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
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
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.xml import unpack_xml
32
from bzrlib.delta import compare_trees
33
from bzrlib.tree import EmptyTree, RevisionTree
38
38
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
39
39
## TODO: Maybe include checks for common corruption of newlines, etc?
42
# TODO: Some operations like log might retrieve the same revisions
43
# repeatedly to calculate deltas. We could perhaps have a weakref
44
# cache in memory to make this faster.
46
# TODO: please move the revision-string syntax stuff out of the branch
47
# object; it's clutter
50
def find_branch(f, **args):
51
if f and (f.startswith('http://') or f.startswith('https://')):
53
return remotebranch.RemoteBranch(f, **args)
55
return Branch(f, **args)
58
def find_cached_branch(f, cache_root, **args):
59
from remotebranch import RemoteBranch
60
br = find_branch(f, **args)
61
def cacheify(br, store_name):
62
from meta_store import CachedStore
63
cache_path = os.path.join(cache_root, store_name)
65
new_store = CachedStore(getattr(br, store_name), cache_path)
66
setattr(br, store_name, new_store)
68
if isinstance(br, RemoteBranch):
69
cacheify(br, 'inventory_store')
70
cacheify(br, 'text_store')
71
cacheify(br, 'revision_store')
75
def _relpath(base, path):
76
"""Return path relative to base, or raise exception.
78
The path may be either an absolute path or a path relative to the
79
current working directory.
81
Lifted out of Branch.relpath for ease of testing.
83
os.path.commonprefix (python2.4) has a bad bug that it works just
84
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
85
avoids that problem."""
86
rp = os.path.abspath(path)
90
while len(head) >= len(base):
93
head, tail = os.path.split(head)
97
from errors import NotBranchError
98
raise NotBranchError("path %r is not within branch %r" % (rp, base))
100
return os.sep.join(s)
43
103
def find_branch_root(f=None):
44
104
"""Find the branch root enclosing f, or pwd.
106
f may be a filename or a URL.
46
108
It is not necessary that f exists.
48
110
Basically we keep looking up until we find the control directory or
111
run into the root. If there isn't one, raises NotBranchError.
52
115
elif hasattr(os.path, 'realpath'):
53
116
f = os.path.realpath(f)
55
118
f = os.path.abspath(f)
119
if not os.path.exists(f):
120
raise BzrError('%r does not exist' % f)
61
126
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
63
128
head, tail = os.path.split(f)
65
130
# reached the root, whatever that may be
66
bailout('%r is not in a branch' % orig_f)
131
raise bzrlib.errors.NotBranchError('%s is not in a branch' % orig_f)
136
# XXX: move into bzrlib.errors; subclass BzrError
137
class DivergedBranches(Exception):
138
def __init__(self, branch1, branch2):
139
self.branch1 = branch1
140
self.branch2 = branch2
141
Exception.__init__(self, "These branches have diverged.")
71
144
######################################################################
147
class Branch(object):
75
148
"""Branch holding a history of revisions.
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.
151
Base directory of the branch.
157
If _lock_mode is true, a positive count of the number of times the
161
Lock object from bzrlib.lock.
168
# Map some sort of prefix into a namespace
169
# stuff like "revno:10", "revid:", etc.
170
# This should match a prefix with a function which accepts
171
REVISION_NAMESPACES = {}
90
173
def __init__(self, base, init=False, find_root=True):
91
174
"""Create new branch object at a particular location.
93
:param base: Base directory for the branch.
176
base -- Base directory for the branch.
95
:param init: If True, create new control files in a previously
178
init -- If True, create new control files in a previously
96
179
unversioned directory. If False, the branch must already
99
:param find_root: If true and init is false, find the root of the
182
find_root -- If true and init is false, find the root of the
100
183
existing branch containing base.
102
185
In the test suite, creation of new trees is tested using the
103
186
`ScratchBranch` class.
188
from bzrlib.store import ImmutableStore
106
190
self.base = os.path.realpath(base)
107
191
self._make_control()
127
212
__repr__ = __str__
216
if self._lock_mode or self._lock:
217
from warnings import warn
218
warn("branch %r was not explicitly unlocked" % self)
222
def lock_write(self):
224
if self._lock_mode != 'w':
225
from errors import LockError
226
raise LockError("can't upgrade to a write lock from %r" %
228
self._lock_count += 1
230
from bzrlib.lock import WriteLock
232
self._lock = WriteLock(self.controlfilename('branch-lock'))
233
self._lock_mode = 'w'
239
assert self._lock_mode in ('r', 'w'), \
240
"invalid lock mode %r" % self._lock_mode
241
self._lock_count += 1
243
from bzrlib.lock import ReadLock
245
self._lock = ReadLock(self.controlfilename('branch-lock'))
246
self._lock_mode = 'r'
250
if not self._lock_mode:
251
from errors import LockError
252
raise LockError('branch %r is not locked' % (self))
254
if self._lock_count > 1:
255
self._lock_count -= 1
259
self._lock_mode = self._lock_count = None
130
261
def abspath(self, name):
131
262
"""Return absolute filename for something in the branch"""
132
263
return os.path.join(self.base, name)
135
265
def relpath(self, path):
136
266
"""Return path relative to this branch of something inside it.
