16
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
22
import traceback, socket, fnmatch, difflib, time
23
from binascii import hexlify
26
from inventory import Inventory
27
from trace import mutter, note
28
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
29
from inventory import InventoryEntry, Inventory
30
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, chomp, \
31
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
32
joinpath, sha_string, file_kind, local_time_offset
33
from store import ImmutableStore
34
from revision import Revision
35
from errors import bailout
36
from textui import show_status
37
from diff import diff_trees
22
from bzrlib.trace import mutter, note
23
from bzrlib.osutils import isdir, quotefn, compact_date, rand_bytes, \
25
sha_file, appendpath, file_kind
27
from bzrlib.errors import BzrError, InvalidRevisionNumber, InvalidRevisionId, \
28
DivergedBranches, NotBranchError
29
from bzrlib.textui import show_status
30
from bzrlib.revision import Revision
31
from bzrlib.delta import compare_trees
32
from bzrlib.tree import EmptyTree, RevisionTree
39
38
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
40
39
## TODO: Maybe include checks for common corruption of newlines, etc?
42
# TODO: Some operations like log might retrieve the same revisions
43
# repeatedly to calculate deltas. We could perhaps have a weakref
44
# cache in memory to make this faster.
46
# TODO: please move the revision-string syntax stuff out of the branch
47
# object; it's clutter
50
def find_branch(f, **args):
51
if f and (f.startswith('http://') or f.startswith('https://')):
52
from bzrlib.remotebranch import RemoteBranch
53
return RemoteBranch(f, **args)
55
return Branch(f, **args)
58
def find_cached_branch(f, cache_root, **args):
59
from bzrlib.remotebranch import RemoteBranch
60
br = find_branch(f, **args)
61
def cacheify(br, store_name):
62
from bzrlib.meta_store import CachedStore
63
cache_path = os.path.join(cache_root, store_name)
65
new_store = CachedStore(getattr(br, store_name), cache_path)
66
setattr(br, store_name, new_store)
68
if isinstance(br, RemoteBranch):
69
cacheify(br, 'inventory_store')
70
cacheify(br, 'text_store')
71
cacheify(br, 'revision_store')
75
def _relpath(base, path):
76
"""Return path relative to base, or raise exception.
78
The path may be either an absolute path or a path relative to the
79
current working directory.
81
Lifted out of Branch.relpath for ease of testing.
83
os.path.commonprefix (python2.4) has a bad bug that it works just
84
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
85
avoids that problem."""
86
rp = os.path.abspath(path)
90
while len(head) >= len(base):
93
head, tail = os.path.split(head)
97
raise NotBranchError("path %r is not within branch %r" % (rp, base))
102
def find_branch_root(f=None):
103
"""Find the branch root enclosing f, or pwd.
105
f may be a filename or a URL.
107
It is not necessary that f exists.
109
Basically we keep looking up until we find the control directory or
110
run into the root. If there isn't one, raises NotBranchError.
114
elif hasattr(os.path, 'realpath'):
115
f = os.path.realpath(f)
117
f = os.path.abspath(f)
118
if not os.path.exists(f):
119
raise BzrError('%r does not exist' % f)
125
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
127
head, tail = os.path.split(f)
129
# reached the root, whatever that may be
130
raise NotBranchError('%s is not in a branch' % orig_f)
46
136
######################################################################
139
class Branch(object):
50
140
"""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.
143
Base directory of the branch.
149
If _lock_mode is true, a positive count of the number of times the
153
Lock object from bzrlib.lock.
65
def __init__(self, base, init=False):
160
# Map some sort of prefix into a namespace
161
# stuff like "revno:10", "revid:", etc.
162
# This should match a prefix with a function which accepts
163
REVISION_NAMESPACES = {}
165
def __init__(self, base, init=False, find_root=True):
66
166
"""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
168
base -- Base directory for the branch.
170
init -- If True, create new control files in a previously
71
171
unversioned directory. If False, the branch must already
174
find_root -- If true and init is false, find the root of the
175
existing branch containing base.
74
177
In the test suite, creation of new trees is tested using the
75
178
`ScratchBranch` class.
77
self.base = os.path.realpath(base)
180
from bzrlib.store import ImmutableStore
182
self.base = os.path.realpath(base)
79
183
self._make_control()
185
self.base = find_branch_root(base)
187
self.base = os.path.realpath(base)
81
188
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'])
189
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
190
['use "bzr init" to initialize a new working tree',
191
'current bzr can only operate from top-of-tree'])
87
194
self.text_store = ImmutableStore(self.controlfilename('text-store'))
88
195
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
96
203
__repr__ = __str__
100
"""Return filename relative to branch top"""
207
if self._lock_mode or self._lock:
208
from bzrlib.warnings import warn
209
warn("branch %r was not explicitly unlocked" % self)
213
def lock_write(self):
215
if self._lock_mode != 'w':
216
from bzrlib.errors import LockError
217
raise LockError("can't upgrade to a write lock from %r" %
219
self._lock_count += 1
221
from bzrlib.lock import WriteLock
223
self._lock = WriteLock(self.controlfilename('branch-lock'))
224
self._lock_mode = 'w'
230
assert self._lock_mode in ('r', 'w'), \
231
"invalid lock mode %r" % self._lock_mode
232
self._lock_count += 1
234
from bzrlib.lock import ReadLock
236
self._lock = ReadLock(self.controlfilename('branch-lock'))
237
self._lock_mode = 'r'
241
if not self._lock_mode:
242
from bzrlib.errors import LockError
243
raise LockError('branch %r is not locked' % (self))
245
if self._lock_count > 1:
246
self._lock_count -= 1
250
self._lock_mode = self._lock_count = None
252
def abspath(self, name):
253
"""Return absolute filename for something in the branch"""
101
254
return os.path.join(self.base, name)
256
def relpath(self, path):
257
"""Return path relative to this branch of something inside it.
