15
16
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20
from cStringIO import StringIO
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
23
from bzrlib.trace import mutter, note
24
from bzrlib.osutils import isdir, quotefn, compact_date, rand_bytes, \
26
sha_file, appendpath, file_kind
28
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
30
from bzrlib.textui import show_status
31
from bzrlib.revision import Revision
32
from bzrlib.delta import compare_trees
33
from bzrlib.tree import EmptyTree, RevisionTree
34
from bzrlib.inventory import Inventory
35
from bzrlib.weavestore import WeaveStore
36
from bzrlib.store import ImmutableStore
41
INVENTORY_FILEID = '__inventory'
42
ANCESTRY_FILEID = '__ancestry'
45
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
46
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
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
39
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
47
40
## TODO: Maybe include checks for common corruption of newlines, etc?
50
# TODO: Some operations like log might retrieve the same revisions
51
# repeatedly to calculate deltas. We could perhaps have a weakref
52
# cache in memory to make this faster. In general anything can be
53
# cached in memory between lock and unlock operations.
55
# TODO: please move the revision-string syntax stuff out of the branch
56
# object; it's clutter
59
def find_branch(f, **args):
60
if f and (f.startswith('http://') or f.startswith('https://')):
62
return remotebranch.RemoteBranch(f, **args)
64
return Branch(f, **args)
67
def find_cached_branch(f, cache_root, **args):
68
from remotebranch import RemoteBranch
69
br = find_branch(f, **args)
70
def cacheify(br, store_name):
71
from meta_store import CachedStore
72
cache_path = os.path.join(cache_root, store_name)
74
new_store = CachedStore(getattr(br, store_name), cache_path)
75
setattr(br, store_name, new_store)
77
if isinstance(br, RemoteBranch):
78
cacheify(br, 'inventory_store')
79
cacheify(br, 'text_store')
80
cacheify(br, 'revision_store')
84
def _relpath(base, path):
85
"""Return path relative to base, or raise exception.
87
The path may be either an absolute path or a path relative to the
88
current working directory.
90
Lifted out of Branch.relpath for ease of testing.
92
os.path.commonprefix (python2.4) has a bad bug that it works just
93
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
94
avoids that problem."""
95
rp = os.path.abspath(path)
99
while len(head) >= len(base):
102
head, tail = os.path.split(head)
106
from errors import NotBranchError
107
raise NotBranchError("path %r is not within branch %r" % (rp, base))
109
return os.sep.join(s)
112
def find_branch_root(f=None):
113
"""Find the branch root enclosing f, or pwd.
115
f may be a filename or a URL.
117
It is not necessary that f exists.
119
Basically we keep looking up until we find the control directory or
120
run into the root. If there isn't one, raises NotBranchError.
124
elif hasattr(os.path, 'realpath'):
125
f = os.path.realpath(f)
127
f = os.path.abspath(f)
128
if not os.path.exists(f):
129
raise BzrError('%r does not exist' % f)
135
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
137
head, tail = os.path.split(f)
139
# reached the root, whatever that may be
140
raise bzrlib.errors.NotBranchError('%s is not in a branch' % orig_f)
145
# XXX: move into bzrlib.errors; subclass BzrError
146
class DivergedBranches(Exception):
147
def __init__(self, branch1, branch2):
148
self.branch1 = branch1
149
self.branch2 = branch2
150
Exception.__init__(self, "These branches have diverged.")
153
46
######################################################################
156
class Branch(object):
157
50
"""Branch holding a history of revisions.
160
Base directory of the branch.
166
If _lock_mode is true, a positive count of the number of times the
170
Lock object from bzrlib.lock.
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.
176
_inventory_weave = None
178
# Map some sort of prefix into a namespace
179
# stuff like "revno:10", "revid:", etc.
180
# This should match a prefix with a function which accepts
181
REVISION_NAMESPACES = {}
183
def __init__(self, base, init=False, find_root=True):
65
def __init__(self, base, init=False):
184
66
"""Create new branch object at a particular location.
186
base -- Base directory for the branch.
188
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
189
71
unversioned directory. If False, the branch must already
192
find_root -- If true and init is false, find the root of the
193
existing branch containing base.
195
74
In the test suite, creation of new trees is tested using the
196
75
`ScratchBranch` class.
77
self.base = os.path.realpath(base)
199
self.base = os.path.realpath(base)
200
79
self._make_control()
202
self.base = find_branch_root(base)
204
self.base = os.path.realpath(base)
205
81
if not isdir(self.controlfilename('.')):
206
from errors import NotBranchError
207
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
208
['use "bzr init" to initialize a new working tree',
209
'current bzr can only operate from top-of-tree'])
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'])
212
self.weave_store = WeaveStore(self.controlfilename('weaves'))
87
self.text_store = ImmutableStore(self.controlfilename('text-store'))
213
88
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
89
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
216
92
def __str__(self):
220
96
__repr__ = __str__
224
if self._lock_mode or self._lock:
225
from warnings import warn
226
warn("branch %r was not explicitly unlocked" % self)
230
def lock_write(self):
232
if self._lock_mode != 'w':
233
from errors import LockError
234
raise LockError("can't upgrade to a write lock from %r" %
236
self._lock_count += 1
238
from bzrlib.lock import WriteLock
240
self._lock = WriteLock(self.controlfilename('branch-lock'))
241
self._lock_mode = 'w'
247
assert self._lock_mode in ('r', 'w'), \
248
"invalid lock mode %r" % self._lock_mode
249
self._lock_count += 1
251
from bzrlib.lock import ReadLock
253
self._lock = ReadLock(self.controlfilename('branch-lock'))
254
self._lock_mode = 'r'
258
if not self._lock_mode:
259
from errors import LockError
260
raise LockError('branch %r is not locked' % (self))
262
if self._lock_count > 1:
263
self._lock_count -= 1
267
self._lock_mode = self._lock_count = None
269
def abspath(self, name):
270
"""Return absolute filename for something in the branch"""
100
"""Return filename relative to branch top"""
271
101
return os.path.join(self.base, name)
273
def relpath(self, path):
274
"""Return path relative to this branch of something inside it.