138
268
Raises an error if path is not in this branch."""
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)
269
return _relpath(self.base, path)
148
271
def controlfilename(self, file_or_path):
149
272
"""Return location relative to branch."""
150
if isinstance(file_or_path, types.StringTypes):
273
if isinstance(file_or_path, basestring):
151
274
file_or_path = [file_or_path]
152
275
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
155
278
def controlfile(self, file_or_path, mode='r'):
156
"""Open a control file for this branch"""
157
return file(self.controlfilename(file_or_path), mode)
279
"""Open a control file for this branch.
281
There are two classes of file in the control directory: text
282
and binary. binary files are untranslated byte streams. Text
283
control files are stored with Unix newlines and in UTF-8, even
284
if the platform or locale defaults are different.
286
Controlfiles should almost never be opened in write mode but
287
rather should be atomically copied and replaced using atomicfile.
290
fn = self.controlfilename(file_or_path)
292
if mode == 'rb' or mode == 'wb':
293
return file(fn, mode)
294
elif mode == 'r' or mode == 'w':
295
# open in binary mode anyhow so there's no newline translation;
296
# codecs uses line buffering by default; don't want that.
298
return codecs.open(fn, mode + 'b', 'utf-8',
301
raise BzrError("invalid controlfile mode %r" % mode)
160
303
def _make_control(self):
304
from bzrlib.inventory import Inventory
305
from bzrlib.xml import pack_xml
161
307
os.mkdir(self.controlfilename([]))
162
308
self.controlfile('README', 'w').write(
163
309
"This is a Bazaar-NG control directory.\n"
164
"Do not change any files in this directory.")
165
self.controlfile('branch-format', 'wb').write(BZR_BRANCH_FORMAT)
310
"Do not change any files in this directory.\n")
311
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
166
312
for d in ('text-store', 'inventory-store', 'revision-store'):
167
313
os.mkdir(self.controlfilename(d))
168
314
for f in ('revision-history', 'merged-patches',
169
'pending-merged-patches', 'branch-name'):
315
'pending-merged-patches', 'branch-name',
170
318
self.controlfile(f, 'w').write('')
171
319
mutter('created control directory in ' + self.base)
172
Inventory().write_xml(self.controlfile('inventory','w'))
321
# if we want per-tree root ids then this is the place to set
322
# them; they're not needed for now and so ommitted for
324
pack_xml(Inventory(), self.controlfile('inventory','w'))
175
326
def _check_format(self):
176
327
"""Check this branch format is supported.
183
334
# This ignores newlines so that we can open branches created
184
335
# on Windows from Linux and so on. I think it might be better
185
336
# to always make all internal files in unix format.
186
fmt = self.controlfile('branch-format', 'rb').read()
337
fmt = self.controlfile('branch-format', 'r').read()
187
338
fmt.replace('\r\n', '')
188
339
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'])
340
raise BzrError('sorry, branch format %r not supported' % fmt,
341
['use a different bzr version',
342
'or remove the .bzr directory and "bzr init" again'])
344
def get_root_id(self):
345
"""Return the id of this branches root"""
346
inv = self.read_working_inventory()
347
return inv.root.file_id
349
def set_root_id(self, file_id):
350
inv = self.read_working_inventory()
351
orig_root_id = inv.root.file_id
352
del inv._byid[inv.root.file_id]
353
inv.root.file_id = file_id
354
inv._byid[inv.root.file_id] = inv.root
357
if entry.parent_id in (None, orig_root_id):
358
entry.parent_id = inv.root.file_id
359
self._write_inventory(inv)
194
361
def read_working_inventory(self):
195
362
"""Read the working inventory."""
197
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
198
mutter("loaded inventory of %d items in %f"
199
% (len(inv), time.time() - before))
363
from bzrlib.inventory import Inventory
364
from bzrlib.xml import unpack_xml
365
from time import time
369
# ElementTree does its own conversion from UTF-8, so open in
371
inv = unpack_xml(Inventory,
372
self.controlfile('inventory', 'rb'))
373
mutter("loaded inventory of %d items in %f"
374
% (len(inv), time() - before))
203
380
def _write_inventory(self, inv):
204
381
"""Update the working inventory.
