15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21
from warnings import warn
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
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"
25
from bzrlib.trace import mutter, note
26
from bzrlib.osutils import (isdir, quotefn, compact_date, rand_bytes,
27
rename, splitpath, sha_file, appendpath,
29
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
30
NoSuchRevision, HistoryMissing, NotBranchError,
31
DivergedBranches, LockError)
32
from bzrlib.textui import show_status
33
from bzrlib.revision import Revision, validate_revision_id
34
from bzrlib.delta import compare_trees
35
from bzrlib.tree import EmptyTree, RevisionTree
36
from bzrlib.inventory import Inventory
37
from bzrlib.weavestore import WeaveStore
38
from bzrlib.store import ImmutableStore
43
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
44
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
39
45
## TODO: Maybe include checks for common corruption of newlines, etc?
48
# TODO: Some operations like log might retrieve the same revisions
49
# repeatedly to calculate deltas. We could perhaps have a weakref
50
# cache in memory to make this faster. In general anything can be
51
# cached in memory between lock and unlock operations.
53
def find_branch(*ignored, **ignored_too):
54
# XXX: leave this here for about one release, then remove it
55
raise NotImplementedError('find_branch() is not supported anymore, '
56
'please use one of the new branch constructors')
58
def _relpath(base, path):
59
"""Return path relative to base, or raise exception.
61
The path may be either an absolute path or a path relative to the
62
current working directory.
64
Lifted out of Branch.relpath for ease of testing.
66
os.path.commonprefix (python2.4) has a bad bug that it works just
67
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
68
avoids that problem."""
69
rp = os.path.abspath(path)
73
while len(head) >= len(base):
76
head, tail = os.path.split(head)
80
raise NotBranchError("path %r is not within branch %r" % (rp, base))
43
85
def find_branch_root(f=None):
44
86
"""Find the branch root enclosing f, or pwd.
88
f may be a filename or a URL.
46
90
It is not necessary that f exists.
48
92
Basically we keep looking up until we find the control directory or
93
run into the root. If there isn't one, raises NotBranchError.
52
97
elif hasattr(os.path, 'realpath'):
53
98
f = os.path.realpath(f)
55
100
f = os.path.abspath(f)
101
if not os.path.exists(f):
102
raise BzrError('%r does not exist' % f)
61
108
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
63
110
head, tail = os.path.split(f)
65
112
# reached the root, whatever that may be
66
bailout('%r is not in a branch' % orig_f)
113
raise NotBranchError('%s is not in a branch' % orig_f)
71
119
######################################################################
122
class Branch(object):
75
123
"""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.
90
def __init__(self, base, init=False, find_root=True):
126
Base directory/url of the branch.
130
def __init__(self, *ignored, **ignored_too):
131
raise NotImplementedError('The Branch class is abstract')
135
"""Open an existing branch, rooted at 'base' (url)"""
136
if base and (base.startswith('http://') or base.startswith('https://')):
137
from bzrlib.remotebranch import RemoteBranch
138
return RemoteBranch(base, find_root=False)
140
return LocalBranch(base, find_root=False)
143
def open_containing(url):
144
"""Open an existing branch which contains url.
146
This probes for a branch at url, and searches upwards from there.
148
if url and (url.startswith('http://') or url.startswith('https://')):
149
from bzrlib.remotebranch import RemoteBranch
150
return RemoteBranch(url)
152
return LocalBranch(url)
155
def initialize(base):
156
"""Create a new branch, rooted at 'base' (url)"""
157
if base and (base.startswith('http://') or base.startswith('https://')):
158
from bzrlib.remotebranch import RemoteBranch
159
return RemoteBranch(base, init=True)
161
return LocalBranch(base, init=True)
163
def setup_caching(self, cache_root):
164
"""Subclasses that care about caching should override this, and set
165
up cached stores located under cache_root.
169
class LocalBranch(Branch):
170
"""A branch stored in the actual filesystem.
172
Note that it's "local" in the context of the filesystem; it doesn't
173
really matter if it's on an nfs/smb/afs/coda/... share, as long as
174
it's writable, and can be accessed via the normal filesystem API.
180
If _lock_mode is true, a positive count of the number of times the
184
Lock object from bzrlib.lock.