259
Raises an error if path is not in this branch."""
260
return _relpath(self.base, path)
104
262
def controlfilename(self, file_or_path):
105
263
"""Return location relative to branch."""
106
if isinstance(file_or_path, types.StringTypes):
264
if isinstance(file_or_path, basestring):
107
265
file_or_path = [file_or_path]
108
266
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
111
269
def controlfile(self, file_or_path, mode='r'):
112
"""Open a control file for this branch"""
113
return file(self.controlfilename(file_or_path), mode)
270
"""Open a control file for this branch.
272
There are two classes of file in the control directory: text
273
and binary. binary files are untranslated byte streams. Text
274
control files are stored with Unix newlines and in UTF-8, even
275
if the platform or locale defaults are different.
277
Controlfiles should almost never be opened in write mode but
278
rather should be atomically copied and replaced using atomicfile.
281
fn = self.controlfilename(file_or_path)
283
if mode == 'rb' or mode == 'wb':
284
return file(fn, mode)
285
elif mode == 'r' or mode == 'w':
286
# open in binary mode anyhow so there's no newline translation;
287
# codecs uses line buffering by default; don't want that.
289
return codecs.open(fn, mode + 'b', 'utf-8',
292
raise BzrError("invalid controlfile mode %r" % mode)
116
294
def _make_control(self):
295
from bzrlib.inventory import Inventory
117
297
os.mkdir(self.controlfilename([]))
118
298
self.controlfile('README', 'w').write(
119
299
"This is a Bazaar-NG control directory.\n"
120
"Do not change any files in this directory.")
300
"Do not change any files in this directory.\n")
121
301
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
122
302
for d in ('text-store', 'inventory-store', 'revision-store'):
123
303
os.mkdir(self.controlfilename(d))
124
304
for f in ('revision-history', 'merged-patches',
125
'pending-merged-patches', 'branch-name'):
305
'pending-merged-patches', 'branch-name',
126
308
self.controlfile(f, 'w').write('')
127
309
mutter('created control directory in ' + self.base)
128
Inventory().write_xml(self.controlfile('inventory','w'))
311
# if we want per-tree root ids then this is the place to set
312
# them; they're not needed for now and so ommitted for
314
f = self.controlfile('inventory','w')
315
bzrlib.xml.serializer_v4.write_inventory(Inventory(), f)
131
318
def _check_format(self):
136
323
In the future, we might need different in-memory Branch
137
324
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()
326
# This ignores newlines so that we can open branches created
327
# on Windows from Linux and so on. I think it might be better
328
# to always make all internal files in unix format.
329
fmt = self.controlfile('branch-format', 'r').read()
330
fmt = fmt.replace('\r\n', '\n')
141
331
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'])
332
raise BzrError('sorry, branch format %r not supported' % fmt,
333
['use a different bzr version',
334
'or remove the .bzr directory and "bzr init" again'])
336
def get_root_id(self):
337
"""Return the id of this branches root"""
338
inv = self.read_working_inventory()
339
return inv.root.file_id
341
def set_root_id(self, file_id):
342
inv = self.read_working_inventory()
343
orig_root_id = inv.root.file_id
344
del inv._byid[inv.root.file_id]
345
inv.root.file_id = file_id
346
inv._byid[inv.root.file_id] = inv.root
349
if entry.parent_id in (None, orig_root_id):
350
entry.parent_id = inv.root.file_id
351
self._write_inventory(inv)
147
353
def read_working_inventory(self):
148
354
"""Read the working inventory."""
150
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
151
mutter("loaded inventory of %d items in %f"
152
% (len(inv), time.time() - before))
355
from bzrlib.inventory import Inventory
358
# ElementTree does its own conversion from UTF-8, so open in
360
f = self.controlfile('inventory', 'rb')
361
return bzrlib.xml.serializer_v4.read_inventory(f)
156
366
def _write_inventory(self, inv):
157
367
"""Update the working inventory.
159
369
That is to say, the inventory describing changes underway, that
160
370
will be committed to the next revision.
162
## TODO: factor out to atomicfile? is rename safe on windows?
163
tmpfname = self.controlfilename('inventory.tmp')
164
tmpf = file(tmpfname, 'w')
167
os.rename(tmpfname, self.controlfilename('inventory'))
372
from bzrlib.atomicfile import AtomicFile
376
f = AtomicFile(self.controlfilename('inventory'), 'wb')
378
bzrlib.xml.serializer_v4.write_inventory(inv, f)
168
385
mutter('wrote working inventory')
171
388
inventory = property(read_working_inventory, _write_inventory, None,
172
389
"""Inventory for the working copy.""")