276
Raises an error if path is not in this branch."""
277
return _relpath(self.base, path)
279
104
def controlfilename(self, file_or_path):
280
105
"""Return location relative to branch."""
281
if isinstance(file_or_path, basestring):
106
if isinstance(file_or_path, types.StringTypes):
282
107
file_or_path = [file_or_path]
283
108
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
286
111
def controlfile(self, file_or_path, mode='r'):
287
"""Open a control file for this branch.
289
There are two classes of file in the control directory: text
290
and binary. binary files are untranslated byte streams. Text
291
control files are stored with Unix newlines and in UTF-8, even
292
if the platform or locale defaults are different.
294
Controlfiles should almost never be opened in write mode but
295
rather should be atomically copied and replaced using atomicfile.
298
fn = self.controlfilename(file_or_path)
300
if mode == 'rb' or mode == 'wb':
301
return file(fn, mode)
302
elif mode == 'r' or mode == 'w':
303
# open in binary mode anyhow so there's no newline translation;
304
# codecs uses line buffering by default; don't want that.
306
return codecs.open(fn, mode + 'b', 'utf-8',
309
raise BzrError("invalid controlfile mode %r" % mode)
112
"""Open a control file for this branch"""
113
return file(self.controlfilename(file_or_path), mode)
311
116
def _make_control(self):
312
117
os.mkdir(self.controlfilename([]))
313
118
self.controlfile('README', 'w').write(
314
119
"This is a Bazaar-NG control directory.\n"
315
"Do not change any files in this directory.\n")
316
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT_5)
317
for d in ('text-store', 'revision-store',
120
"Do not change any files in this directory.")
121
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
122
for d in ('text-store', 'inventory-store', 'revision-store'):
319
123
os.mkdir(self.controlfilename(d))
320
124
for f in ('revision-history', 'merged-patches',
321
'pending-merged-patches', 'branch-name',
125
'pending-merged-patches', 'branch-name'):
324
126
self.controlfile(f, 'w').write('')
325
127
mutter('created control directory in ' + self.base)
327
# if we want per-tree root ids then this is the place to set
328
# them; they're not needed for now and so ommitted for
330
f = self.controlfile('inventory','w')
331
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), f)
128
Inventory().write_xml(self.controlfile('inventory','w'))
335
131
def _check_format(self):
336
132
"""Check this branch format is supported.
338
The format level is stored, as an integer, in
339
self._branch_format for code that needs to check it later.
134
The current tool only supports the current unstable format.
341
136
In the future, we might need different in-memory Branch
342
137
classes to support downlevel branches. But not yet.
344
fmt = self.controlfile('branch-format', 'r').read()
345
if fmt == BZR_BRANCH_FORMAT_5:
346
self._branch_format = 5
348
raise BzrError('sorry, branch format "%s" not supported; '
349
'use a different bzr version, '
350
'or run "bzr upgrade", '
351
'or remove the .bzr directory and "bzr init" again'
352
% fmt.rstrip('\n\r'))
354
def get_root_id(self):
355
"""Return the id of this branches root"""
356
inv = self.read_working_inventory()
357
return inv.root.file_id
359
def set_root_id(self, file_id):
360
inv = self.read_working_inventory()
361
orig_root_id = inv.root.file_id
362
del inv._byid[inv.root.file_id]
363
inv.root.file_id = file_id
364
inv._byid[inv.root.file_id] = inv.root
367
if entry.parent_id in (None, orig_root_id):
368
entry.parent_id = inv.root.file_id
369
self._write_inventory(inv)
139
# read in binary mode to detect newline wierdness.
140
fmt = self.controlfile('branch-format', 'rb').read()
141
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'])
371
147
def read_working_inventory(self):
372
148
"""Read the working inventory."""
375
# ElementTree does its own conversion from UTF-8, so open in
377
f = self.controlfile('inventory', 'rb')
378
return bzrlib.xml5.serializer_v5.read_inventory(f)
150
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
151
mutter("loaded inventory of %d items in %f"
152
% (len(inv), time.time() - before))
383
156
def _write_inventory(self, inv):
384
157
"""Update the working inventory.
386
159
That is to say, the inventory describing changes underway, that
387
160
will be committed to the next revision.
389
from bzrlib.atomicfile import AtomicFile
393
f = AtomicFile(self.controlfilename('inventory'), 'wb')
395
bzrlib.xml5.serializer_v5.write_inventory(inv, f)
402
mutter('wrote working inventory')
162
inv.write_xml(self.controlfile('inventory', 'w'))
163
mutter('wrote inventory to %s' % quotefn(self.controlfilename('inventory')))
405
166
inventory = property(read_working_inventory, _write_inventory, None,
406
167
"""Inventory for the working copy.""")
409
def add(self, files, ids=None):
170
def add(self, files, verbose=False):
410
171
"""Make files versioned.
412
Note that the command line normally calls smart_add instead,
413
which can automatically recurse.
