15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
21
import traceback, socket, fnmatch, difflib, time
22
from binascii import hexlify
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,
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
33
from bzrlib.inventory import Inventory
34
from bzrlib.weavestore import WeaveStore
40
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
41
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
25
from inventory import Inventory
26
from trace import mutter, note
27
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
28
from inventory import InventoryEntry, Inventory
29
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, chomp, \
30
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
31
joinpath, sha_string, file_kind, local_time_offset, appendpath
32
from store import ImmutableStore
33
from revision import Revision
34
from errors import bailout
35
from textui import show_status
36
from diff import diff_trees
38
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
42
39
## TODO: Maybe include checks for common corruption of newlines, etc?
45
# TODO: Some operations like log might retrieve the same revisions
46
# repeatedly to calculate deltas. We could perhaps have a weakref
47
# cache in memory to make this faster.
49
# TODO: please move the revision-string syntax stuff out of the branch
50
# object; it's clutter
53
def find_branch(f, **args):
54
if f and (f.startswith('http://') or f.startswith('https://')):
56
return remotebranch.RemoteBranch(f, **args)
58
return Branch(f, **args)
61
def find_cached_branch(f, cache_root, **args):
62
from remotebranch import RemoteBranch
63
br = find_branch(f, **args)
64
def cacheify(br, store_name):
65
from meta_store import CachedStore
66
cache_path = os.path.join(cache_root, store_name)
68
new_store = CachedStore(getattr(br, store_name), cache_path)
69
setattr(br, store_name, new_store)
71
if isinstance(br, RemoteBranch):
72
cacheify(br, 'inventory_store')
73
cacheify(br, 'text_store')
74
cacheify(br, 'revision_store')
78
def _relpath(base, path):
79
"""Return path relative to base, or raise exception.
81
The path may be either an absolute path or a path relative to the
82
current working directory.
84
Lifted out of Branch.relpath for ease of testing.
86
os.path.commonprefix (python2.4) has a bad bug that it works just
87
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
88
avoids that problem."""
89
rp = os.path.abspath(path)
93
while len(head) >= len(base):
96
head, tail = os.path.split(head)
100
from errors import NotBranchError
101
raise NotBranchError("path %r is not within branch %r" % (rp, base))
103
return os.sep.join(s)
106
43
def find_branch_root(f=None):
107
44
"""Find the branch root enclosing f, or pwd.
109
f may be a filename or a URL.
111
46
It is not necessary that f exists.
113
48
Basically we keep looking up until we find the control directory or
114
run into the root. If there isn't one, raises NotBranchError.
118
52
elif hasattr(os.path, 'realpath'):
119
53
f = os.path.realpath(f)
121
55
f = os.path.abspath(f)
122
if not os.path.exists(f):
123
raise BzrError('%r does not exist' % f)
129
61
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
131
63
head, tail = os.path.split(f)
133
65
# reached the root, whatever that may be
134
raise bzrlib.errors.NotBranchError('%s is not in a branch' % orig_f)
66
bailout('%r is not in a branch' % orig_f)
139
# XXX: move into bzrlib.errors; subclass BzrError
140
class DivergedBranches(Exception):
141
def __init__(self, branch1, branch2):
142
self.branch1 = branch1
143
self.branch2 = branch2
144
Exception.__init__(self, "These branches have diverged.")
147
71
######################################################################
150
class Branch(object):
151
75
"""Branch holding a history of revisions.
154
Base directory of the branch.
160
If _lock_mode is true, a positive count of the number of times the
164
Lock object from bzrlib.lock.
77
:todo: Perhaps use different stores for different classes of object,
78
so that we can keep track of how much space each one uses,
79
or garbage-collect them.
81
:todo: Add a RemoteBranch subclass. For the basic case of read-only
82
HTTP access this should be very easy by,
83
just redirecting controlfile access into HTTP requests.
84
We would need a RemoteStore working similarly.
86
:todo: Keep the on-disk branch locked while the object exists.
88
:todo: mkdir() method.
171
# Map some sort of prefix into a namespace
172
# stuff like "revno:10", "revid:", etc.
173
# This should match a prefix with a function which accepts
174
REVISION_NAMESPACES = {}
176
90
def __init__(self, base, init=False, find_root=True):
177
91
"""Create new branch object at a particular location.
179
base -- Base directory for the branch.
93
:param base: Base directory for the branch.
181
init -- If True, create new control files in a previously
95
:param init: If True, create new control files in a previously
182
96
unversioned directory. If False, the branch must already
185
find_root -- If true and init is false, find the root of the
99
:param find_root: If true and init is false, find the root of the
186
100
existing branch containing base.
188
102
In the test suite, creation of new trees is tested using the
189
103
`ScratchBranch` class.
191
from bzrlib.store import ImmutableStore
193
106
self.base = os.path.realpath(base)
194
107
self._make_control()
215
127
__repr__ = __str__
219
if self._lock_mode or self._lock:
220
from warnings import warn
221
warn("branch %r was not explicitly unlocked" % self)
225
def lock_write(self):
227
if self._lock_mode != 'w':
228
from errors import LockError
229
raise LockError("can't upgrade to a write lock from %r" %
231
self._lock_count += 1
233
from bzrlib.lock import WriteLock
235
self._lock = WriteLock(self.controlfilename('branch-lock'))
236
self._lock_mode = 'w'
242
assert self._lock_mode in ('r', 'w'), \
243
"invalid lock mode %r" % self._lock_mode
244
self._lock_count += 1
246
from bzrlib.lock import ReadLock
248
self._lock = ReadLock(self.controlfilename('branch-lock'))
249
self._lock_mode = 'r'
253
if not self._lock_mode:
254
from errors import LockError
255
raise LockError('branch %r is not locked' % (self))
257
if self._lock_count > 1:
258
self._lock_count -= 1
262
self._lock_mode = self._lock_count = None
264
130
def abspath(self, name):
265
131
"""Return absolute filename for something in the branch"""
266
132
return os.path.join(self.base, name)
268
135
def relpath(self, path):
269
136
"""Return path relative to this branch of something inside it.