175
def add(self, files, verbose=False):
392
def add(self, files, ids=None):
176
393
"""Make files versioned.
395
Note that the command line normally calls smart_add instead,
396
which can automatically recurse.
178
398
This puts the files in the Added state, so that they will be
179
399
recorded by the next commit.
181
: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
185
is easy to retrieve them if they're needed.
187
:todo: Option to specify file id.
189
:todo: Adding a directory should optionally recurse down and
190
add all non-ignored children. Perhaps do that in a
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', [])
402
List of paths to add, relative to the base of the tree.
405
If set, use these instead of automatically generated ids.
406
Must be the same length as the list of files, but may
407
contain None for ids that are to be autogenerated.
409
TODO: Perhaps have an option to add the ids even if the files do
412
TODO: Perhaps yield the ids and paths as they're added.
216
414
# TODO: Re-adding a file that is removed in the working copy
217
415
# should probably put it back with the previous ID.
218
if isinstance(files, types.StringTypes):
416
if isinstance(files, basestring):
417
assert(ids is None or isinstance(ids, basestring))
221
inv = self.read_working_inventory()
223
if is_control_file(f):
224
bailout("cannot add control file %s" % quotefn(f))
229
bailout("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))
253
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))
257
self._write_inventory(inv)
423
ids = [None] * len(files)
425
assert(len(ids) == len(files))
429
inv = self.read_working_inventory()
430
for f,file_id in zip(files, ids):
431
if is_control_file(f):
432
raise BzrError("cannot add control file %s" % quotefn(f))
437
raise BzrError("cannot add top-level %r" % f)
439
fullpath = os.path.normpath(self.abspath(f))
442
kind = file_kind(fullpath)
444
# maybe something better?
445
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
447
if kind != 'file' and kind != 'directory':
448
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
451
file_id = gen_file_id(f)
452
inv.add_path(f, kind=kind, file_id=file_id)
454
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
456
self._write_inventory(inv)
461
def print_file(self, file, revno):
462
"""Print `file` to stdout."""
465
tree = self.revision_tree(self.lookup_revision(revno))
466
# use inventory as it was in that revision
467
file_id = tree.inventory.path2id(file)
469
raise BzrError("%r is not present in revision %s" % (file, revno))
470
tree.print_file(file_id)
261
475
def remove(self, files, verbose=False):
264
478
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
480
TODO: Refuse to remove modified files unless --force is given?
482
TODO: Do something useful with directories.
484
TODO: Should this remove the text or not? Tough call; not
293
485
removing may be useful and the user can just use use rm, and
294
486
is the opposite of add. Removing it is consistent with most
295
487
other tools. Maybe an option.
297
489
## TODO: Normalize names
298
490
## TODO: Remove nested loops; better scalability
300
if isinstance(files, types.StringTypes):
491
if isinstance(files, basestring):
303
tree = self.working_tree()
306
# do this before any modifications
310
bailout("cannot remove unversioned file %s" % quotefn(f))
311
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
313
# having remove it, it must be either ignored or unknown
314
if tree.is_ignored(f):
318
show_status(new_status, inv[fid].kind, quotefn(f))
497
tree = self.working_tree()
500
# do this before any modifications
504
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
505
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
507
# having remove it, it must be either ignored or unknown
508
if tree.is_ignored(f):
512
show_status(new_status, inv[fid].kind, quotefn(f))
515
self._write_inventory(inv)
520
# FIXME: this doesn't need to be a branch method
521
def set_inventory(self, new_inventory_list):
522
from bzrlib.inventory import Inventory, InventoryEntry
523
inv = Inventory(self.get_root_id())
524
for path, file_id, parent, kind in new_inventory_list:
525
name = os.path.basename(path)
528
inv.add(InventoryEntry(file_id, name, kind, parent))
321
529
self._write_inventory(inv)
340
548
return self.working_tree().unknowns()
343
def commit(self, message, timestamp=None, timezone=None,
346
"""Commit working copy as a new revision.
348
The basic approach is to add all the file texts into the
349
store, then the inventory, then make a new revision pointing
350
to that inventory and store that.
352
This is not quite safe if the working copy changes during the
353
commit; for the moment that is simply not allowed. A better
354
approach is to make a temporary copy of the files before
355
computing their hashes, and then add those hashes in turn to
356
the inventory. This should mean at least that there are no
357
broken hash pointers. There is no way we can get a snapshot
358
of the whole directory at an instant. This would also have to
359
be robust against files disappearing, moving, etc. So the
360
whole thing is a bit hard.
362
:param timestamp: if not None, seconds-since-epoch for a
363
postdated/predated commit.