206
383
That is to say, the inventory describing changes underway, that
207
384
will be committed to the next revision.
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)
386
from bzrlib.atomicfile import AtomicFile
387
from bzrlib.xml import pack_xml
391
f = AtomicFile(self.controlfilename('inventory'), 'wb')
219
400
mutter('wrote working inventory')
222
403
inventory = property(read_working_inventory, _write_inventory, None,
223
404
"""Inventory for the working copy.""")
226
def add(self, files, verbose=False):
407
def add(self, files, verbose=False, ids=None):
227
408
"""Make files versioned.
410
Note that the command line normally calls smart_add instead.
229
412
This puts the files in the Added state, so that they will be
230
413
recorded by the next commit.
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', [])
416
List of paths to add, relative to the base of the tree.
419
If set, use these instead of automatically generated ids.
420
Must be the same length as the list of files, but may
421
contain None for ids that are to be autogenerated.
423
TODO: Perhaps have an option to add the ids even if the files do
426
TODO: Perhaps return the ids of the files? But then again it
427
is easy to retrieve them if they're needed.
429
TODO: Adding a directory should optionally recurse down and
430
add all non-ignored children. Perhaps do that in a
267
433
# TODO: Re-adding a file that is removed in the working copy
268
434
# should probably put it back with the previous ID.
269
if isinstance(files, types.StringTypes):
435
if isinstance(files, basestring):
436
assert(ids is None or isinstance(ids, basestring))
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)
442
ids = [None] * len(files)
444
assert(len(ids) == len(files))
448
inv = self.read_working_inventory()
449
for f,file_id in zip(files, ids):
450
if is_control_file(f):
451
raise BzrError("cannot add control file %s" % quotefn(f))
456
raise BzrError("cannot add top-level %r" % f)
458
fullpath = os.path.normpath(self.abspath(f))
461
kind = file_kind(fullpath)
463
# maybe something better?
464
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
466
if kind != 'file' and kind != 'directory':
467
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
470
file_id = gen_file_id(f)
471
inv.add_path(f, kind=kind, file_id=file_id)
474
print 'added', quotefn(f)
476
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
478
self._write_inventory(inv)
483
def print_file(self, file, revno):
484
"""Print `file` to stdout."""
487
tree = self.revision_tree(self.lookup_revision(revno))
488
# use inventory as it was in that revision
489
file_id = tree.inventory.path2id(file)
491
raise BzrError("%r is not present in revision %s" % (file, revno))
492
tree.print_file(file_id)
305
497
def remove(self, files, verbose=False):
308
500
This does not remove their text. This does not run on
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
502
TODO: Refuse to remove modified files unless --force is given?
504
TODO: Do something useful with directories.
506
TODO: Should this remove the text or not? Tough call; not
337
507
removing may be useful and the user can just use use rm, and
338
508
is the opposite of add. Removing it is consistent with most
339
509
other tools. Maybe an option.
341
511
## TODO: Normalize names
342
512
## TODO: Remove nested loops; better scalability
344
if isinstance(files, types.StringTypes):
513
if isinstance(files, basestring):
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))
519
tree = self.working_tree()
522
# do this before any modifications
526
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
527
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
529
# having remove it, it must be either ignored or unknown
530
if tree.is_ignored(f):
534
show_status(new_status, inv[fid].kind, quotefn(f))
537
self._write_inventory(inv)
542
# FIXME: this doesn't need to be a branch method
543
def set_inventory(self, new_inventory_list):
544
from bzrlib.inventory import Inventory, InventoryEntry
545
inv = Inventory(self.get_root_id())
546
for path, file_id, parent, kind in new_inventory_list:
547
name = os.path.basename(path)
550
inv.add(InventoryEntry(file_id, name, kind, parent))
365
551
self._write_inventory(inv)
384
570
return self.working_tree().unknowns()
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())
573
def append_revision(self, *revision_ids):
574
from bzrlib.atomicfile import AtomicFile
576
for revision_id in revision_ids:
577
mutter("add {%s} to revision-history" % revision_id)
579
rev_history = self.revision_history()
580
rev_history.extend(revision_ids)
582
f = AtomicFile(self.controlfilename('revision-history'))
584
for rev_id in rev_history:
591
def get_revision_xml(self, revision_id):
592
"""Return XML file object for revision object."""