271
138
Raises an error if path is not in this branch."""
272
return _relpath(self.base, path)
139
rp = os.path.realpath(path)
141
if not rp.startswith(self.base):
142
bailout("path %r is not within branch %r" % (rp, self.base))
143
rp = rp[len(self.base):]
144
rp = rp.lstrip(os.sep)
274
148
def controlfilename(self, file_or_path):
275
149
"""Return location relative to branch."""
276
if isinstance(file_or_path, basestring):
150
if isinstance(file_or_path, types.StringTypes):
277
151
file_or_path = [file_or_path]
278
152
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
281
155
def controlfile(self, file_or_path, mode='r'):
282
"""Open a control file for this branch.
284
There are two classes of file in the control directory: text
285
and binary. binary files are untranslated byte streams. Text
286
control files are stored with Unix newlines and in UTF-8, even
287
if the platform or locale defaults are different.
289
Controlfiles should almost never be opened in write mode but
290
rather should be atomically copied and replaced using atomicfile.
293
fn = self.controlfilename(file_or_path)
295
if mode == 'rb' or mode == 'wb':
296
return file(fn, mode)
297
elif mode == 'r' or mode == 'w':
298
# open in binary mode anyhow so there's no newline translation;
299
# codecs uses line buffering by default; don't want that.
301
return codecs.open(fn, mode + 'b', 'utf-8',
304
raise BzrError("invalid controlfile mode %r" % mode)
156
"""Open a control file for this branch"""
157
return file(self.controlfilename(file_or_path), mode)
306
160
def _make_control(self):
307
161
os.mkdir(self.controlfilename([]))
308
162
self.controlfile('README', 'w').write(
309
163
"This is a Bazaar-NG control directory.\n"
310
"Do not change any files in this directory.\n")
311
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT_5)
312
for d in ('text-store', 'inventory-store', 'revision-store',
164
"Do not change any files in this directory.")
165
self.controlfile('branch-format', 'wb').write(BZR_BRANCH_FORMAT)
166
for d in ('text-store', 'inventory-store', 'revision-store'):
314
167
os.mkdir(self.controlfilename(d))
315
168
for f in ('revision-history', 'merged-patches',
316
'pending-merged-patches', 'branch-name',
169
'pending-merged-patches', 'branch-name'):
319
170
self.controlfile(f, 'w').write('')
320
171
mutter('created control directory in ' + self.base)
322
# if we want per-tree root ids then this is the place to set
323
# them; they're not needed for now and so ommitted for
325
f = self.controlfile('inventory','w')
326
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), f)
172
Inventory().write_xml(self.controlfile('inventory','w'))
329
175
def _check_format(self):
330
176
"""Check this branch format is supported.
332
The format level is stored, as an integer, in
333
self._branch_format for code that needs to check it later.
178
The current tool only supports the current unstable format.
335
180
In the future, we might need different in-memory Branch
336
181
classes to support downlevel branches. But not yet.
338
fmt = self.controlfile('branch-format', 'r').read()
339
if fmt == BZR_BRANCH_FORMAT_5:
340
self._branch_format = 5
342
raise BzrError('sorry, branch format "%s" not supported; '
343
'use a different bzr version, '
344
'or run "bzr upgrade", '
345
'or remove the .bzr directory and "bzr init" again'
346
% fmt.rstrip('\n\r'))
348
def get_root_id(self):
349
"""Return the id of this branches root"""
350
inv = self.read_working_inventory()
351
return inv.root.file_id
353
def set_root_id(self, file_id):
354
inv = self.read_working_inventory()
355
orig_root_id = inv.root.file_id
356
del inv._byid[inv.root.file_id]
357
inv.root.file_id = file_id
358
inv._byid[inv.root.file_id] = inv.root
361
if entry.parent_id in (None, orig_root_id):
362
entry.parent_id = inv.root.file_id
363
self._write_inventory(inv)
183
# This ignores newlines so that we can open branches created
184
# on Windows from Linux and so on. I think it might be better
185
# to always make all internal files in unix format.
186
fmt = self.controlfile('branch-format', 'rb').read()
187
fmt.replace('\r\n', '')
188
if fmt != BZR_BRANCH_FORMAT:
189
bailout('sorry, branch format %r not supported' % fmt,
190
['use a different bzr version',
191
'or remove the .bzr directory and "bzr init" again'])
365
194
def read_working_inventory(self):
366
195
"""Read the working inventory."""
369
# ElementTree does its own conversion from UTF-8, so open in
371
f = self.controlfile('inventory', 'rb')
372
return bzrlib.xml5.serializer_v5.read_inventory(f)
197
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
198
mutter("loaded inventory of %d items in %f"
199
% (len(inv), time.time() - before))
377
203
def _write_inventory(self, inv):
378
204
"""Update the working inventory.
380
206
That is to say, the inventory describing changes underway, that
381
207
will be committed to the next revision.
383
from bzrlib.atomicfile import AtomicFile
387
f = AtomicFile(self.controlfilename('inventory'), 'wb')
389
bzrlib.xml5.serializer_v5.write_inventory(inv, f)
209
## TODO: factor out to atomicfile? is rename safe on windows?
210
## TODO: Maybe some kind of clean/dirty marker on inventory?
211
tmpfname = self.controlfilename('inventory.tmp')
212
tmpf = file(tmpfname, 'w')
215
inv_fname = self.controlfilename('inventory')
216
if sys.platform == 'win32':
218
os.rename(tmpfname, inv_fname)
396
219
mutter('wrote working inventory')
399
222
inventory = property(read_working_inventory, _write_inventory, None,
400
223
"""Inventory for the working copy.""")