366
## TODO: Show branch names
368
# TODO: Don't commit if there are no changes, unless forced?
370
# First walk over the working inventory; and both update that
371
# and also build a new revision inventory. The revision
372
# inventory needs to hold the text-id, sha1 and size of the
373
# actual file versions committed in the revision. (These are
374
# not present in the working inventory.) We also need to
375
# detect missing/deleted files, and remove them from the
378
work_inv = self.read_working_inventory()
380
basis = self.basis_tree()
381
basis_inv = basis.inventory
383
for path, entry in work_inv.iter_entries():
384
## TODO: Cope with files that have gone missing.
386
## TODO: Check that the file kind has not changed from the previous
387
## revision of this file (if any).
392
file_id = entry.file_id
393
mutter('commit prep file %s, id %r ' % (p, file_id))
395
if not os.path.exists(p):
396
mutter(" file is missing, removing from inventory")
398
show_status('D', entry.kind, quotefn(path))
399
missing_ids.append(file_id)
402
# TODO: Handle files that have been deleted
404
# TODO: Maybe a special case for empty files? Seems a
405
# waste to store them many times.
409
if basis_inv.has_id(file_id):
410
old_kind = basis_inv[file_id].kind
411
if old_kind != entry.kind:
412
bailout("entry %r changed kind from %r to %r"
413
% (file_id, old_kind, entry.kind))
415
if entry.kind == 'directory':
417
bailout("%s is entered as directory but not a directory" % quotefn(p))
418
elif entry.kind == 'file':
420
bailout("%s is entered as file but is not a file" % quotefn(p))
422
content = file(p, 'rb').read()
424
entry.text_sha1 = sha_string(content)
425
entry.text_size = len(content)
427
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
429
and (old_ie.text_size == entry.text_size)
430
and (old_ie.text_sha1 == entry.text_sha1)):
431
## assert content == basis.get_file(file_id).read()
432
entry.text_id = basis_inv[file_id].text_id
433
mutter(' unchanged from previous text_id {%s}' %
437
entry.text_id = _gen_file_id(entry.name)
438
self.text_store.add(content, entry.text_id)
439
mutter(' stored with text_id {%s}' % entry.text_id)
443
elif (old_ie.name == entry.name
444
and old_ie.parent_id == entry.parent_id):
449
show_status(state, entry.kind, quotefn(path))
451
for file_id in missing_ids:
452
# have to do this later so we don't mess up the iterator.
453
# since parents may be removed before their children we
456
# FIXME: There's probably a better way to do this; perhaps
457
# the workingtree should know how to filter itself.
458
if work_inv.has_id(file_id):
459
del work_inv[file_id]
462
inv_id = rev_id = _gen_revision_id(time.time())
464
inv_tmp = tempfile.TemporaryFile()
465
inv.write_xml(inv_tmp)
467
self.inventory_store.add(inv_tmp, inv_id)
468
mutter('new inventory_id is {%s}' % inv_id)
470
self._write_inventory(work_inv)
472
if timestamp == None:
473
timestamp = time.time()
475
if committer == None:
476
committer = username()
479
timezone = local_time_offset()
481
mutter("building commit log message")
482
rev = Revision(timestamp=timestamp,
485
precursor = self.last_patch(),
490
rev_tmp = tempfile.TemporaryFile()
491
rev.write_xml(rev_tmp)
493
self.revision_store.add(rev_tmp, rev_id)
494
mutter("new revision_id is {%s}" % rev_id)
496
## XXX: Everything up to here can simply be orphaned if we abort
497
## the commit; it will leave junk files behind but that doesn't
500
## TODO: Read back the just-generated changeset, and make sure it
501
## applies and recreates the right state.
503
## TODO: Also calculate and store the inventory SHA1
504
mutter("committing patch r%d" % (self.revno() + 1))
506
mutter("append to revision-history")
507
self.controlfile('revision-history', 'at').write(rev_id + '\n')
551
def append_revision(self, *revision_ids):
552
from bzrlib.atomicfile import AtomicFile
554
for revision_id in revision_ids:
555
mutter("add {%s} to revision-history" % revision_id)
557
rev_history = self.revision_history()
558
rev_history.extend(revision_ids)
560
f = AtomicFile(self.controlfilename('revision-history'))
562
for rev_id in rev_history:
569
def get_revision_xml_file(self, revision_id):
570
"""Return XML file object for revision object."""
571
if not revision_id or not isinstance(revision_id, basestring):
572
raise InvalidRevisionId(revision_id)
577
return self.revision_store[revision_id]
579
raise bzrlib.errors.NoSuchRevision(self, revision_id)
585
get_revision_xml = get_revision_xml_file
512
588
def get_revision(self, revision_id):
513
589
"""Return the Revision object for a named revision"""
514
r = Revision.read_xml(self.revision_store[revision_id])
590
xml_file = self.get_revision_xml_file(revision_id)
593
r = bzrlib.xml.serializer_v4.read_revision(xml_file)
594
except SyntaxError, e:
595
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
515
599
assert r.revision_id == revision_id
603
def get_revision_delta(self, revno):
604
"""Return the delta for one revision.
606
The delta is relative to its mainline predecessor, or the
607
empty tree for revision 1.
609
assert isinstance(revno, int)
610
rh = self.revision_history()
611
if not (1 <= revno <= len(rh)):
612
raise InvalidRevisionNumber(revno)
614
# revno is 1-based; list is 0-based
616
new_tree = self.revision_tree(rh[revno-1])
618
old_tree = EmptyTree()
620
old_tree = self.revision_tree(rh[revno-2])
622
return compare_trees(old_tree, new_tree)
626
def get_revision_sha1(self, revision_id):
627
"""Hash the stored value of a revision, and return it."""