593
if not revision_id or not isinstance(revision_id, basestring):
594
raise InvalidRevisionId(revision_id)
599
return self.revision_store[revision_id]
601
raise bzrlib.errors.NoSuchRevision(self, revision_id)
559
606
def get_revision(self, revision_id):
560
607
"""Return the Revision object for a named revision"""
561
r = Revision.read_xml(self.revision_store[revision_id])
608
xml_file = self.get_revision_xml(revision_id)
611
r = unpack_xml(Revision, xml_file)
612
except SyntaxError, e:
613
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
562
617
assert r.revision_id == revision_id
621
def get_revision_delta(self, revno):
622
"""Return the delta for one revision.
624
The delta is relative to its mainline predecessor, or the
625
empty tree for revision 1.
627
assert isinstance(revno, int)
628
rh = self.revision_history()
629
if not (1 <= revno <= len(rh)):
630
raise InvalidRevisionNumber(revno)
632
# revno is 1-based; list is 0-based
634
new_tree = self.revision_tree(rh[revno-1])
636
old_tree = EmptyTree()
638
old_tree = self.revision_tree(rh[revno-2])
640
return compare_trees(old_tree, new_tree)
644
def get_revision_sha1(self, revision_id):
645
"""Hash the stored value of a revision, and return it."""
646
# In the future, revision entries will be signed. At that
647
# point, it is probably best *not* to include the signature
648
# in the revision hash. Because that lets you re-sign
649
# the revision, (add signatures/remove signatures) and still
650
# have all hash pointers stay consistent.
651
# But for now, just hash the contents.
652
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
566
655
def get_inventory(self, inventory_id):
567
656
"""Get Inventory object by hash.
569
:todo: Perhaps for this and similar methods, take a revision
658
TODO: Perhaps for this and similar methods, take a revision
570
659
parameter which can be either an integer revno or a
572
i = Inventory.read_xml(self.inventory_store[inventory_id])
661
from bzrlib.inventory import Inventory
662
from bzrlib.xml import unpack_xml
664
return unpack_xml(Inventory, self.get_inventory_xml(inventory_id))
667
def get_inventory_xml(self, inventory_id):
668
"""Get inventory XML as a file object."""
669
return self.inventory_store[inventory_id]
672
def get_inventory_sha1(self, inventory_id):
673
"""Return the sha1 hash of the inventory entry
675
return sha_file(self.get_inventory_xml(inventory_id))
576
678
def get_revision_inventory(self, revision_id):
577
679
"""Return inventory of a past revision."""
680
# bzr 0.0.6 imposes the constraint that the inventory_id
681
# must be the same as its revision, so this is trivial.
578
682
if revision_id == None:
683
from bzrlib.inventory import Inventory
684
return Inventory(self.get_root_id())
581
return self.get_inventory(self.get_revision(revision_id).inventory_id)
686
return self.get_inventory(revision_id)
584
689
def revision_history(self):
596
751
That is equivalent to the number of revisions committed to
599
>>> b = ScratchBranch()
602
>>> b.commit('no foo')
606
754
return len(self.revision_history())
609
757
def last_patch(self):
610
758
"""Return last patch hash, or None if no history.
612
>>> ScratchBranch().last_patch() == None
615
760
ph = self.revision_history()
620
def lookup_revision(self, revno):
621
"""Return revision hash for revision number."""
626
# list is 0-based; revisions are 1-based
627
return self.revision_history()[revno-1]
629
bailout("no such revision %s" % revno)
767
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
769
If self and other have not diverged, return a list of the revisions
770
present in other, but missing from self.
772
>>> from bzrlib.commit import commit
773
>>> bzrlib.trace.silent = True
774
>>> br1 = ScratchBranch()
775
>>> br2 = ScratchBranch()
776
>>> br1.missing_revisions(br2)
778
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
779
>>> br1.missing_revisions(br2)
781
>>> br2.missing_revisions(br1)
783
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
784
>>> br1.missing_revisions(br2)
786
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
787
>>> br1.missing_revisions(br2)
789
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
790
>>> br1.missing_revisions(br2)
791
Traceback (most recent call last):
792
DivergedBranches: These branches have diverged.
794
self_history = self.revision_history()
795
self_len = len(self_history)
796
other_history = other.revision_history()
797
other_len = len(other_history)
798
common_index = min(self_len, other_len) -1
799
if common_index >= 0 and \
800
self_history[common_index] != other_history[common_index]:
801
raise DivergedBranches(self, other)
803
if stop_revision is None:
804
stop_revision = other_len
805
elif stop_revision > other_len:
806
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
808
return other_history[self_len:stop_revision]
811
def update_revisions(self, other, stop_revision=None):
812
"""Pull in all new revisions from other branch.