415
173
This puts the files in the Added state, so that they will be
416
174
recorded by the next commit.
419
List of paths to add, relative to the base of the tree.
422
If set, use these instead of automatically generated ids.
423
Must be the same length as the list of files, but may
424
contain None for ids that are to be autogenerated.
426
TODO: Perhaps have an option to add the ids even if the files do
429
TODO: Perhaps yield the ids and paths as they're added.
176
:todo: Perhaps have an option to add the ids even if the files do
179
:todo: Perhaps return the ids of the files? But then again it
180
is easy to retrieve them if they're needed.
182
:todo: Option to specify file id.
184
:todo: Adding a directory should optionally recurse down and
185
add all non-ignored children. Perhaps do that in a
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', [])
431
211
# TODO: Re-adding a file that is removed in the working copy
432
212
# should probably put it back with the previous ID.
433
if isinstance(files, basestring):
434
assert(ids is None or isinstance(ids, basestring))
213
if isinstance(files, types.StringTypes):
440
ids = [None] * len(files)
442
assert(len(ids) == len(files))
446
inv = self.read_working_inventory()
447
for f,file_id in zip(files, ids):
448
if is_control_file(f):
449
raise BzrError("cannot add control file %s" % quotefn(f))
454
raise BzrError("cannot add top-level %r" % f)
456
fullpath = os.path.normpath(self.abspath(f))
459
kind = file_kind(fullpath)
461
# maybe something better?
462
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
464
if kind != 'file' and kind != 'directory':
465
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
468
file_id = gen_file_id(f)
469
inv.add_path(f, kind=kind, file_id=file_id)
471
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
473
self._write_inventory(inv)
478
def print_file(self, file, revno):
479
"""Print `file` to stdout."""
482
tree = self.revision_tree(self.lookup_revision(revno))
483
# use inventory as it was in that revision
484
file_id = tree.inventory.path2id(file)
486
raise BzrError("%r is not present in revision %s" % (file, revno))
487
tree.print_file(file_id)
216
inv = self.read_working_inventory()
218
if is_control_file(f):
219
bailout("cannot add control file %s" % quotefn(f))
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))
248
show_status('A', kind, quotefn(f))
250
mutter("add file %s file_id:{%s} kind=%r parent_id={%s}"
251
% (f, file_id, kind, parent_id))
252
self._write_inventory(inv)
492
256
def remove(self, files, verbose=False):
495
259
This does not remove their text. This does not run on
497
TODO: Refuse to remove modified files unless --force is given?
499
TODO: Do something useful with directories.
501
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
502
288
removing may be useful and the user can just use use rm, and
503
289
is the opposite of add. Removing it is consistent with most
504
290
other tools. Maybe an option.
506
292
## TODO: Normalize names
507
293
## TODO: Remove nested loops; better scalability
508
if isinstance(files, basestring):
295
if isinstance(files, types.StringTypes):
514
tree = self.working_tree()
517
# do this before any modifications
521
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
522
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
524
# having remove it, it must be either ignored or unknown
525
if tree.is_ignored(f):
529
show_status(new_status, inv[fid].kind, quotefn(f))
532
self._write_inventory(inv)
537
# FIXME: this doesn't need to be a branch method
538
def set_inventory(self, new_inventory_list):
539
from bzrlib.inventory import Inventory, InventoryEntry
540
inv = Inventory(self.get_root_id())
541
for path, file_id, parent, kind in new_inventory_list:
542
name = os.path.basename(path)
545
inv.add(InventoryEntry(file_id, name, kind, parent))
298
inv = self.read_working_inventory()
300
# do this before any modifications
304
bailout("cannot remove unversioned file %s" % quotefn(f))
305
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
307
show_status('D', inv[fid].kind, quotefn(f))
546
310
self._write_inventory(inv)
565
329
return self.working_tree().unknowns()
568
def append_revision(self, *revision_ids):
569
from bzrlib.atomicfile import AtomicFile
571
for revision_id in revision_ids:
572
mutter("add {%s} to revision-history" % revision_id)
574
rev_history = self.revision_history()
575
rev_history.extend(revision_ids)
577
f = AtomicFile(self.controlfilename('revision-history'))
579
for rev_id in rev_history:
586
def get_revision_xml_file(self, revision_id):
587
"""Return XML file object for revision object."""
588
if not revision_id or not isinstance(revision_id, basestring):
589
raise InvalidRevisionId(revision_id)
594
return self.revision_store[revision_id]
596
raise bzrlib.errors.NoSuchRevision(self, revision_id)
602
get_revision_xml = get_revision_xml_file
332
def commit(self, message, timestamp=None, timezone=None,
335
"""Commit working copy as a new revision.
337
The basic approach is to add all the file texts into the
338
store, then the inventory, then make a new revision pointing
339
to that inventory and store that.
341
This is not quite safe if the working copy changes during the
342
commit; for the moment that is simply not allowed. A better
343
approach is to make a temporary copy of the files before
344
computing their hashes, and then add those hashes in turn to
345
the inventory. This should mean at least that there are no
346
broken hash pointers. There is no way we can get a snapshot
347
of the whole directory at an instant. This would also have to
348
be robust against files disappearing, moving, etc. So the
349
whole thing is a bit hard.
351
:param timestamp: if not None, seconds-since-epoch for a
352
postdated/predated commit.