628
# In the future, revision entries will be signed. At that
629
# point, it is probably best *not* to include the signature
630
# in the revision hash. Because that lets you re-sign
631
# the revision, (add signatures/remove signatures) and still
632
# have all hash pointers stay consistent.
633
# But for now, just hash the contents.
634
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
519
637
def get_inventory(self, inventory_id):
520
638
"""Get Inventory object by hash.
522
:todo: Perhaps for this and similar methods, take a revision
640
TODO: Perhaps for this and similar methods, take a revision
523
641
parameter which can be either an integer revno or a
525
i = Inventory.read_xml(self.inventory_store[inventory_id])
643
from bzrlib.inventory import Inventory
645
f = self.get_inventory_xml_file(inventory_id)
646
return bzrlib.xml.serializer_v4.read_inventory(f)
649
def get_inventory_xml(self, inventory_id):
650
"""Get inventory XML as a file object."""
651
return self.inventory_store[inventory_id]
653
get_inventory_xml_file = get_inventory_xml
656
def get_inventory_sha1(self, inventory_id):
657
"""Return the sha1 hash of the inventory entry
659
return sha_file(self.get_inventory_xml(inventory_id))
529
662
def get_revision_inventory(self, revision_id):
530
663
"""Return inventory of a past revision."""
664
# bzr 0.0.6 imposes the constraint that the inventory_id
665
# must be the same as its revision, so this is trivial.
531
666
if revision_id == None:
667
from bzrlib.inventory import Inventory
668
return Inventory(self.get_root_id())
534
return self.get_inventory(self.get_revision(revision_id).inventory_id)
670
return self.get_inventory(revision_id)
537
673
def revision_history(self):
549
735
That is equivalent to the number of revisions committed to
552
>>> b = ScratchBranch()
555
>>> b.commit('no foo')
559
738
return len(self.revision_history())
562
741
def last_patch(self):
563
742
"""Return last patch hash, or None if no history.
565
>>> ScratchBranch().last_patch() == None
568
744
ph = self.revision_history()
573
def lookup_revision(self, revno):
574
"""Return revision hash for revision number."""
751
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
753
If self and other have not diverged, return a list of the revisions
754
present in other, but missing from self.
756
>>> from bzrlib.commit import commit
757
>>> bzrlib.trace.silent = True
758
>>> br1 = ScratchBranch()
759
>>> br2 = ScratchBranch()
760
>>> br1.missing_revisions(br2)
762
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
763
>>> br1.missing_revisions(br2)
765
>>> br2.missing_revisions(br1)
767
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
768
>>> br1.missing_revisions(br2)
770
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
771
>>> br1.missing_revisions(br2)
773
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
774
>>> br1.missing_revisions(br2)
775
Traceback (most recent call last):
776
DivergedBranches: These branches have diverged.
778
self_history = self.revision_history()
779
self_len = len(self_history)
780
other_history = other.revision_history()
781
other_len = len(other_history)
782
common_index = min(self_len, other_len) -1
783
if common_index >= 0 and \
784
self_history[common_index] != other_history[common_index]:
785
raise DivergedBranches(self, other)
787
if stop_revision is None:
788
stop_revision = other_len
789
elif stop_revision > other_len:
790
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
792
return other_history[self_len:stop_revision]
795
def update_revisions(self, other, stop_revision=None):
796
"""Pull in all new revisions from other branch.
798
from bzrlib.fetch import greedy_fetch
800
pb = bzrlib.ui.ui_factory.progress_bar()
801
pb.update('comparing histories')
803
revision_ids = self.missing_revisions(other, stop_revision)
805
if len(revision_ids) > 0:
806
count = greedy_fetch(self, other, revision_ids[-1], pb)[0]
809
self.append_revision(*revision_ids)
810
## note("Added %d revisions." % count)
813
def install_revisions(self, other, revision_ids, pb):
814
if hasattr(other.revision_store, "prefetch"):
815
other.revision_store.prefetch(revision_ids)
816
if hasattr(other.inventory_store, "prefetch"):
817
inventory_ids = [other.get_revision(r).inventory_id
818
for r in revision_ids]
819
other.inventory_store.prefetch(inventory_ids)
822
pb = bzrlib.ui.ui_factory.progress_bar()
829
for i, rev_id in enumerate(revision_ids):
830
pb.update('fetching revision', i+1, len(revision_ids))
832
rev = other.get_revision(rev_id)
833
except bzrlib.errors.NoSuchRevision:
837
revisions.append(rev)
838
inv = other.get_inventory(str(rev.inventory_id))
839
for key, entry in inv.iter_entries():
840
if entry.text_id is None:
842
if entry.text_id not in self.text_store:
843
needed_texts.add(entry.text_id)
847
count, cp_fail = self.text_store.copy_multi(other.text_store,
849
#print "Added %d texts." % count
850
inventory_ids = [ f.inventory_id for f in revisions ]
851
count, cp_fail = self.inventory_store.copy_multi(other.inventory_store,
853
#print "Added %d inventories." % count
854
revision_ids = [ f.revision_id for f in revisions]
856
count, cp_fail = self.revision_store.copy_multi(other.revision_store,
859
assert len(cp_fail) == 0
860
return count, failures
863
def commit(self, *args, **kw):
864
from bzrlib.commit import commit
865
commit(self, *args, **kw)
868
def lookup_revision(self, revision):
869
"""Return the revision identifier for a given revision information."""