814
from bzrlib.fetch import greedy_fetch
816
pb = bzrlib.ui.ui_factory.progress_bar()
817
pb.update('comparing histories')
819
revision_ids = self.missing_revisions(other, stop_revision)
821
if len(revision_ids) > 0:
822
count = greedy_fetch(self, other, revision_ids[-1], pb)[0]
825
self.append_revision(*revision_ids)
826
## note("Added %d revisions." % count)
829
def install_revisions(self, other, revision_ids, pb):
830
if hasattr(other.revision_store, "prefetch"):
831
other.revision_store.prefetch(revision_ids)
832
if hasattr(other.inventory_store, "prefetch"):
833
inventory_ids = [other.get_revision(r).inventory_id
834
for r in revision_ids]
835
other.inventory_store.prefetch(inventory_ids)
838
pb = bzrlib.ui.ui_factory.progress_bar()
845
for i, rev_id in enumerate(revision_ids):
846
pb.update('fetching revision', i+1, len(revision_ids))
848
rev = other.get_revision(rev_id)
849
except bzrlib.errors.NoSuchRevision:
853
revisions.append(rev)
854
inv = other.get_inventory(str(rev.inventory_id))
855
for key, entry in inv.iter_entries():
856
if entry.text_id is None:
858
if entry.text_id not in self.text_store:
859
needed_texts.add(entry.text_id)
863
count, cp_fail = self.text_store.copy_multi(other.text_store,
865
print "Added %d texts." % count
866
inventory_ids = [ f.inventory_id for f in revisions ]
867
count, cp_fail = self.inventory_store.copy_multi(other.inventory_store,
869
print "Added %d inventories." % count
870
revision_ids = [ f.revision_id for f in revisions]
872
count, cp_fail = self.revision_store.copy_multi(other.revision_store,
875
assert len(cp_fail) == 0
876
return count, failures
879
def commit(self, *args, **kw):
880
from bzrlib.commit import commit
881
commit(self, *args, **kw)
884
def lookup_revision(self, revision):
885
"""Return the revision identifier for a given revision information."""
886
revno, info = self.get_revision_info(revision)
890
def revision_id_to_revno(self, revision_id):
891
"""Given a revision id, return its revno"""
892
history = self.revision_history()
894
return history.index(revision_id) + 1
896
raise bzrlib.errors.NoSuchRevision(self, revision_id)
899
def get_revision_info(self, revision):
900
"""Return (revno, revision id) for revision identifier.
902
revision can be an integer, in which case it is assumed to be revno (though
903
this will translate negative values into positive ones)
904
revision can also be a string, in which case it is parsed for something like
905
'date:' or 'revid:' etc.
910
try:# Convert to int if possible
911
revision = int(revision)
914
revs = self.revision_history()
915
if isinstance(revision, int):
918
# Mabye we should do this first, but we don't need it if revision == 0
920
revno = len(revs) + revision + 1
923
elif isinstance(revision, basestring):
924
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
925
if revision.startswith(prefix):
926
revno = func(self, revs, revision)
929
raise BzrError('No namespace registered for string: %r' % revision)
931
if revno is None or revno <= 0 or revno > len(revs):
932
raise BzrError("no such revision %s" % revision)
933
return revno, revs[revno-1]
935
def _namespace_revno(self, revs, revision):
936
"""Lookup a revision by revision number"""
937
assert revision.startswith('revno:')
939
return int(revision[6:])
942
REVISION_NAMESPACES['revno:'] = _namespace_revno
944
def _namespace_revid(self, revs, revision):
945
assert revision.startswith('revid:')
947
return revs.index(revision[6:]) + 1
950
REVISION_NAMESPACES['revid:'] = _namespace_revid
952
def _namespace_last(self, revs, revision):
953
assert revision.startswith('last:')
955
offset = int(revision[5:])
960
raise BzrError('You must supply a positive value for --revision last:XXX')
961
return len(revs) - offset + 1
962
REVISION_NAMESPACES['last:'] = _namespace_last
964
def _namespace_tag(self, revs, revision):
965
assert revision.startswith('tag:')
966
raise BzrError('tag: namespace registered, but not implemented.')