355
## TODO: Show branch names
357
# TODO: Don't commit if there are no changes, unless forced?
359
# First walk over the working inventory; and both update that
360
# and also build a new revision inventory. The revision
361
# inventory needs to hold the text-id, sha1 and size of the
362
# actual file versions committed in the revision. (These are
363
# not present in the working inventory.) We also need to
364
# detect missing/deleted files, and remove them from the
367
work_inv = self.read_working_inventory()
369
basis = self.basis_tree()
370
basis_inv = basis.inventory
372
for path, entry in work_inv.iter_entries():
373
## TODO: Cope with files that have gone missing.
375
## TODO: Check that the file kind has not changed from the previous
376
## revision of this file (if any).
381
file_id = entry.file_id
382
mutter('commit prep file %s, id %r ' % (p, file_id))
384
if not os.path.exists(p):
385
mutter(" file is missing, removing from inventory")
387
show_status('D', entry.kind, quotefn(path))
388
missing_ids.append(file_id)
391
# TODO: Handle files that have been deleted
393
# TODO: Maybe a special case for empty files? Seems a
394
# waste to store them many times.
398
if basis_inv.has_id(file_id):
399
old_kind = basis_inv[file_id].kind
400
if old_kind != entry.kind:
401
bailout("entry %r changed kind from %r to %r"
402
% (file_id, old_kind, entry.kind))
404
if entry.kind == 'directory':
406
bailout("%s is entered as directory but not a directory" % quotefn(p))
407
elif entry.kind == 'file':
409
bailout("%s is entered as file but is not a file" % quotefn(p))
411
content = file(p, 'rb').read()
413
entry.text_sha1 = sha_string(content)
414
entry.text_size = len(content)
416
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
418
and (old_ie.text_size == entry.text_size)
419
and (old_ie.text_sha1 == entry.text_sha1)):
420
## assert content == basis.get_file(file_id).read()
421
entry.text_id = basis_inv[file_id].text_id
422
mutter(' unchanged from previous text_id {%s}' %
426
entry.text_id = _gen_file_id(entry.name)
427
self.text_store.add(content, entry.text_id)
428
mutter(' stored with text_id {%s}' % entry.text_id)
432
elif (old_ie.name == entry.name
433
and old_ie.parent_id == entry.parent_id):
438
show_status(state, entry.kind, quotefn(path))
440
for file_id in missing_ids:
441
# have to do this later so we don't mess up the iterator.
442
# since parents may be removed before their children we
445
# FIXME: There's probably a better way to do this; perhaps
446
# the workingtree should know how to filter itself.
447
if work_inv.has_id(file_id):
448
del work_inv[file_id]
451
inv_id = rev_id = _gen_revision_id(time.time())
453
inv_tmp = tempfile.TemporaryFile()
454
inv.write_xml(inv_tmp)
456
self.inventory_store.add(inv_tmp, inv_id)
457
mutter('new inventory_id is {%s}' % inv_id)
459
self._write_inventory(work_inv)
461
if timestamp == None:
462
timestamp = time.time()
464
if committer == None:
465
committer = username()
468
timezone = local_time_offset()
470
mutter("building commit log message")
471
rev = Revision(timestamp=timestamp,
474
precursor = self.last_patch(),
479
rev_tmp = tempfile.TemporaryFile()
480
rev.write_xml(rev_tmp)
482
self.revision_store.add(rev_tmp, rev_id)
483
mutter("new revision_id is {%s}" % rev_id)
485
## XXX: Everything up to here can simply be orphaned if we abort
486
## the commit; it will leave junk files behind but that doesn't
489
## TODO: Read back the just-generated changeset, and make sure it
490
## applies and recreates the right state.
492
## TODO: Also calculate and store the inventory SHA1
493
mutter("committing patch r%d" % (self.revno() + 1))
495
mutter("append to revision-history")
496
self.controlfile('revision-history', 'at').write(rev_id + '\n')
605
501
def get_revision(self, revision_id):
606
502
"""Return the Revision object for a named revision"""
607
xml_file = self.get_revision_xml_file(revision_id)
610
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
611
except SyntaxError, e:
612
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
503
r = Revision.read_xml(self.revision_store[revision_id])
616
504
assert r.revision_id == revision_id
620
def get_revision_delta(self, revno):
621
"""Return the delta for one revision.
623
The delta is relative to its mainline predecessor, or the
624
empty tree for revision 1.
626
assert isinstance(revno, int)
627
rh = self.revision_history()
628
if not (1 <= revno <= len(rh)):
629
raise InvalidRevisionNumber(revno)
631
# revno is 1-based; list is 0-based
633
new_tree = self.revision_tree(rh[revno-1])
635
old_tree = EmptyTree()
637
old_tree = self.revision_tree(rh[revno-2])
639
return compare_trees(old_tree, new_tree)
643
def get_revision_sha1(self, revision_id):
644
"""Hash the stored value of a revision, and return it."""
645
# In the future, revision entries will be signed. At that
646
# point, it is probably best *not* to include the signature
647
# in the revision hash. Because that lets you re-sign
648
# the revision, (add signatures/remove signatures) and still
649
# have all hash pointers stay consistent.
650
# But for now, just hash the contents.
651
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
654
def get_ancestry(self, revision_id):
655
"""Return a list of revision-ids integrated by a revision.
657
w = self.weave_store.get_weave(ANCESTRY_FILEID)
659
return [l[:-1] for l in w.get_iter(w.lookup(revision_id))]
662
def get_inventory_weave(self):
663
return self.weave_store.get_weave(INVENTORY_FILEID)
666
def get_inventory(self, revision_id):
667
"""Get Inventory object by hash."""