870
revno, info = self._get_revision_info(revision)
874
def revision_id_to_revno(self, revision_id):
875
"""Given a revision id, return its revno"""
876
history = self.revision_history()
878
return history.index(revision_id) + 1
880
raise bzrlib.errors.NoSuchRevision(self, revision_id)
883
def get_revision_info(self, revision):
884
"""Return (revno, revision id) for revision identifier.
886
revision can be an integer, in which case it is assumed to be revno (though
887
this will translate negative values into positive ones)
888
revision can also be a string, in which case it is parsed for something like
889
'date:' or 'revid:' etc.
891
revno, rev_id = self._get_revision_info(revision)
893
raise bzrlib.errors.NoSuchRevision(self, revision)
896
def get_rev_id(self, revno, history=None):
897
"""Find the revision id of the specified revno."""
579
# list is 0-based; revisions are 1-based
580
return self.revision_history()[revno-1]
582
bailout("no such revision %s" % revno)
901
history = self.revision_history()
902
elif revno <= 0 or revno > len(history):
903
raise bzrlib.errors.NoSuchRevision(self, revno)
904
return history[revno - 1]
906
def _get_revision_info(self, revision):
907
"""Return (revno, revision id) for revision specifier.
909
revision can be an integer, in which case it is assumed to be revno
910
(though this will translate negative values into positive ones)
911
revision can also be a string, in which case it is parsed for something
912
like 'date:' or 'revid:' etc.
914
A revid is always returned. If it is None, the specifier referred to
915
the null revision. If the revid does not occur in the revision
916
history, revno will be None.
922
try:# Convert to int if possible
923
revision = int(revision)
926
revs = self.revision_history()
927
if isinstance(revision, int):
929
revno = len(revs) + revision + 1
932
rev_id = self.get_rev_id(revno, revs)
933
elif isinstance(revision, basestring):
934
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
935
if revision.startswith(prefix):
936
result = func(self, revs, revision)
938
revno, rev_id = result
941
rev_id = self.get_rev_id(revno, revs)
944
raise BzrError('No namespace registered for string: %r' %
947
raise TypeError('Unhandled revision type %s' % revision)
951
raise bzrlib.errors.NoSuchRevision(self, revision)
954
def _namespace_revno(self, revs, revision):
955
"""Lookup a revision by revision number"""
956
assert revision.startswith('revno:')
958
return (int(revision[6:]),)
961
REVISION_NAMESPACES['revno:'] = _namespace_revno
963
def _namespace_revid(self, revs, revision):
964
assert revision.startswith('revid:')
965
rev_id = revision[len('revid:'):]
967
return revs.index(rev_id) + 1, rev_id
970
REVISION_NAMESPACES['revid:'] = _namespace_revid
972
def _namespace_last(self, revs, revision):
973
assert revision.startswith('last:')
975
offset = int(revision[5:])
980
raise BzrError('You must supply a positive value for --revision last:XXX')
981
return (len(revs) - offset + 1,)
982
REVISION_NAMESPACES['last:'] = _namespace_last
984
def _namespace_tag(self, revs, revision):
985
assert revision.startswith('tag:')
986
raise BzrError('tag: namespace registered, but not implemented.')
987
REVISION_NAMESPACES['tag:'] = _namespace_tag
989
def _namespace_date(self, revs, revision):
990
assert revision.startswith('date:')
992
# Spec for date revisions:
994
# value can be 'yesterday', 'today', 'tomorrow' or a YYYY-MM-DD string.
995
# it can also start with a '+/-/='. '+' says match the first
996
# entry after the given date. '-' is match the first entry before the date
997
# '=' is match the first entry after, but still on the given date.
999
# +2005-05-12 says find the first matching entry after May 12th, 2005 at 0:00
1000
# -2005-05-12 says find the first matching entry before May 12th, 2005 at 0:00
1001
# =2005-05-12 says find the first match after May 12th, 2005 at 0:00 but before
1002
# May 13th, 2005 at 0:00
1004
# So the proper way of saying 'give me all entries for today' is:
1005
# -r {date:+today}:{date:-tomorrow}
1006
# The default is '=' when not supplied
1009
if val[:1] in ('+', '-', '='):
1010
match_style = val[:1]
1013
today = datetime.datetime.today().replace(hour=0,minute=0,second=0,microsecond=0)
1014
if val.lower() == 'yesterday':
1015
dt = today - datetime.timedelta(days=1)
1016
elif val.lower() == 'today':
1018
elif val.lower() == 'tomorrow':
1019
dt = today + datetime.timedelta(days=1)
1022
# This should be done outside the function to avoid recompiling it.
1023
_date_re = re.compile(
1024
r'(?P<date>(?P<year>\d\d\d\d)-(?P<month>\d\d)-(?P<day>\d\d))?'
1026
r'(?P<time>(?P<hour>\d\d):(?P<minute>\d\d)(:(?P<second>\d\d))?)?'