403
def add(self, files, ids=None):
226
def add(self, files, verbose=False):
404
227
"""Make files versioned.
406
Note that the command line normally calls smart_add instead,
407
which can automatically recurse.
409
229
This puts the files in the Added state, so that they will be
410
230
recorded by the next commit.
413
List of paths to add, relative to the base of the tree.
416
If set, use these instead of automatically generated ids.
417
Must be the same length as the list of files, but may
418
contain None for ids that are to be autogenerated.
420
TODO: Perhaps have an option to add the ids even if the files do
423
TODO: Perhaps yield the ids and paths as they're added.
232
:todo: Perhaps have an option to add the ids even if the files do
235
:todo: Perhaps return the ids of the files? But then again it
236
is easy to retrieve them if they're needed.
238
:todo: Option to specify file id.
240
:todo: Adding a directory should optionally recurse down and
241
add all non-ignored children. Perhaps do that in a
244
>>> b = ScratchBranch(files=['foo'])
245
>>> 'foo' in b.unknowns()
250
>>> 'foo' in b.unknowns()
252
>>> bool(b.inventory.path2id('foo'))
258
Traceback (most recent call last):
260
BzrError: ('foo is already versioned', [])
262
>>> b.add(['nothere'])
263
Traceback (most recent call last):
264
BzrError: ('cannot add: not a regular file or directory: nothere', [])
425
267
# TODO: Re-adding a file that is removed in the working copy
426
268
# should probably put it back with the previous ID.
427
if isinstance(files, basestring):
428
assert(ids is None or isinstance(ids, basestring))
269
if isinstance(files, types.StringTypes):
434
ids = [None] * len(files)
436
assert(len(ids) == len(files))
440
inv = self.read_working_inventory()
441
for f,file_id in zip(files, ids):
442
if is_control_file(f):
443
raise BzrError("cannot add control file %s" % quotefn(f))
448
raise BzrError("cannot add top-level %r" % f)
450
fullpath = os.path.normpath(self.abspath(f))
453
kind = file_kind(fullpath)
455
# maybe something better?
456
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
458
if kind != 'file' and kind != 'directory':
459
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
462
file_id = gen_file_id(f)
463
inv.add_path(f, kind=kind, file_id=file_id)
465
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
467
self._write_inventory(inv)
472
def print_file(self, file, revno):
473
"""Print `file` to stdout."""
476
tree = self.revision_tree(self.lookup_revision(revno))
477
# use inventory as it was in that revision
478
file_id = tree.inventory.path2id(file)
480
raise BzrError("%r is not present in revision %s" % (file, revno))
481
tree.print_file(file_id)
272
inv = self.read_working_inventory()
274
if is_control_file(f):
275
bailout("cannot add control file %s" % quotefn(f))
280
bailout("cannot add top-level %r" % f)
282
fullpath = os.path.normpath(self.abspath(f))
285
kind = file_kind(fullpath)
287
# maybe something better?
288
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
290
if kind != 'file' and kind != 'directory':
291
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
293
file_id = gen_file_id(f)
294
inv.add_path(f, kind=kind, file_id=file_id)
297
show_status('A', kind, quotefn(f))
299
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
301
self._write_inventory(inv)
486
305
def remove(self, files, verbose=False):
489
308
This does not remove their text. This does not run on
491
TODO: Refuse to remove modified files unless --force is given?
493
TODO: Do something useful with directories.
495
TODO: Should this remove the text or not? Tough call; not
310
:todo: Refuse to remove modified files unless --force is given?
312
>>> b = ScratchBranch(files=['foo'])
314
>>> b.inventory.has_filename('foo')
317
>>> b.working_tree().has_filename('foo')
319
>>> b.inventory.has_filename('foo')
322
>>> b = ScratchBranch(files=['foo'])
327
>>> b.inventory.has_filename('foo')
329
>>> b.basis_tree().has_filename('foo')
331
>>> b.working_tree().has_filename('foo')
334
:todo: Do something useful with directories.
336
:todo: Should this remove the text or not? Tough call; not
496
337
removing may be useful and the user can just use use rm, and
497
338
is the opposite of add. Removing it is consistent with most
498
339
other tools. Maybe an option.
500
341
## TODO: Normalize names
501
342
## TODO: Remove nested loops; better scalability
502
if isinstance(files, basestring):
344
if isinstance(files, types.StringTypes):
508
tree = self.working_tree()
511
# do this before any modifications
515
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
516
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
518
# having remove it, it must be either ignored or unknown
519
if tree.is_ignored(f):
523
show_status(new_status, inv[fid].kind, quotefn(f))
526
self._write_inventory(inv)
531
# FIXME: this doesn't need to be a branch method
532
def set_inventory(self, new_inventory_list):
533
from bzrlib.inventory import Inventory, InventoryEntry
534
inv = Inventory(self.get_root_id())
535
for path, file_id, parent, kind in new_inventory_list:
536
name = os.path.basename(path)
539
inv.add(InventoryEntry(file_id, name, kind, parent))
347
tree = self.working_tree()
350
# do this before any modifications
354
bailout("cannot remove unversioned file %s" % quotefn(f))
355
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
357
# having remove it, it must be either ignored or unknown
358
if tree.is_ignored(f):
362
show_status(new_status, inv[fid].kind, quotefn(f))
540
365
self._write_inventory(inv)
559
384
return self.working_tree().unknowns()
562
def append_revision(self, *revision_ids):
563
from bzrlib.atomicfile import AtomicFile
565
for revision_id in revision_ids:
566
mutter("add {%s} to revision-history" % revision_id)
568
rev_history = self.revision_history()
569
rev_history.extend(revision_ids)
571
f = AtomicFile(self.controlfilename('revision-history'))
573
for rev_id in rev_history:
580
def get_revision_xml_file(self, revision_id):
581
"""Return XML file object for revision object."""
