise BzrError("destination %r is not a directory" % to_abs)
694
if not tree.has_filename(to_name):
695
raise BzrError("destination %r not in working directory" % to_abs)
696
to_dir_id = inv.path2id(to_name)
697
if to_dir_id == None and to_name != '':
698
raise BzrError("destination %r is not a versioned directory" % to_name)
699
to_dir_ie = inv[to_dir_id]
700
if to_dir_ie.kind not in ('directory', 'root_directory'):
701
raise BzrError("destination %r is not a directory" % to_abs)
703
to_idpath = inv.get_idpath(to_dir_id)
706
if not tree.has_filename(f):
707
raise BzrError("%r does not exist in working tree" % f)
708
f_id = inv.path2id(f)
710
raise BzrError("%r is not versioned" % f)
711
name_tail = splitpath(f)[-1]
712
dest_path = appendpath(to_name, name_tail)
713
if tree.has_filename(dest_path):
714
raise BzrError("destination %r already exists" % dest_path)
715
if f_id in to_idpath:
716
raise BzrError("can't move %r to a subdirectory of itself" % f)
718
# OK, so there's a race here, it's possible that someone will
719
# create a file in this interval and then the rename might be
720
# left half-done. But we should have caught most problems.
723
name_tail = splitpath(f)[-1]
724
dest_path = appendpath(to_name, name_tail)
725
print "%s => %s" % (f, dest_path)
726
inv.rename(inv.path2id(f), to_dir_id, name_tail)
728
os.rename(self.abspath(f), self.abspath(dest_path))
730
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
731
["rename rolled back"])
733
self._write_inventory(inv)
610
def write_log(self, utc=False):
611
"""Write out human-readable log of commits to this branch
613
:param utc: If true, show dates in universal time, not local time."""
616
for p in self.revision_history():
618
print 'revno:', revno
619
## TODO: Show hash if --id is given.
620
##print 'revision-hash:', p
621
rev = self.get_revision(p)
622
print 'committer:', rev.committer
623
print 'timestamp: %s' % (format_date(rev.timestamp, utc))
625
## opportunistic consistency check, same as check_patch_chaining
626
if rev.precursor != precursor:
627
bailout("mismatched precursor!")
631
print ' (no message)'
633
for l in rev.message.split('\n'):
641
def show_status(branch, show_all=False):
642
"""Display single-line status for non-ignored working files.
644
The list is show sorted in order by file name.
646
>>> b = ScratchBranch(files=['foo', 'foo~'])
652
>>> b.commit("add foo")
655
:todo: Get state for single files.
657
:todo: Perhaps show a slash at the end of directory names.
661
# We have to build everything into a list first so that it can
662
# sorted by name, incorporating all the different sources.
664
# FIXME: Rather than getting things in random order and then sorting,
665
# just step through in order.
667
# Interesting case: the old ID for a file has been removed,
668
# but a new file has been created under that name.
670
old = branch.basis_tree()
671
old_inv = old.inventory
672
new = branch.working_tree()
673
new_inv = new.inventory
675
for fs, fid, oldname, newname, kind in diff_trees(old, new):
677
show_status(fs, kind,
678
oldname + ' => ' + newname)
679
elif fs == 'A' or fs == 'M':
680
show_status(fs, kind, newname)
682
show_status(fs, kind, oldname)
685
show_status(fs, kind, newname)
688
show_status(fs, kind, newname)
690
show_status(fs, kind, newname)
692
bailout("wierd file state %r" % ((fs, fid),))
738
696
class ScratchBranch(Branch):