1028
m = _date_re.match(val)
1029
if not m or (not m.group('date') and not m.group('time')):
1030
raise BzrError('Invalid revision date %r' % revision)
1033
year, month, day = int(m.group('year')), int(m.group('month')), int(m.group('day'))
1035
year, month, day = today.year, today.month, today.day
1037
hour = int(m.group('hour'))
1038
minute = int(m.group('minute'))
1039
if m.group('second'):
1040
second = int(m.group('second'))
1044
hour, minute, second = 0,0,0
1046
dt = datetime.datetime(year=year, month=month, day=day,
1047
hour=hour, minute=minute, second=second)
1051
if match_style == '-':
1053
elif match_style == '=':
1054
last = dt + datetime.timedelta(days=1)
1057
for i in range(len(revs)-1, -1, -1):
1058
r = self.get_revision(revs[i])
1059
# TODO: Handle timezone.
1060
dt = datetime.datetime.fromtimestamp(r.timestamp)
1061
if first >= dt and (last is None or dt >= last):
1064
for i in range(len(revs)):
1065
r = self.get_revision(revs[i])
1066
# TODO: Handle timezone.
1067
dt = datetime.datetime.fromtimestamp(r.timestamp)
1068
if first <= dt and (last is None or dt <= last):
1070
REVISION_NAMESPACES['date:'] = _namespace_date
585
1072
def revision_tree(self, revision_id):
586
1073
"""Return Tree for a revision on this branch.
588
1075
`revision_id` may be None for the null revision, in which case
589
1076
an `EmptyTree` is returned."""
1077
# TODO: refactor this to use an existing revision object
1078
# so we don't need to read it in twice.
591
1079
if revision_id == None:
592
1080
return EmptyTree()
626
def write_log(self, show_timezone='original'):
627
"""Write out human-readable log of commits to this branch
629
:param utc: If true, show dates in universal time, not local time."""
630
## TODO: Option to choose either original, utc or local timezone
633
for p in self.revision_history():
635
print 'revno:', revno
636
## TODO: Show hash if --id is given.
637
##print 'revision-hash:', p
638
rev = self.get_revision(p)
639
print 'committer:', rev.committer
640
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
643
## opportunistic consistency check, same as check_patch_chaining
644
if rev.precursor != precursor:
645
bailout("mismatched precursor!")
649
print ' (no message)'
651
for l in rev.message.split('\n'):
659
def show_status(branch, show_all=False):
660
"""Display single-line status for non-ignored working files.
662
The list is show sorted in order by file name.
664
>>> b = ScratchBranch(files=['foo', 'foo~'])
670
>>> b.commit("add foo")
672
>>> os.unlink(b._rel('foo'))
677
:todo: Get state for single files.
679
:todo: Perhaps show a slash at the end of directory names.
683
# We have to build everything into a list first so that it can
684
# sorted by name, incorporating all the different sources.
686
# FIXME: Rather than getting things in random order and then sorting,
687
# just step through in order.
689
# Interesting case: the old ID for a file has been removed,
690
# but a new file has been created under that name.
692
old = branch.basis_tree()
693
old_inv = old.inventory
694
new = branch.working_tree()
695
new_inv = new.inventory
697
for fs, fid, oldname, newname, kind in diff_trees(old, new):
699
show_status(fs, kind,
700
oldname + ' => ' + newname)
701
elif fs == 'A' or fs == 'M':
702
show_status(fs, kind, newname)
704
show_status(fs, kind, oldname)
707
show_status(fs, kind, newname)
710
show_status(fs, kind, newname)
712
show_status(fs, kind, newname)
714
bailout("wierd file state %r" % ((fs, fid),))
1105
def rename_one(self, from_rel, to_rel):
1108
This can change the directory or the filename or both.
1112
tree = self.working_tree()
1113
inv = tree.inventory
1114
if not tree.has_filename(from_rel):
1115
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1116
if tree.has_filename(to_rel):
1117
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1119
file_id = inv.path2id(from_rel)
1121
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1123
if inv.path2id(to_rel):
1124
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1126
to_dir, to_tail = os.path.split(to_rel)
1127
to_dir_id = inv.path2id(to_dir)
1128
if to_dir_id == None and to_dir != '':
1129
raise BzrError("can't determine destination directory id for %r" % to_dir)
1131
mutter("rename_one:")
1132
mutter(" file_id {%s}" % file_id)
1133
mutter(" from_rel %r" % from_rel)
1134
mutter(" to_rel %r" % to_rel)
1135
mutter(" to_dir %r" % to_dir)
1136
mutter(" to_dir_id {%s}" % to_dir_id)
1138
inv.rename(file_id, to_dir_id, to_tail)
1140
from_abs = self.abspath(from_rel)
1141
to_abs = self.abspath(to_rel)
1143
os.rename(from_abs, to_abs)
1145
raise BzrError("failed to rename %r to %r: %s"
1146
% (from_abs, to_abs, e[1]),
1147
["rename rolled back"])
1149
self._write_inventory(inv)
1154
def move(self, from_paths, to_name):
1157
to_name must exist as a versioned directory.