668
# FIXME: The text gets passed around a lot coming from the weave.
669
f = StringIO(self.get_inventory_xml(revision_id))
670
return bzrlib.xml5.serializer_v5.read_inventory(f)
673
def get_inventory_xml(self, revision_id):
674
"""Get inventory XML as a file object."""
676
assert isinstance(revision_id, basestring), type(revision_id)
677
iw = self.get_inventory_weave()
678
return iw.get_text(iw.lookup(revision_id))
680
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
683
def get_inventory_sha1(self, revision_id):
684
"""Return the sha1 hash of the inventory entry
686
return self.get_revision(revision_id).inventory_sha1
508
def get_inventory(self, inventory_id):
509
"""Get Inventory object by hash.
511
:todo: Perhaps for this and similar methods, take a revision
512
parameter which can be either an integer revno or a
514
i = Inventory.read_xml(self.inventory_store[inventory_id])
689
518
def get_revision_inventory(self, revision_id):
690
519
"""Return inventory of a past revision."""
691
# bzr 0.0.6 and later imposes the constraint that the inventory_id
692
# must be the same as its revision, so this is trivial.
693
520
if revision_id == None:
694
return Inventory(self.get_root_id())
696
return self.get_inventory(revision_id)
523
return self.get_inventory(self.get_revision(revision_id).inventory_id)
699
526
def revision_history(self):
760
538
That is equivalent to the number of revisions committed to
541
>>> b = ScratchBranch()
544
>>> b.commit('no foo')
763
548
return len(self.revision_history())
766
551
def last_patch(self):
767
552
"""Return last patch hash, or None if no history.
554
>>> ScratchBranch().last_patch() == None
769
557
ph = self.revision_history()
776
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
778
If self and other have not diverged, return a list of the revisions
779
present in other, but missing from self.
781
>>> from bzrlib.commit import commit
782
>>> bzrlib.trace.silent = True
783
>>> br1 = ScratchBranch()
784
>>> br2 = ScratchBranch()
785
>>> br1.missing_revisions(br2)
787
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
788
>>> br1.missing_revisions(br2)
790
>>> br2.missing_revisions(br1)
792
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
793
>>> br1.missing_revisions(br2)
795
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
796
>>> br1.missing_revisions(br2)
798
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
799
>>> br1.missing_revisions(br2)
800
Traceback (most recent call last):
801
DivergedBranches: These branches have diverged.
803
self_history = self.revision_history()
804
self_len = len(self_history)
805
other_history = other.revision_history()
806
other_len = len(other_history)
807
common_index = min(self_len, other_len) -1
808
if common_index >= 0 and \
809
self_history[common_index] != other_history[common_index]:
810
raise DivergedBranches(self, other)
812
if stop_revision is None:
813
stop_revision = other_len
814
elif stop_revision > other_len:
815
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
817
return other_history[self_len:stop_revision]
820
def update_revisions(self, other, stop_revision=None):
821
"""Pull in all new revisions from other branch.
823
from bzrlib.fetch import greedy_fetch
825
pb = bzrlib.ui.ui_factory.progress_bar()
826
pb.update('comparing histories')
828
revision_ids = self.missing_revisions(other, stop_revision)
830
if len(revision_ids) > 0:
831
count = greedy_fetch(self, other, revision_ids[-1], pb)[0]
834
self.append_revision(*revision_ids)
835
## note("Added %d revisions." % count)
839
def commit(self, *args, **kw):
840
from bzrlib.commit import Commit
841
Commit().commit(self, *args, **kw)
844
def lookup_revision(self, revision):
845
"""Return the revision identifier for a given revision information."""
846
revno, info = self._get_revision_info(revision)
850
def revision_id_to_revno(self, revision_id):
851
"""Given a revision id, return its revno"""
852
history = self.revision_history()
854
return history.index(revision_id) + 1
856
raise bzrlib.errors.NoSuchRevision(self, revision_id)
859
def get_revision_info(self, revision):
860
"""Return (revno, revision id) for revision identifier.
862
revision can be an integer, in which case it is assumed to be revno (though
863
this will translate negative values into positive ones)
864
revision can also be a string, in which case it is parsed for something like
865
'date:' or 'revid:' etc.
867
revno, rev_id = self._get_revision_info(revision)
869
raise bzrlib.errors.NoSuchRevision(self, revision)
872
def get_rev_id(self, revno, history=None):
873
"""Find the revision id of the specified revno."""
562
def lookup_revision(self, revno):
563
"""Return revision hash for revision number."""
877
history = self.revision_history()
878
elif revno <= 0 or revno > len(history):
879
raise bzrlib.errors.NoSuchRevision(self, revno)
880
return history[revno - 1]
882
def _get_revision_info(self, revision):
883
"""Return (revno, revision id) for revision specifier.
885
revision can be an integer, in which case it is assumed to be revno
886
(though this will translate negative values into positive ones)
887
revision can also be a string, in which case it is parsed for something
888
like 'date:' or 'revid:' etc.
890
A revid is always returned. If it is None, the specifier referred to
891
the null revision. If the revid does not occur in the revision
892
history, revno will be None.