582
if not revision_id or not isinstance(revision_id, basestring):
583
raise InvalidRevisionId(revision_id)
588
return self.revision_store[revision_id]
590
raise bzrlib.errors.NoSuchRevision(self, revision_id)
596
get_revision_xml = get_revision_xml_file
387
def commit(self, message, timestamp=None, timezone=None,
390
"""Commit working copy as a new revision.
392
The basic approach is to add all the file texts into the
393
store, then the inventory, then make a new revision pointing
394
to that inventory and store that.
396
This is not quite safe if the working copy changes during the
397
commit; for the moment that is simply not allowed. A better
398
approach is to make a temporary copy of the files before
399
computing their hashes, and then add those hashes in turn to
400
the inventory. This should mean at least that there are no
401
broken hash pointers. There is no way we can get a snapshot
402
of the whole directory at an instant. This would also have to
403
be robust against files disappearing, moving, etc. So the
404
whole thing is a bit hard.
406
:param timestamp: if not None, seconds-since-epoch for a
407
postdated/predated commit.
410
## TODO: Show branch names
412
# TODO: Don't commit if there are no changes, unless forced?
414
# First walk over the working inventory; and both update that
415
# and also build a new revision inventory. The revision
416
# inventory needs to hold the text-id, sha1 and size of the
417
# actual file versions committed in the revision. (These are
418
# not present in the working inventory.) We also need to
419
# detect missing/deleted files, and remove them from the
422
work_inv = self.read_working_inventory()
424
basis = self.basis_tree()
425
basis_inv = basis.inventory
427
for path, entry in work_inv.iter_entries():
428
## TODO: Cope with files that have gone missing.
430
## TODO: Check that the file kind has not changed from the previous
431
## revision of this file (if any).
435
p = self.abspath(path)
436
file_id = entry.file_id
437
mutter('commit prep file %s, id %r ' % (p, file_id))
439
if not os.path.exists(p):
440
mutter(" file is missing, removing from inventory")
442
show_status('D', entry.kind, quotefn(path))
443
missing_ids.append(file_id)
446
# TODO: Handle files that have been deleted
448
# TODO: Maybe a special case for empty files? Seems a
449
# waste to store them many times.
453
if basis_inv.has_id(file_id):
454
old_kind = basis_inv[file_id].kind
455
if old_kind != entry.kind:
456
bailout("entry %r changed kind from %r to %r"
457
% (file_id, old_kind, entry.kind))
459
if entry.kind == 'directory':
461
bailout("%s is entered as directory but not a directory" % quotefn(p))
462
elif entry.kind == 'file':
464
bailout("%s is entered as file but is not a file" % quotefn(p))
466
content = file(p, 'rb').read()
468
entry.text_sha1 = sha_string(content)
469
entry.text_size = len(content)
471
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
473
and (old_ie.text_size == entry.text_size)
474
and (old_ie.text_sha1 == entry.text_sha1)):
475
## assert content == basis.get_file(file_id).read()
476
entry.text_id = basis_inv[file_id].text_id
477
mutter(' unchanged from previous text_id {%s}' %
481
entry.text_id = gen_file_id(entry.name)
482
self.text_store.add(content, entry.text_id)
483
mutter(' stored with text_id {%s}' % entry.text_id)
487
elif (old_ie.name == entry.name
488
and old_ie.parent_id == entry.parent_id):
493
show_status(state, entry.kind, quotefn(path))
495
for file_id in missing_ids:
496
# have to do this later so we don't mess up the iterator.
497
# since parents may be removed before their children we
500
# FIXME: There's probably a better way to do this; perhaps
501
# the workingtree should know how to filter itself.
502
if work_inv.has_id(file_id):
503
del work_inv[file_id]
506
inv_id = rev_id = _gen_revision_id(time.time())
508
inv_tmp = tempfile.TemporaryFile()
509
inv.write_xml(inv_tmp)
511
self.inventory_store.add(inv_tmp, inv_id)
512
mutter('new inventory_id is {%s}' % inv_id)
514
self._write_inventory(work_inv)
516
if timestamp == None:
517
timestamp = time.time()
519
if committer == None:
520
committer = username()
523
timezone = local_time_offset()
525
mutter("building commit log message")
526
rev = Revision(timestamp=timestamp,
529
precursor = self.last_patch(),
534
rev_tmp = tempfile.TemporaryFile()
535
rev.write_xml(rev_tmp)
537
self.revision_store.add(rev_tmp, rev_id)
538
mutter("new revision_id is {%s}" % rev_id)
540
## XXX: Everything up to here can simply be orphaned if we abort
541
## the commit; it will leave junk files behind but that doesn't
544
## TODO: Read back the just-generated changeset, and make sure it
545
## applies and recreates the right state.
547
## TODO: Also calculate and store the inventory SHA1
548
mutter("committing patch r%d" % (self.revno() + 1))
550
mutter("append to revision-history")
551
f = self.controlfile('revision-history', 'at')
552
f.write(rev_id + '\n')
556
note("commited r%d" % self.revno())
599
559
def get_revision(self, revision_id):
600
560
"""Return the Revision object for a named revision"""
601
xml_file = self.get_revision_xml_file(revision_id)
604
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
605
except SyntaxError, e:
606
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
561
r = Revision.read_xml(self.revision_store[revision_id])
610
562
assert r.revision_id == revision_id
614
def get_revision_delta(self, revno):
615
"""Return the delta for one revision.
617
The delta is relative to its mainline predecessor, or the
618
empty tree for revision 1.