1159
If to_name exists and is a directory, the files are moved into
1160
it, keeping their old names. If it is a directory,
1162
Note that to_name is only the last component of the new name;
1163
this doesn't change the directory.
1165
This returns a list of (from_path, to_path) pairs for each
1166
entry that is moved.
1171
## TODO: Option to move IDs only
1172
assert not isinstance(from_paths, basestring)
1173
tree = self.working_tree()
1174
inv = tree.inventory
1175
to_abs = self.abspath(to_name)
1176
if not isdir(to_abs):
1177
raise BzrError("destination %r is not a directory" % to_abs)
1178
if not tree.has_filename(to_name):
1179
raise BzrError("destination %r not in working directory" % to_abs)
1180
to_dir_id = inv.path2id(to_name)
1181
if to_dir_id == None and to_name != '':
1182
raise BzrError("destination %r is not a versioned directory" % to_name)
1183
to_dir_ie = inv[to_dir_id]
1184
if to_dir_ie.kind not in ('directory', 'root_directory'):
1185
raise BzrError("destination %r is not a directory" % to_abs)
1187
to_idpath = inv.get_idpath(to_dir_id)
1189
for f in from_paths:
1190
if not tree.has_filename(f):
1191
raise BzrError("%r does not exist in working tree" % f)
1192
f_id = inv.path2id(f)
1194
raise BzrError("%r is not versioned" % f)
1195
name_tail = splitpath(f)[-1]
1196
dest_path = appendpath(to_name, name_tail)
1197
if tree.has_filename(dest_path):
1198
raise BzrError("destination %r already exists" % dest_path)
1199
if f_id in to_idpath:
1200
raise BzrError("can't move %r to a subdirectory of itself" % f)
1202
# OK, so there's a race here, it's possible that someone will
1203
# create a file in this interval and then the rename might be
1204
# left half-done. But we should have caught most problems.
1206
for f in from_paths:
1207
name_tail = splitpath(f)[-1]
1208
dest_path = appendpath(to_name, name_tail)
1209
result.append((f, dest_path))
1210
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1212
os.rename(self.abspath(f), self.abspath(dest_path))
1214
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1215
["rename rolled back"])
1217
self._write_inventory(inv)
1224
def revert(self, filenames, old_tree=None, backups=True):
1225
"""Restore selected files to the versions from a previous tree.
1228
If true (default) backups are made of files before
1231
from bzrlib.errors import NotVersionedError, BzrError
1232
from bzrlib.atomicfile import AtomicFile
1233
from bzrlib.osutils import backup_file
1235
inv = self.read_working_inventory()
1236
if old_tree is None:
1237
old_tree = self.basis_tree()
1238
old_inv = old_tree.inventory
1241
for fn in filenames:
1242
file_id = inv.path2id(fn)
1244
raise NotVersionedError("not a versioned file", fn)
1245
if not old_inv.has_id(file_id):
1246
raise BzrError("file not present in old tree", fn, file_id)
1247
nids.append((fn, file_id))
1249
# TODO: Rename back if it was previously at a different location
1251
# TODO: If given a directory, restore the entire contents from
1252
# the previous version.
1254
# TODO: Make a backup to a temporary file.
1256
# TODO: If the file previously didn't exist, delete it?
1257
for fn, file_id in nids:
1260
f = AtomicFile(fn, 'wb')
1262
f.write(old_tree.get_file(file_id).read())
1268
def pending_merges(self):
1269
"""Return a list of pending merges.
1271
These are revisions that have been merged into the working
1272
directory but not yet committed.
1274
cfn = self.controlfilename('pending-merges')
1275
if not os.path.exists(cfn):
1278
for l in self.controlfile('pending-merges', 'r').readlines():
1279
p.append(l.rstrip('\n'))
1283
def add_pending_merge(self, revision_id):
1284
from bzrlib.revision import validate_revision_id
1286
validate_revision_id(revision_id)
1288
p = self.pending_merges()
1289
if revision_id in p:
1291
p.append(revision_id)
1292
self.set_pending_merges(p)
1295
def set_pending_merges(self, rev_list):
1296
from bzrlib.atomicfile import AtomicFile
1299
f = AtomicFile(self.controlfilename('pending-merges'))
1310
def get_parent(self):
1311
"""Return the parent location of the branch.
1313
This is the default location for push/pull/missing. The usual
1314
pattern is that the user can override it by specifying a
1318
_locs = ['parent', 'pull', 'x-pull']
1321
return self.controlfile(l, 'r').read().strip('\n')
1323
if e.errno != errno.ENOENT:
1328
def set_parent(self, url):
1329
# TODO: Maybe delete old location files?
1330
from bzrlib.atomicfile import AtomicFile
1333
f = AtomicFile(self.controlfilename('parent'))
1342
def check_revno(self, revno):
1344
Check whether a revno corresponds to any revision.
1345
Zero (the NULL revision) is considered valid.
1348
self.check_real_revno(revno)
1350
def check_real_revno(self, revno):
1352
Check whether a revno corresponds to a real revision.
1353
Zero (the NULL revision) is considered invalid
1355
if revno < 1 or revno > self.revno():
1356
raise InvalidRevisionNumber(revno)
718
1361
class ScratchBranch(Branch):