186
# We actually expect this class to be somewhat short-lived; part of its
187
# purpose is to try to isolate what bits of the branch logic are tied to
188
# filesystem access, so that in a later step, we can extricate them to
189
# a separarte ("storage") class.
193
_inventory_weave = None
195
# Map some sort of prefix into a namespace
196
# stuff like "revno:10", "revid:", etc.
197
# This should match a prefix with a function which accepts
198
REVISION_NAMESPACES = {}
200
def __init__(self, base, init=False, find_root=True,
201
relax_version_check=False):
91
202
"""Create new branch object at a particular location.
93
:param base: Base directory for the branch.
204
base -- Base directory for the branch. May be a file:// url.
95
:param init: If True, create new control files in a previously
206
init -- If True, create new control files in a previously
96
207
unversioned directory. If False, the branch must already
99
:param find_root: If true and init is false, find the root of the
210
find_root -- If true and init is false, find the root of the
100
211
existing branch containing base.
213
relax_version_check -- If true, the usual check for the branch
214
version is not applied. This is intended only for
215
upgrade/recovery type use; it's not guaranteed that
216
all operations will work on old format branches.
102
218
In the test suite, creation of new trees is tested using the
103
219
`ScratchBranch` class.
127
253
__repr__ = __str__
257
if self._lock_mode or self._lock:
258
# XXX: This should show something every time, and be suitable for
259
# headless operation and embedding
260
warn("branch %r was not explicitly unlocked" % self)
263
def lock_write(self):
265
if self._lock_mode != 'w':
266
raise LockError("can't upgrade to a write lock from %r" %
268
self._lock_count += 1
270
from bzrlib.lock import WriteLock
272
self._lock = WriteLock(self.controlfilename('branch-lock'))
273
self._lock_mode = 'w'
279
assert self._lock_mode in ('r', 'w'), \
280
"invalid lock mode %r" % self._lock_mode
281
self._lock_count += 1
283
from bzrlib.lock import ReadLock
285
self._lock = ReadLock(self.controlfilename('branch-lock'))
286
self._lock_mode = 'r'
290
if not self._lock_mode:
291
raise LockError('branch %r is not locked' % (self))
293
if self._lock_count > 1:
294
self._lock_count -= 1
298
self._lock_mode = self._lock_count = None
130
300
def abspath(self, name):
131
301
"""Return absolute filename for something in the branch"""
132
302
return os.path.join(self.base, name)
135
304
def relpath(self, path):
136
305
"""Return path relative to this branch of something inside it.
138
307
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)
308
return _relpath(self.base, path)
148
310
def controlfilename(self, file_or_path):
149
311
"""Return location relative to branch."""
150
if isinstance(file_or_path, types.StringTypes):
312
if isinstance(file_or_path, basestring):
151
313
file_or_path = [file_or_path]
152
314
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
155
317
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)
318
"""Open a control file for this branch.
320
There are two classes of file in the control directory: text
321
and binary. binary files are untranslated byte streams. Text
322
control files are stored with Unix newlines and in UTF-8, even
323
if the platform or locale defaults are different.
325
Controlfiles should almost never be opened in write mode but
326
rather should be atomically copied and replaced using atomicfile.
329
fn = self.controlfilename(file_or_path)
331
if mode == 'rb' or mode == 'wb':
332
return file(fn, mode)
333
elif mode == 'r' or mode == 'w':
334
# open in binary mode anyhow so there's no newline translation;
335
# codecs uses line buffering by default; don't want that.
337
return codecs.open(fn, mode + 'b', 'utf-8',
340
raise BzrError("invalid controlfile mode %r" % mode)
160
342
def _make_control(self):
161
343
os.mkdir(self.controlfilename([]))
162
344
self.controlfile('README', 'w').write(
163
345
"This is a Bazaar-NG control directory.\n"
164
"Do not change any files in this directory.")
165
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
166
for d in ('text-store', 'inventory-store', 'revision-store'):
346
"Do not change any files in this directory.\n")
347
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT_5)
348
for d in ('text-store', 'revision-store',
167
350
os.mkdir(self.controlfilename(d))
168
for f in ('revision-history', 'merged-patches',
169
'pending-merged-patches', 'branch-name'):
351
for f in ('revision-history',
170
355
self.controlfile(f, 'w').write('')
171
356
mutter('created control directory in ' + self.base)
172
Inventory().write_xml(self.controlfile('inventory','w'))
175
def _check_format(self):
358
# if we want per-tree root ids then this is the place to set
359
# them; they're not needed for now and so ommitted for
361
f = self.controlfile('inventory','w')
362
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), f)
365
def _check_format(self, relax_version_check):
176
366
"""Check this branch format is supported.