967
REVISION_NAMESPACES['tag:'] = _namespace_tag
969
def _namespace_date(self, revs, revision):
970
assert revision.startswith('date:')
972
# Spec for date revisions:
974
# value can be 'yesterday', 'today', 'tomorrow' or a YYYY-MM-DD string.
975
# it can also start with a '+/-/='. '+' says match the first
976
# entry after the given date. '-' is match the first entry before the date
977
# '=' is match the first entry after, but still on the given date.
979
# +2005-05-12 says find the first matching entry after May 12th, 2005 at 0:00
980
# -2005-05-12 says find the first matching entry before May 12th, 2005 at 0:00
981
# =2005-05-12 says find the first match after May 12th, 2005 at 0:00 but before
982
# May 13th, 2005 at 0:00
984
# So the proper way of saying 'give me all entries for today' is:
985
# -r {date:+today}:{date:-tomorrow}
986
# The default is '=' when not supplied
989
if val[:1] in ('+', '-', '='):
990
match_style = val[:1]
993
today = datetime.datetime.today().replace(hour=0,minute=0,second=0,microsecond=0)
994
if val.lower() == 'yesterday':
995
dt = today - datetime.timedelta(days=1)
996
elif val.lower() == 'today':
998
elif val.lower() == 'tomorrow':
999
dt = today + datetime.timedelta(days=1)
1002
# This should be done outside the function to avoid recompiling it.
1003
_date_re = re.compile(
1004
r'(?P<date>(?P<year>\d\d\d\d)-(?P<month>\d\d)-(?P<day>\d\d))?'
1006
r'(?P<time>(?P<hour>\d\d):(?P<minute>\d\d)(:(?P<second>\d\d))?)?'
1008
m = _date_re.match(val)
1009
if not m or (not m.group('date') and not m.group('time')):
1010
raise BzrError('Invalid revision date %r' % revision)
1013
year, month, day = int(m.group('year')), int(m.group('month')), int(m.group('day'))
1015
year, month, day = today.year, today.month, today.day
1017
hour = int(m.group('hour'))
1018
minute = int(m.group('minute'))
1019
if m.group('second'):
1020
second = int(m.group('second'))
1024
hour, minute, second = 0,0,0
1026
dt = datetime.datetime(year=year, month=month, day=day,
1027
hour=hour, minute=minute, second=second)
1031
if match_style == '-':
1033
elif match_style == '=':
1034
last = dt + datetime.timedelta(days=1)
1037
for i in range(len(revs)-1, -1, -1):
1038
r = self.get_revision(revs[i])
1039
# TODO: Handle timezone.
1040
dt = datetime.datetime.fromtimestamp(r.timestamp)
1041
if first >= dt and (last is None or dt >= last):
1044
for i in range(len(revs)):
1045
r = self.get_revision(revs[i])
1046
# TODO: Handle timezone.
1047
dt = datetime.datetime.fromtimestamp(r.timestamp)
1048
if first <= dt and (last is None or dt <= last):
1050
REVISION_NAMESPACES['date:'] = _namespace_date
632
1052
def revision_tree(self, revision_id):
633
1053
"""Return Tree for a revision on this branch.
635
1055
`revision_id` may be None for the null revision, in which case
636
1056
an `EmptyTree` is returned."""
1057
# TODO: refactor this to use an existing revision object
1058
# so we don't need to read it in twice.
638
1059
if revision_id == None:
639
1060
return EmptyTree()
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'):
705
1085
def rename_one(self, from_rel, to_rel):
706
tree = self.working_tree()
708
if not tree.has_filename(from_rel):
709
bailout("can't rename: old working file %r does not exist" % from_rel)
710
if tree.has_filename(to_rel):
711
bailout("can't rename: new working file %r already exists" % to_rel)
713
file_id = inv.path2id(from_rel)
715
bailout("can't rename: old name %r is not versioned" % from_rel)
717
if inv.path2id(to_rel):
718
bailout("can't rename: new name %r is already versioned" % to_rel)
720
to_dir, to_tail = os.path.split(to_rel)
721
to_dir_id = inv.path2id(to_dir)
722
if to_dir_id == None and to_dir != '':
723
bailout("can't determine destination directory id for %r" % to_dir)
725
mutter("rename_one:")
726
mutter(" file_id {%s}" % file_id)
727
mutter(" from_rel %r" % from_rel)
728
mutter(" to_rel %r" % to_rel)
729
mutter(" to_dir %r" % to_dir)
730
mutter(" to_dir_id {%s}" % to_dir_id)
732
inv.rename(file_id, to_dir_id, to_tail)
733
os.rename(self.abspath(from_rel), self.abspath(to_rel))
735
self._write_inventory(inv)
739
def rename(self, from_paths, to_name):
1088
This can change the directory or the filename or both.