898
try:# Convert to int if possible
899
revision = int(revision)
902
revs = self.revision_history()
903
if isinstance(revision, int):
905
revno = len(revs) + revision + 1
908
rev_id = self.get_rev_id(revno, revs)
909
elif isinstance(revision, basestring):
910
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
911
if revision.startswith(prefix):
912
result = func(self, revs, revision)
914
revno, rev_id = result
917
rev_id = self.get_rev_id(revno, revs)
920
raise BzrError('No namespace registered for string: %r' %
923
raise TypeError('Unhandled revision type %s' % revision)
927
raise bzrlib.errors.NoSuchRevision(self, revision)
930
def _namespace_revno(self, revs, revision):
931
"""Lookup a revision by revision number"""
932
assert revision.startswith('revno:')
934
return (int(revision[6:]),)
937
REVISION_NAMESPACES['revno:'] = _namespace_revno
939
def _namespace_revid(self, revs, revision):
940
assert revision.startswith('revid:')
941
rev_id = revision[len('revid:'):]
943
return revs.index(rev_id) + 1, rev_id
946
REVISION_NAMESPACES['revid:'] = _namespace_revid
948
def _namespace_last(self, revs, revision):
949
assert revision.startswith('last:')
951
offset = int(revision[5:])
956
raise BzrError('You must supply a positive value for --revision last:XXX')
957
return (len(revs) - offset + 1,)
958
REVISION_NAMESPACES['last:'] = _namespace_last
960
def _namespace_tag(self, revs, revision):
961
assert revision.startswith('tag:')
962
raise BzrError('tag: namespace registered, but not implemented.')
963
REVISION_NAMESPACES['tag:'] = _namespace_tag
965
def _namespace_date(self, revs, revision):
966
assert revision.startswith('date:')
968
# Spec for date revisions:
970
# value can be 'yesterday', 'today', 'tomorrow' or a YYYY-MM-DD string.
971
# it can also start with a '+/-/='. '+' says match the first
972
# entry after the given date. '-' is match the first entry before the date
973
# '=' is match the first entry after, but still on the given date.
975
# +2005-05-12 says find the first matching entry after May 12th, 2005 at 0:00
976
# -2005-05-12 says find the first matching entry before May 12th, 2005 at 0:00
977
# =2005-05-12 says find the first match after May 12th, 2005 at 0:00 but before
978
# May 13th, 2005 at 0:00
980
# So the proper way of saying 'give me all entries for today' is:
981
# -r {date:+today}:{date:-tomorrow}
982
# The default is '=' when not supplied
985
if val[:1] in ('+', '-', '='):
986
match_style = val[:1]
989
today = datetime.datetime.today().replace(hour=0,minute=0,second=0,microsecond=0)
990
if val.lower() == 'yesterday':
991
dt = today - datetime.timedelta(days=1)
992
elif val.lower() == 'today':
994
elif val.lower() == 'tomorrow':
995
dt = today + datetime.timedelta(days=1)
998
# This should be done outside the function to avoid recompiling it.
999
_date_re = re.compile(
1000
r'(?P<date>(?P<year>\d\d\d\d)-(?P<month>\d\d)-(?P<day>\d\d))?'
1002
r'(?P<time>(?P<hour>\d\d):(?P<minute>\d\d)(:(?P<second>\d\d))?)?'
1004
m = _date_re.match(val)
1005
if not m or (not m.group('date') and not m.group('time')):
1006
raise BzrError('Invalid revision date %r' % revision)
1009
year, month, day = int(m.group('year')), int(m.group('month')), int(m.group('day'))
1011
year, month, day = today.year, today.month, today.day
1013
hour = int(m.group('hour'))
1014
minute = int(m.group('minute'))
1015
if m.group('second'):
1016
second = int(m.group('second'))
1020
hour, minute, second = 0,0,0
1022
dt = datetime.datetime(year=year, month=month, day=day,
1023
hour=hour, minute=minute, second=second)
1027
if match_style == '-':
1029
elif match_style == '=':
1030
last = dt + datetime.timedelta(days=1)
1033
for i in range(len(revs)-1, -1, -1):
1034
r = self.get_revision(revs[i])
1035
# TODO: Handle timezone.
1036
dt = datetime.datetime.fromtimestamp(r.timestamp)
1037
if first >= dt and (last is None or dt >= last):
1040
for i in range(len(revs)):
1041
r = self.get_revision(revs[i])
1042
# TODO: Handle timezone.
1043
dt = datetime.datetime.fromtimestamp(r.timestamp)
1044
if first <= dt and (last is None or dt <= last):
1046
REVISION_NAMESPACES['date:'] = _namespace_date
568
# list is 0-based; revisions are 1-based
569
return self.revision_history()[revno-1]
571
bailout("no such revision %s" % revno)
1048
574
def revision_tree(self, revision_id):
1049
575
"""Return Tree for a revision on this branch.
1051
577
`revision_id` may be None for the null revision, in which case
1052
578
an `EmptyTree` is returned."""
1053
# TODO: refactor this to use an existing revision object
1054
# so we don't need to read it in twice.
1055
580
if revision_id == None:
1056
581
return EmptyTree()
1058
583
inv = self.get_revision_inventory(revision_id)
1059
return RevisionTree(self.weave_store, inv, revision_id)
584
return RevisionTree(self.text_store, inv)
1062
587
def working_tree(self):
1063
588
"""Return a `Tree` for the working copy."""
1064
from workingtree import WorkingTree
1065
589
return WorkingTree(self.base, self.read_working_inventory())
1069
593
"""Return `Tree` object for last revision.
1071
595
If there are no revisions yet, return an `EmptyTree`.
1073
return self.revision_tree(self.last_patch())
1076
def rename_one(self, from_rel, to_rel):
1079
This can change the directory or the filename or both.