620
assert isinstance(revno, int)
621
rh = self.revision_history()
622
if not (1 <= revno <= len(rh)):
623
raise InvalidRevisionNumber(revno)
625
# revno is 1-based; list is 0-based
627
new_tree = self.revision_tree(rh[revno-1])
629
old_tree = EmptyTree()
631
old_tree = self.revision_tree(rh[revno-2])
633
return compare_trees(old_tree, new_tree)
637
def get_revision_sha1(self, revision_id):
638
"""Hash the stored value of a revision, and return it."""
639
# In the future, revision entries will be signed. At that
640
# point, it is probably best *not* to include the signature
641
# in the revision hash. Because that lets you re-sign
642
# the revision, (add signatures/remove signatures) and still
643
# have all hash pointers stay consistent.
644
# But for now, just hash the contents.
645
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
648
def get_inventory(self, revision_id):
566
def get_inventory(self, inventory_id):
649
567
"""Get Inventory object by hash.
651
TODO: Perhaps for this and similar methods, take a revision
569
:todo: Perhaps for this and similar methods, take a revision
652
570
parameter which can be either an integer revno or a
654
f = self.get_inventory_xml_file(revision_id)
655
return bzrlib.xml5.serializer_v5.read_inventory(f)
658
def get_inventory_xml(self, revision_id):
659
"""Get inventory XML as a file object."""
661
assert isinstance(revision_id, basestring), type(revision_id)
662
return self.inventory_store[revision_id]
664
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
666
get_inventory_xml_file = get_inventory_xml
669
def get_inventory_sha1(self, revision_id):
670
"""Return the sha1 hash of the inventory entry
672
return sha_file(self.get_inventory_xml_file(revision_id))
572
i = Inventory.read_xml(self.inventory_store[inventory_id])
675
576
def get_revision_inventory(self, revision_id):
676
577
"""Return inventory of a past revision."""
677
# bzr 0.0.6 imposes the constraint that the inventory_id
678
# must be the same as its revision, so this is trivial.
679
578
if revision_id == None:
680
return Inventory(self.get_root_id())
682
return self.get_inventory(revision_id)
581
return self.get_inventory(self.get_revision(revision_id).inventory_id)
685
584
def revision_history(self):
747
596
That is equivalent to the number of revisions committed to
599
>>> b = ScratchBranch()
602
>>> b.commit('no foo')
750
606
return len(self.revision_history())
753
609
def last_patch(self):
754
610
"""Return last patch hash, or None if no history.
612
>>> ScratchBranch().last_patch() == None
756
615
ph = self.revision_history()
763
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
765
If self and other have not diverged, return a list of the revisions
766
present in other, but missing from self.
768
>>> from bzrlib.commit import commit
769
>>> bzrlib.trace.silent = True
770
>>> br1 = ScratchBranch()
771
>>> br2 = ScratchBranch()
772
>>> br1.missing_revisions(br2)
774
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
775
>>> br1.missing_revisions(br2)
777
>>> br2.missing_revisions(br1)
779
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
780
>>> br1.missing_revisions(br2)
782
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
783
>>> br1.missing_revisions(br2)
785
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
786
>>> br1.missing_revisions(br2)
787
Traceback (most recent call last):
788
DivergedBranches: These branches have diverged.
790
self_history = self.revision_history()
791
self_len = len(self_history)
792
other_history = other.revision_history()
793
other_len = len(other_history)
794
common_index = min(self_len, other_len) -1
795
if common_index >= 0 and \
796
self_history[common_index] != other_history[common_index]:
797
raise DivergedBranches(self, other)
799
if stop_revision is None:
800
stop_revision = other_len
801
elif stop_revision > other_len:
802
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
804
return other_history[self_len:stop_revision]
807
def update_revisions(self, other, stop_revision=None):
808
"""Pull in all new revisions from other branch.
810
from bzrlib.fetch import greedy_fetch
812
pb = bzrlib.ui.ui_factory.progress_bar()
813
pb.update('comparing histories')
815
revision_ids = self.missing_revisions(other, stop_revision)
817
if len(revision_ids) > 0:
818
count = greedy_fetch(self, other, revision_ids[-1], pb)[0]
821
self.append_revision(*revision_ids)
822
## note("Added %d revisions." % count)
825
def install_revisions(self, other, revision_ids, pb):
826
if hasattr(other.revision_store, "prefetch"):
827
other.revision_store.prefetch(revision_ids)
828
if hasattr(other.inventory_store, "prefetch"):
829
inventory_ids = [other.get_revision(r).inventory_id
830
for r in revision_ids]
831
other.inventory_store.prefetch(inventory_ids)
834
pb = bzrlib.ui.ui_factory.progress_bar()
841
for i, rev_id in enumerate(revision_ids):
842
pb.update('fetching revision', i+1, len(revision_ids))
844
rev = other.get_revision(rev_id)
845
except bzrlib.errors.NoSuchRevision:
849
revisions.append(rev)
850
inv = other.get_inventory(str(rev.inventory_id))
851
for key, entry in inv.iter_entries():
852
if entry.text_id is None:
854
if entry.text_id not in self.text_store:
855
needed_texts.add(entry.text_id)
859
count, cp_fail = self.text_store.copy_multi(other.text_store,
861
#print "Added %d texts." % count
862
inventory_ids = [ f.inventory_id for f in revisions ]
863
count, cp_fail = self.inventory_store.copy_multi(other.inventory_store,
865
#print "Added %d inventories." % count
866
revision_ids = [ f.revision_id for f in revisions]
868
count, cp_fail = self.revision_store.copy_multi(other.revision_store,
871
assert len(cp_fail) == 0
872
return count, failures
875
def commit(self, *args, **kw):
876
from bzrlib.commit import Commit
877
Commit().commit(self, *args, **kw)
880
def lookup_revision(self, revision):
881
"""Return the revision identifier for a given revision information."""