1092
tree = self.working_tree()
1093
inv = tree.inventory
1094
if not tree.has_filename(from_rel):
1095
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1096
if tree.has_filename(to_rel):
1097
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1099
file_id = inv.path2id(from_rel)
1101
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1103
if inv.path2id(to_rel):
1104
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1106
to_dir, to_tail = os.path.split(to_rel)
1107
to_dir_id = inv.path2id(to_dir)
1108
if to_dir_id == None and to_dir != '':
1109
raise BzrError("can't determine destination directory id for %r" % to_dir)
1111
mutter("rename_one:")
1112
mutter(" file_id {%s}" % file_id)
1113
mutter(" from_rel %r" % from_rel)
1114
mutter(" to_rel %r" % to_rel)
1115
mutter(" to_dir %r" % to_dir)
1116
mutter(" to_dir_id {%s}" % to_dir_id)
1118
inv.rename(file_id, to_dir_id, to_tail)
1120
print "%s => %s" % (from_rel, to_rel)
1122
from_abs = self.abspath(from_rel)
1123
to_abs = self.abspath(to_rel)
1125
os.rename(from_abs, to_abs)
1127
raise BzrError("failed to rename %r to %r: %s"
1128
% (from_abs, to_abs, e[1]),
1129
["rename rolled back"])
1131
self._write_inventory(inv)
1136
def move(self, from_paths, to_name):
740
1137
"""Rename files.
1139
to_name must exist as a versioned directory.
742
1141
If to_name exists and is a directory, the files are moved into
743
1142
it, keeping their old names. If it is a directory,
745
1144
Note that to_name is only the last component of the new name;
746
1145
this doesn't change the directory.
748
## TODO: Option to move IDs only
749
assert not isinstance(from_paths, basestring)
750
tree = self.working_tree()
752
dest_dir = isdir(self.abspath(to_name))
754
# TODO: Wind back properly if some can't be moved?
755
dest_dir_id = inv.path2id(to_name)
756
if not dest_dir_id and to_name != '':
757
bailout("destination %r is not a versioned directory" % to_name)
1149
## TODO: Option to move IDs only
1150
assert not isinstance(from_paths, basestring)
1151
tree = self.working_tree()
1152
inv = tree.inventory
1153
to_abs = self.abspath(to_name)
1154
if not isdir(to_abs):
1155
raise BzrError("destination %r is not a directory" % to_abs)
1156
if not tree.has_filename(to_name):
1157
raise BzrError("destination %r not in working directory" % to_abs)
1158
to_dir_id = inv.path2id(to_name)
1159
if to_dir_id == None and to_name != '':
1160
raise BzrError("destination %r is not a versioned directory" % to_name)
1161
to_dir_ie = inv[to_dir_id]
1162
if to_dir_ie.kind not in ('directory', 'root_directory'):
1163
raise BzrError("destination %r is not a directory" % to_abs)
1165
to_idpath = inv.get_idpath(to_dir_id)
1167
for f in from_paths:
1168
if not tree.has_filename(f):
1169
raise BzrError("%r does not exist in working tree" % f)
1170
f_id = inv.path2id(f)
1172
raise BzrError("%r is not versioned" % f)
1173
name_tail = splitpath(f)[-1]
1174
dest_path = appendpath(to_name, name_tail)
1175
if tree.has_filename(dest_path):
1176
raise BzrError("destination %r already exists" % dest_path)
1177
if f_id in to_idpath:
1178
raise BzrError("can't move %r to a subdirectory of itself" % f)
1180
# OK, so there's a race here, it's possible that someone will
1181
# create a file in this interval and then the rename might be
1182
# left half-done. But we should have caught most problems.
758
1184
for f in from_paths:
759
1185
name_tail = splitpath(f)[-1]
760
1186
dest_path = appendpath(to_name, name_tail)
761
1187
print "%s => %s" % (f, dest_path)
762
inv.rename(inv.path2id(f), dest_dir_id, name_tail)
763
os.rename(self.abspath(f), self.abspath(dest_path))
1188
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1190
os.rename(self.abspath(f), self.abspath(dest_path))
1192
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1193
["rename rolled back"])
764
1195
self._write_inventory(inv)
766
if len(from_paths) != 1:
767
bailout("when moving multiple files, destination must be a directory")
768
bailout("rename to non-directory %r not implemented sorry" % to_name)
772
def show_status(branch, show_all=False):
773
"""Display single-line status for non-ignored working files.
775
The list is show sorted in order by file name.
777
>>> b = ScratchBranch(files=['foo', 'foo~'])
783
>>> b.commit("add foo")
785
>>> os.unlink(b.abspath('foo'))
1200
def revert(self, filenames, old_tree=None, backups=True):
1201
"""Restore selected files to the versions from a previous tree.
1204
If true (default) backups are made of files before
1207
from bzrlib.errors import NotVersionedError, BzrError
1208
from bzrlib.atomicfile import AtomicFile
1209
from bzrlib.osutils import backup_file
790
:todo: Get state for single files.
792
:todo: Perhaps show a slash at the end of directory names.