1083
tree = self.working_tree()
1084
inv = tree.inventory
1085
if not tree.has_filename(from_rel):
1086
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1087
if tree.has_filename(to_rel):
1088
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1090
file_id = inv.path2id(from_rel)
1092
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1094
if inv.path2id(to_rel):
1095
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1097
to_dir, to_tail = os.path.split(to_rel)
1098
to_dir_id = inv.path2id(to_dir)
1099
if to_dir_id == None and to_dir != '':
1100
raise BzrError("can't determine destination directory id for %r" % to_dir)
1102
mutter("rename_one:")
1103
mutter(" file_id {%s}" % file_id)
1104
mutter(" from_rel %r" % from_rel)
1105
mutter(" to_rel %r" % to_rel)
1106
mutter(" to_dir %r" % to_dir)
1107
mutter(" to_dir_id {%s}" % to_dir_id)
1109
inv.rename(file_id, to_dir_id, to_tail)
1111
from_abs = self.abspath(from_rel)
1112
to_abs = self.abspath(to_rel)
1114
os.rename(from_abs, to_abs)
1116
raise BzrError("failed to rename %r to %r: %s"
1117
% (from_abs, to_abs, e[1]),
1118
["rename rolled back"])
1120
self._write_inventory(inv)
1125
def move(self, from_paths, to_name):
1128
to_name must exist as a versioned directory.
1130
If to_name exists and is a directory, the files are moved into
1131
it, keeping their old names. If it is a directory,
1133
Note that to_name is only the last component of the new name;
1134
this doesn't change the directory.
1136
This returns a list of (from_path, to_path) pairs for each
1137
entry that is moved.
1142
## TODO: Option to move IDs only
1143
assert not isinstance(from_paths, basestring)
1144
tree = self.working_tree()
1145
inv = tree.inventory
1146
to_abs = self.abspath(to_name)
1147
if not isdir(to_abs):
1148
raise BzrError("destination %r is not a directory" % to_abs)
1149
if not tree.has_filename(to_name):
1150
raise BzrError("destination %r not in working directory" % to_abs)
1151
to_dir_id = inv.path2id(to_name)
1152
if to_dir_id == None and to_name != '':
1153
raise BzrError("destination %r is not a versioned directory" % to_name)
1154
to_dir_ie = inv[to_dir_id]
1155
if to_dir_ie.kind not in ('directory', 'root_directory'):
1156
raise BzrError("destination %r is not a directory" % to_abs)
1158
to_idpath = inv.get_idpath(to_dir_id)
1160
for f in from_paths:
1161
if not tree.has_filename(f):
1162
raise BzrError("%r does not exist in working tree" % f)
1163
f_id = inv.path2id(f)
1165
raise BzrError("%r is not versioned" % f)
1166
name_tail = splitpath(f)[-1]
1167
dest_path = appendpath(to_name, name_tail)
1168
if tree.has_filename(dest_path):
1169
raise BzrError("destination %r already exists" % dest_path)
1170
if f_id in to_idpath:
1171
raise BzrError("can't move %r to a subdirectory of itself" % f)
1173
# OK, so there's a race here, it's possible that someone will
1174
# create a file in this interval and then the rename might be
1175
# left half-done. But we should have caught most problems.
1177
for f in from_paths:
1178
name_tail = splitpath(f)[-1]
1179
dest_path = appendpath(to_name, name_tail)
1180
result.append((f, dest_path))
1181
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1183
os.rename(self.abspath(f), self.abspath(dest_path))
1185
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1186
["rename rolled back"])
1188
self._write_inventory(inv)
1195
def revert(self, filenames, old_tree=None, backups=True):
1196
"""Restore selected files to the versions from a previous tree.
1199
If true (default) backups are made of files before
1202
from bzrlib.errors import NotVersionedError, BzrError
1203
from bzrlib.atomicfile import AtomicFile
1204
from bzrlib.osutils import backup_file
1206
inv = self.read_working_inventory()
1207
if old_tree is None:
1208
old_tree = self.basis_tree()
1209
old_inv = old_tree.inventory
1212
for fn in filenames:
1213
file_id = inv.path2id(fn)
1215
raise NotVersionedError("not a versioned file", fn)
1216
if not old_inv.has_id(file_id):
1217
raise BzrError("file not present in old tree", fn, file_id)
1218
nids.append((fn, file_id))
1220
# TODO: Rename back if it was previously at a different location
1222
# TODO: If given a directory, restore the entire contents from
1223
# the previous version.
1225
# TODO: Make a backup to a temporary file.
1227
# TODO: If the file previously didn't exist, delete it?
1228
for fn, file_id in nids:
1231
f = AtomicFile(fn, 'wb')
1233
f.write(old_tree.get_file(file_id).read())
1239
def pending_merges(self):
1240
"""Return a list of pending merges.
1242
These are revisions that have been merged into the working
1243
directory but not yet committed.
1245
cfn = self.controlfilename('pending-merges')
1246
if not os.path.exists(cfn):
1249
for l in self.controlfile('pending-merges', 'r').readlines():
1250
p.append(l.rstrip('\n'))
1254
def add_pending_merge(self, revision_id):
1255
from bzrlib.revision import validate_revision_id
1257
validate_revision_id(revision_id)
1259
p = self.pending_merges()
1260
if revision_id in p:
1262
p.append(revision_id)
1263
self.set_pending_merges(p)
1266
def set_pending_merges(self, rev_list):
1267
from bzrlib.atomicfile import AtomicFile
1270
f = AtomicFile(self.controlfilename('pending-merges'))
1281
def get_parent(self):
1282
"""Return the parent location of the branch.