882
revno, info = self._get_revision_info(revision)
886
def revision_id_to_revno(self, revision_id):
887
"""Given a revision id, return its revno"""
888
history = self.revision_history()
890
return history.index(revision_id) + 1
892
raise bzrlib.errors.NoSuchRevision(self, revision_id)
895
def get_revision_info(self, revision):
896
"""Return (revno, revision id) for revision identifier.
898
revision can be an integer, in which case it is assumed to be revno (though
899
this will translate negative values into positive ones)
900
revision can also be a string, in which case it is parsed for something like
901
'date:' or 'revid:' etc.
903
revno, rev_id = self._get_revision_info(revision)
905
raise bzrlib.errors.NoSuchRevision(self, revision)
908
def get_rev_id(self, revno, history=None):
909
"""Find the revision id of the specified revno."""
620
def lookup_revision(self, revno):
621
"""Return revision hash for revision number."""
913
history = self.revision_history()
914
elif revno <= 0 or revno > len(history):
915
raise bzrlib.errors.NoSuchRevision(self, revno)
916
return history[revno - 1]
918
def _get_revision_info(self, revision):
919
"""Return (revno, revision id) for revision specifier.
921
revision can be an integer, in which case it is assumed to be revno
922
(though this will translate negative values into positive ones)
923
revision can also be a string, in which case it is parsed for something
924
like 'date:' or 'revid:' etc.
926
A revid is always returned. If it is None, the specifier referred to
927
the null revision. If the revid does not occur in the revision
928
history, revno will be None.
934
try:# Convert to int if possible
935
revision = int(revision)
938
revs = self.revision_history()
939
if isinstance(revision, int):
941
revno = len(revs) + revision + 1
944
rev_id = self.get_rev_id(revno, revs)
945
elif isinstance(revision, basestring):
946
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
947
if revision.startswith(prefix):
948
result = func(self, revs, revision)
950
revno, rev_id = result
953
rev_id = self.get_rev_id(revno, revs)
956
raise BzrError('No namespace registered for string: %r' %
959
raise TypeError('Unhandled revision type %s' % revision)
963
raise bzrlib.errors.NoSuchRevision(self, revision)
966
def _namespace_revno(self, revs, revision):
967
"""Lookup a revision by revision number"""
968
assert revision.startswith('revno:')
970
return (int(revision[6:]),)
973
REVISION_NAMESPACES['revno:'] = _namespace_revno
975
def _namespace_revid(self, revs, revision):
976
assert revision.startswith('revid:')
977
rev_id = revision[len('revid:'):]
979
return revs.index(rev_id) + 1, rev_id
982
REVISION_NAMESPACES['revid:'] = _namespace_revid
984
def _namespace_last(self, revs, revision):
985
assert revision.startswith('last:')
987
offset = int(revision[5:])
992
raise BzrError('You must supply a positive value for --revision last:XXX')
993
return (len(revs) - offset + 1,)
994
REVISION_NAMESPACES['last:'] = _namespace_last
996
def _namespace_tag(self, revs, revision):
997
assert revision.startswith('tag:')
998
raise BzrError('tag: namespace registered, but not implemented.')
999
REVISION_NAMESPACES['tag:'] = _namespace_tag
1001
def _namespace_date(self, revs, revision):
1002
assert revision.startswith('date:')
1004
# Spec for date revisions:
1006
# value can be 'yesterday', 'today', 'tomorrow' or a YYYY-MM-DD string.
1007
# it can also start with a '+/-/='. '+' says match the first
1008
# entry after the given date. '-' is match the first entry before the date
1009
# '=' is match the first entry after, but still on the given date.
1011
# +2005-05-12 says find the first matching entry after May 12th, 2005 at 0:00
1012
# -2005-05-12 says find the first matching entry before May 12th, 2005 at 0:00
1013
# =2005-05-12 says find the first match after May 12th, 2005 at 0:00 but before
1014
# May 13th, 2005 at 0:00
1016
# So the proper way of saying 'give me all entries for today' is:
1017
# -r {date:+today}:{date:-tomorrow}
1018
# The default is '=' when not supplied
1021
if val[:1] in ('+', '-', '='):
1022
match_style = val[:1]
1025
today = datetime.datetime.today().replace(hour=0,minute=0,second=0,microsecond=0)
1026
if val.lower() == 'yesterday':
1027
dt = today - datetime.timedelta(days=1)
1028
elif val.lower() == 'today':
1030
elif val.lower() == 'tomorrow':
1031
dt = today + datetime.timedelta(days=1)
1034
# This should be done outside the function to avoid recompiling it.
1035
_date_re = re.compile(
1036
r'(?P<date>(?P<year>\d\d\d\d)-(?P<month>\d\d)-(?P<day>\d\d))?'
1038
r'(?P<time>(?P<hour>\d\d):(?P<minute>\d\d)(:(?P<second>\d\d))?)?'
1040
m = _date_re.match(val)
1041
if not m or (not m.group('date') and not m.group('time')):
1042
raise BzrError('Invalid revision date %r' % revision)
1045
year, month, day = int(m.group('year')), int(m.group('month')), int(m.group('day'))
1047
year, month, day = today.year, today.month, today.day
1049
hour = int(m.group('hour'))
1050
minute = int(m.group('minute'))
1051
if m.group('second'):
1052
second = int(m.group('second'))
1056
hour, minute, second = 0,0,0
1058
dt = datetime.datetime(year=year, month=month, day=day,
1059
hour=hour, minute=minute, second=second)
1063
if match_style == '-':
1065
elif match_style == '=':
1066
last = dt + datetime.timedelta(days=1)
1069
for i in range(len(revs)-1, -1, -1):
1070
r = self.get_revision(revs[i])
1071
# TODO: Handle timezone.
1072
dt = datetime.datetime.fromtimestamp(r.timestamp)
1073
if first >= dt and (last is None or dt >= last):
1076
for i in range(len(revs)):
1077
r = self.get_revision(revs[i])
1078
# TODO: Handle timezone.
1079
dt = datetime.datetime.fromtimestamp(r.timestamp)
1080
if first <= dt and (last is None or dt <= last):
1082
REVISION_NAMESPACES['date:'] = _namespace_date
626
# list is 0-based; revisions are 1-based
627
return self.revision_history()[revno-1]
629
bailout("no such revision %s" % revno)
1084
632
def revision_tree(self, revision_id):
1085
633
"""Return Tree for a revision on this branch.