1211
inv = self.read_working_inventory()
1212
if old_tree is None:
1213
old_tree = self.basis_tree()
1214
old_inv = old_tree.inventory
1217
for fn in filenames:
1218
file_id = inv.path2id(fn)
1220
raise NotVersionedError("not a versioned file", fn)
1221
if not old_inv.has_id(file_id):
1222
raise BzrError("file not present in old tree", fn, file_id)
1223
nids.append((fn, file_id))
1225
# TODO: Rename back if it was previously at a different location
1227
# TODO: If given a directory, restore the entire contents from
1228
# the previous version.
1230
# TODO: Make a backup to a temporary file.
1232
# TODO: If the file previously didn't exist, delete it?
1233
for fn, file_id in nids:
1236
f = AtomicFile(fn, 'wb')
1238
f.write(old_tree.get_file(file_id).read())
1244
def pending_merges(self):
1245
"""Return a list of pending merges.
1247
These are revisions that have been merged into the working
1248
directory but not yet committed.
796
# We have to build everything into a list first so that it can
797
# sorted by name, incorporating all the different sources.
799
# FIXME: Rather than getting things in random order and then sorting,
800
# just step through in order.
802
# Interesting case: the old ID for a file has been removed,
803
# but a new file has been created under that name.
805
old = branch.basis_tree()
806
old_inv = old.inventory
807
new = branch.working_tree()
808
new_inv = new.inventory
810
for fs, fid, oldname, newname, kind in diff_trees(old, new):
812
show_status(fs, kind,
813
oldname + ' => ' + newname)
814
elif fs == 'A' or fs == 'M':
815
show_status(fs, kind, newname)
817
show_status(fs, kind, oldname)
820
show_status(fs, kind, newname)
823
show_status(fs, kind, newname)
825
show_status(fs, kind, newname)
827
bailout("wierd file state %r" % ((fs, fid),))
1250
cfn = self.controlfilename('pending-merges')
1251
if not os.path.exists(cfn):
1254
for l in self.controlfile('pending-merges', 'r').readlines():
1255
p.append(l.rstrip('\n'))
1259
def add_pending_merge(self, revision_id):
1260
from bzrlib.revision import validate_revision_id
1262
validate_revision_id(revision_id)
1264
p = self.pending_merges()
1265
if revision_id in p:
1267
p.append(revision_id)
1268
self.set_pending_merges(p)
1271
def set_pending_merges(self, rev_list):
1272
from bzrlib.atomicfile import AtomicFile
1275
f = AtomicFile(self.controlfilename('pending-merges'))
831
1287
class ScratchBranch(Branch):
890
def _gen_revision_id(when):
891
"""Return new revision-id."""
892
s = '%s-%s-' % (user_email(), compact_date(when))
893
s += hexlify(rand_bytes(8))
897
1375
def gen_file_id(name):
898
1376
"""Return new file id.
900
1378
This should probably generate proper UUIDs, but for the moment we
901
1379
cope with just randomness because running uuidgen every time is
1382
from binascii import hexlify
1383
from time import time
1385
# get last component
903
1386
idx = name.rfind('/')
905
1388
name = name[idx+1 : ]
1389
idx = name.rfind('\\')
1391
name = name[idx+1 : ]
1393
# make it not a hidden file
907
1394
name = name.lstrip('.')
1396
# remove any wierd characters; we don't escape them but rather
1397
# just pull them out
1398
name = re.sub(r'[^\w.]', '', name)
909
1400
s = hexlify(rand_bytes(8))
910
return '-'.join((name, compact_date(time.time()), s))
1401
return '-'.join((name, compact_date(time()), s))
1405
"""Return a new tree-root file id."""
1406
return gen_file_id('TREE_ROOT')
1409
def pull_loc(branch):
1410
# TODO: Should perhaps just make attribute be 'base' in
1411
# RemoteBranch and Branch?
1412
if hasattr(branch, "baseurl"):
1413
return branch.baseurl
1418
def copy_branch(branch_from, to_location, revision=None):
1419
"""Copy branch_from into the existing directory to_location.
1421
If revision is not None, the head of the new branch will be revision.
1423
from bzrlib.merge import merge
1424
from bzrlib.branch import Branch
1425
br_to = Branch(to_location, init=True)
1426
br_to.set_root_id(branch_from.get_root_id())
1427
if revision is None:
1428
revno = branch_from.revno()
1430
revno, rev_id = branch_from.get_revision_info(revision)
1431
br_to.update_revisions(branch_from, stop_revision=revno)
1432
merge((to_location, -1), (to_location, 0), this_dir=to_location,
1433
check_clean=False, ignore_zero=True)
1434
from_location = pull_loc(branch_from)
1435
br_to.controlfile("x-pull", "wb").write(from_location + "\n")