1284
This is the default location for push/pull/missing. The usual
1285
pattern is that the user can override it by specifying a
1289
_locs = ['parent', 'pull', 'x-pull']
1292
return self.controlfile(l, 'r').read().strip('\n')
1294
if e.errno != errno.ENOENT:
1299
def set_parent(self, url):
1300
# TODO: Maybe delete old location files?
1301
from bzrlib.atomicfile import AtomicFile
1304
f = AtomicFile(self.controlfilename('parent'))
1313
def check_revno(self, revno):
1315
Check whether a revno corresponds to any revision.
1316
Zero (the NULL revision) is considered valid.
1319
self.check_real_revno(revno)
1321
def check_real_revno(self, revno):
1323
Check whether a revno corresponds to a real revision.
1324
Zero (the NULL revision) is considered invalid
1326
if revno < 1 or revno > self.revno():
1327
raise InvalidRevisionNumber(revno)
597
>>> b = ScratchBranch(files=['foo'])
598
>>> b.basis_tree().has_filename('foo')
600
>>> b.working_tree().has_filename('foo')
603
>>> b.commit('add foo')
604
>>> b.basis_tree().has_filename('foo')
607
r = self.last_patch()
611
return RevisionTree(self.text_store, self.get_revision_inventory(r))
615
def write_log(self, utc=False):
616
"""Write out human-readable log of commits to this branch
618
:param utc: If true, show dates in universal time, not local time."""
619
## TODO: Option to choose either original, utc or local timezone
622
for p in self.revision_history():
624
print 'revno:', revno
625
## TODO: Show hash if --id is given.
626
##print 'revision-hash:', p
627
rev = self.get_revision(p)
628
print 'committer:', rev.committer
629
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0))
631
## opportunistic consistency check, same as check_patch_chaining
632
if rev.precursor != precursor:
633
bailout("mismatched precursor!")
637
print ' (no message)'
639
for l in rev.message.split('\n'):
647
def show_status(branch, show_all=False):
648
"""Display single-line status for non-ignored working files.
650
The list is show sorted in order by file name.
652
>>> b = ScratchBranch(files=['foo', 'foo~'])
658
>>> b.commit("add foo")
661
:todo: Get state for single files.
663
:todo: Perhaps show a slash at the end of directory names.
667
# We have to build everything into a list first so that it can
668
# sorted by name, incorporating all the different sources.
670
# FIXME: Rather than getting things in random order and then sorting,
671
# just step through in order.
673
# Interesting case: the old ID for a file has been removed,
674
# but a new file has been created under that name.
676
old = branch.basis_tree()
677
old_inv = old.inventory
678
new = branch.working_tree()
679
new_inv = new.inventory
681
for fs, fid, oldname, newname, kind in diff_trees(old, new):
683
show_status(fs, kind,
684
oldname + ' => ' + newname)
685
elif fs == 'A' or fs == 'M':
686
show_status(fs, kind, newname)
688
show_status(fs, kind, oldname)
691
show_status(fs, kind, newname)
694
show_status(fs, kind, newname)
696
show_status(fs, kind, newname)
698
bailout("wierd file state %r" % ((fs, fid),))
1332
702
class ScratchBranch(Branch):
1412
740
## mutter('check %r for control file' % ((head, tail), ))
1413
741
if tail == bzrlib.BZRDIR:
1415
if filename == head:
1422
def gen_file_id(name):
748
def _gen_revision_id(when):
749
"""Return new revision-id."""
750
s = '%s-%s-' % (user_email(), compact_date(when))
751
s += hexlify(rand_bytes(8))
755
def _gen_file_id(name):
1423
756
"""Return new file id.
1425
758
This should probably generate proper UUIDs, but for the moment we
1426
759
cope with just randomness because running uuidgen every time is
1429
from binascii import hexlify
1430
from time import time
1432
# get last component
1433
idx = name.rfind('/')
1435
name = name[idx+1 : ]
1436
idx = name.rfind('\\')
1438
name = name[idx+1 : ]
1440
# make it not a hidden file
1441
name = name.lstrip('.')
1443
# remove any wierd characters; we don't escape them but rather
1444
# just pull them out
1445
name = re.sub(r'[^\w.]', '', name)
761
assert '/' not in name
762
while name[0] == '.':
1447
764
s = hexlify(rand_bytes(8))
1448
return '-'.join((name, compact_date(time()), s))
1452
"""Return a new tree-root file id."""
1453
return gen_file_id('TREE_ROOT')
1456
def pull_loc(branch):
1457
# TODO: Should perhaps just make attribute be 'base' in
1458
# RemoteBranch and Branch?
1459
if hasattr(branch, "baseurl"):
1460
return branch.baseurl
1465
def copy_branch(branch_from, to_location, revision=None):
1466
"""Copy branch_from into the existing directory to_location.
1469
If not None, only revisions up to this point will be copied.
1470
The head of the new branch will be that revision.
1473
The name of a local directory that exists but is empty.
1475
from bzrlib.merge import merge
1476
from bzrlib.branch import Branch
1478
assert isinstance(branch_from, Branch)
1479
assert isinstance(to_location, basestring)
1481
br_to = Branch(to_location, init=True)
1482
br_to.set_root_id(branch_from.get_root_id())
1483
if revision is None:
1484
revno = branch_from.revno()
1486
revno, rev_id = branch_from.get_revision_info(revision)
1487
br_to.update_revisions(branch_from, stop_revision=revno)
1488
merge((to_location, -1), (to_location, 0), this_dir=to_location,
1489
check_clean=False, ignore_zero=True)
1491
from_location = pull_loc(branch_from)
1492
br_to.set_parent(pull_loc(branch_from))
765
return '-'.join((name, compact_date(time.time()), s))