1087
635
`revision_id` may be None for the null revision, in which case
1088
636
an `EmptyTree` is returned."""
1089
# TODO: refactor this to use an existing revision object
1090
# so we don't need to read it in twice.
1091
638
if revision_id == None:
1092
639
return EmptyTree()
1094
641
inv = self.get_revision_inventory(revision_id)
1095
return RevisionTree(self.weave_store, inv, revision_id)
642
return RevisionTree(self.text_store, inv)
1098
645
def working_tree(self):
1099
646
"""Return a `Tree` for the working copy."""
1100
from workingtree import WorkingTree
1101
647
return WorkingTree(self.base, self.read_working_inventory())
1105
651
"""Return `Tree` object for last revision.
1107
653
If there are no revisions yet, return an `EmptyTree`.
1109
return self.revision_tree(self.last_patch())
1112
def rename_one(self, from_rel, to_rel):
1115
This can change the directory or the filename or both.
1119
tree = self.working_tree()
1120
inv = tree.inventory
1121
if not tree.has_filename(from_rel):
1122
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1123
if tree.has_filename(to_rel):
1124
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1126
file_id = inv.path2id(from_rel)
1128
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1130
if inv.path2id(to_rel):
1131
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1133
to_dir, to_tail = os.path.split(to_rel)
1134
to_dir_id = inv.path2id(to_dir)
1135
if to_dir_id == None and to_dir != '':
1136
raise BzrError("can't determine destination directory id for %r" % to_dir)
1138
mutter("rename_one:")
1139
mutter(" file_id {%s}" % file_id)
1140
mutter(" from_rel %r" % from_rel)
1141
mutter(" to_rel %r" % to_rel)
1142
mutter(" to_dir %r" % to_dir)
1143
mutter(" to_dir_id {%s}" % to_dir_id)
1145
inv.rename(file_id, to_dir_id, to_tail)
1147
from_abs = self.abspath(from_rel)
1148
to_abs = self.abspath(to_rel)
1150
os.rename(from_abs, to_abs)
1152
raise BzrError("failed to rename %r to %r: %s"
1153
% (from_abs, to_abs, e[1]),
1154
["rename rolled back"])
1156
self._write_inventory(inv)
1161
def move(self, from_paths, to_name):
655
>>> b = ScratchBranch(files=['foo'])
656
>>> b.basis_tree().has_filename('foo')
658
>>> b.working_tree().has_filename('foo')
661
>>> b.commit('add foo')
662
>>> b.basis_tree().has_filename('foo')
665
r = self.last_patch()
669
return RevisionTree(self.text_store, self.get_revision_inventory(r))
673
def write_log(self, show_timezone='original'):
674
"""Write out human-readable log of commits to this branch
676
:param utc: If true, show dates in universal time, not local time."""
677
## TODO: Option to choose either original, utc or local timezone
680
for p in self.revision_history():
682
print 'revno:', revno
683
## TODO: Show hash if --id is given.
684
##print 'revision-hash:', p
685
rev = self.get_revision(p)
686
print 'committer:', rev.committer
687
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
690
## opportunistic consistency check, same as check_patch_chaining
691
if rev.precursor != precursor:
692
bailout("mismatched precursor!")
696
print ' (no message)'
698
for l in rev.message.split('\n'):
706
def rename(self, from_paths, to_name):
1162
707
"""Rename files.
1164
to_name must exist as a versioned directory.
1166
709
If to_name exists and is a directory, the files are moved into
1167
710
it, keeping their old names. If it is a directory,
1169
712
Note that to_name is only the last component of the new name;
1170
713
this doesn't change the directory.
1172
This returns a list of (from_path, to_path) pairs for each
1173
entry that is moved.
1178
## TODO: Option to move IDs only
1179
assert not isinstance(from_paths, basestring)
1180
tree = self.working_tree()
1181
inv = tree.inventory
1182
to_abs = self.abspath(to_name)
1183
if not isdir(to_abs):
1184
raise BzrError("destination %r is not a directory" % to_abs)
1185
if not tree.has_filename(to_name):
1186
raise BzrError("destination %r not in working directory" % to_abs)
1187
to_dir_id = inv.path2id(to_name)
1188
if to_dir_id == None and to_name != '':
1189
raise BzrError("destination %r is not a versioned directory" % to_name)
1190
to_dir_ie = inv[to_dir_id]
1191
if to_dir_ie.kind not in ('directory', 'root_directory'):
1192
raise BzrError("destination %r is not a directory" % to_abs)
1194
to_idpath = inv.get_idpath(to_dir_id)
1196
for f in from_paths:
1197
if not tree.has_filename(f):
1198
raise BzrError("%r does not exist in working tree" % f)
1199
f_id = inv.path2id(f)
1201
raise BzrError("%r is not versioned" % f)
1202
name_tail = splitpath(f)[-1]
1203
dest_path = appendpath(to_name, name_tail)
1204
if tree.has_filename(dest_path):
1205
raise BzrError("destination %r already exists" % dest_path)
1206
if f_id in to_idpath:
1207
raise BzrError("can't move %r to a subdirectory of itself" % f)
1209
# OK, so there's a race here, it's possible that someone will
1210
# create a file in this interval and then the rename might be
1211
# left half-done. But we should have caught most problems.
1213
for f in from_paths:
1214
name_tail = splitpath(f)[-1]
1215
dest_path = appendpath(to_name, name_tail)
1216
result.append((f, dest_path))
1217
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1219
os.rename(self.abspath(f), self.abspath(dest_path))
1221
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1222
["rename rolled back"])
715
## TODO: Option to move IDs only
716
assert not isinstance(from_paths, basestring)
717
tree = self.working_tree()
719
dest_dir = isdir(self.abspath(to_name))
721
# TODO: Wind back properly if some can't be moved?
722
dest_dir_id = inv.path2id(to_name)
723
if not dest_dir_id and to_name != '':
724
bailout("destination %r is not a versioned directory" % to_name)
726
name_tail = splitpath(f)[-1]
727
dest_path = appendpath(to_name, name_tail)
728
print "%s => %s" % (f, dest_path)
729
inv.rename(inv.path2id(f), dest_dir_id, name_tail)
730
os.rename(self.abspath(f), self.abspath(dest_path))
1224
731
self._write_inventory(inv)
1231
def revert(self, filenames, old_tree=None, backups=True):
1232
"""Restore selected files to the versions from a previous tree.
1235
If true (default) backups are made of files before
1238
from bzrlib.errors import NotVersionedError, BzrError
1239
from bzrlib.atomicfile import AtomicFile
1240
from bzrlib.osutils import backup_file
1242
inv = self.read_working_inventory()
1243
if old_tree is None:
1244
old_tree = self.basis_tree()
1245
old_inv = old_tree.inventory
1248
for fn in filenames:
1249
file_id = inv.path2id(fn)
1251
raise NotVersionedError("not a versioned file", fn)
1252
if not old_inv.has_id(file_id):
1253
raise BzrError("file not present in old tree", fn, file_id)
1254
nids.append((fn, file_id))
1256
# TODO: Rename back if it was previously at a different location
1258
# TODO: If given a directory, restore the entire contents from
1259
# the previous version.
1261
# TODO: Make a backup to a temporary file.
1263
# TODO: If the file previously didn't exist, delete it?
1264
for fn, file_id in nids:
1267
f = AtomicFile(fn, 'wb')
1269
f.write(old_tree.get_file(file_id).read())
1275
def pending_merges(self):
1276
"""Return a list of pending merges.
1278
These are revisions that have been merged into the working
1279
directory but not yet committed.
1281
cfn = self.controlfilename('pending-merges')
1282
if not os.path.exists(cfn):
1285
for l in self.controlfile('pending-merges', 'r').readlines():
1286
p.append(l.rstrip('\n'))
1290
def add_pending_merge(self, revision_id):
1291
from bzrlib.revision import validate_revision_id
1293
validate_revision_id(revision_id)
1295
p = self.pending_merges()
1296
if revision_id in p:
1298
p.append(revision_id)
1299
self.set_pending_merges(p)
1302
def set_pending_merges(self, rev_list):
1303
from bzrlib.atomicfile import AtomicFile
1306
f = AtomicFile(self.controlfilename('pending-merges'))
1317
def get_parent(self):
1318
"""Return the parent location of the branch.
1320
This is the default location for push/pull/missing. The usual
1321
pattern is that the user can override it by specifying a
1325
_locs = ['parent', 'pull', 'x-pull']
1328
return self.controlfile(l, 'r').read().strip('\n')
1330
if e.errno != errno.ENOENT:
1335
def set_parent(self, url):
1336
# TODO: Maybe delete old location files?
1337
from bzrlib.atomicfile import AtomicFile
1340
f = AtomicFile(self.controlfilename('parent'))
1349
def check_revno(self, revno):
1351
Check whether a revno corresponds to any revision.
1352
Zero (the NULL revision) is considered valid.
1355
self.check_real_revno(revno)
1357
def check_real_revno(self, revno):
1359
Check whether a revno corresponds to a real revision.
1360
Zero (the NULL revision) is considered invalid
1362
if revno < 1 or revno > self.revno():
1363
raise InvalidRevisionNumber(revno)
733
if len(from_paths) != 1:
734
bailout("when moving multiple files, destination must be a directory")
735
bailout("rename to non-directory %r not implemented sorry" % to_name)
739
def show_status(branch, show_all=False):
740
"""Display single-line status for non-ignored working files.
742
The list is show sorted in order by file name.
744
>>> b = ScratchBranch(files=['foo', 'foo~'])
750
>>> b.commit("add foo")
752
>>> os.unlink(b.abspath('foo'))
757
:todo: Get state for single files.
759
:todo: Perhaps show a slash at the end of directory names.
763
# We have to build everything into a list first so that it can
764
# sorted by name, incorporating all the different sources.
766
# FIXME: Rather than getting things in random order and then sorting,
767
# just step through in order.
769
# Interesting case: the old ID for a file has been removed,
770
# but a new file has been created under that name.
772
old = branch.basis_tree()
773
old_inv = old.inventory
774
new = branch.working_tree()
775
new_inv = new.inventory
777
for fs, fid, oldname, newname, kind in diff_trees(old, new):
779
show_status(fs, kind,
780
oldname + ' => ' + newname)
781
elif fs == 'A' or fs == 'M':
782
show_status(fs, kind, newname)
784
show_status(fs, kind, oldname)
787
show_status(fs, kind, newname)
790
show_status(fs, kind, newname)
792
show_status(fs, kind, newname)
794
bailout("wierd file state %r" % ((fs, fid),))
1368
798
class ScratchBranch(Branch):