178
The current tool only supports the current unstable format.
368
The format level is stored, as an integer, in
369
self._branch_format for code that needs to check it later.
180
371
In the future, we might need different in-memory Branch
181
372
classes to support downlevel branches. But not yet.
183
# read in binary mode to detect newline wierdness.
184
fmt = self.controlfile('branch-format', 'rb').read()
185
if fmt != BZR_BRANCH_FORMAT:
186
bailout('sorry, branch format %r not supported' % fmt,
187
['use a different bzr version',
188
'or remove the .bzr directory and "bzr init" again'])
375
fmt = self.controlfile('branch-format', 'r').read()
377
if e.errno == errno.ENOENT:
378
raise NotBranchError(self.base)
382
if fmt == BZR_BRANCH_FORMAT_5:
383
self._branch_format = 5
384
elif fmt == BZR_BRANCH_FORMAT_4:
385
self._branch_format = 4
387
if (not relax_version_check
388
and self._branch_format != 5):
389
raise BzrError('sorry, branch format %r not supported' % fmt,
390
['use a different bzr version',
391
'or remove the .bzr directory and "bzr init" again'])
393
def get_root_id(self):
394
"""Return the id of this branches root"""
395
inv = self.read_working_inventory()
396
return inv.root.file_id
398
def set_root_id(self, file_id):
399
inv = self.read_working_inventory()
400
orig_root_id = inv.root.file_id
401
del inv._byid[inv.root.file_id]
402
inv.root.file_id = file_id
403
inv._byid[inv.root.file_id] = inv.root
406
if entry.parent_id in (None, orig_root_id):
407
entry.parent_id = inv.root.file_id
408
self._write_inventory(inv)
191
410
def read_working_inventory(self):
192
411
"""Read the working inventory."""
194
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
195
mutter("loaded inventory of %d items in %f"
196
% (len(inv), time.time() - before))
414
# ElementTree does its own conversion from UTF-8, so open in
416
f = self.controlfile('inventory', 'rb')
417
return bzrlib.xml5.serializer_v5.read_inventory(f)
200
422
def _write_inventory(self, inv):
201
423
"""Update the working inventory.
203
425
That is to say, the inventory describing changes underway, that
204
426
will be committed to the next revision.
206
## TODO: factor out to atomicfile? is rename safe on windows?
207
## TODO: Maybe some kind of clean/dirty marker on inventory?
208
tmpfname = self.controlfilename('inventory.tmp')
209
tmpf = file(tmpfname, 'w')
212
os.rename(tmpfname, self.controlfilename('inventory'))
428
from bzrlib.atomicfile import AtomicFile
432
f = AtomicFile(self.controlfilename('inventory'), 'wb')
434
bzrlib.xml5.serializer_v5.write_inventory(inv, f)
213
441
mutter('wrote working inventory')
216
444
inventory = property(read_working_inventory, _write_inventory, None,
217
445
"""Inventory for the working copy.""")
220
def add(self, files, verbose=False):
448
def add(self, files, ids=None):
221
449
"""Make files versioned.
451
Note that the command line normally calls smart_add instead,
452
which can automatically recurse.
223
454
This puts the files in the Added state, so that they will be
224
455
recorded by the next commit.
226
:todo: Perhaps have an option to add the ids even if the files do
229
:todo: Perhaps return the ids of the files? But then again it
230
is easy to retrieve them if they're needed.
232
:todo: Option to specify file id.
234
:todo: Adding a directory should optionally recurse down and
235
add all non-ignored children. Perhaps do that in a
238
>>> b = ScratchBranch(files=['foo'])
239
>>> 'foo' in b.unknowns()
244
>>> 'foo' in b.unknowns()
246
>>> bool(b.inventory.path2id('foo'))
252
Traceback (most recent call last):
254
BzrError: ('foo is already versioned', [])
256
>>> b.add(['nothere'])
257
Traceback (most recent call last):
258
BzrError: ('cannot add: not a regular file or directory: nothere', [])
458
List of paths to add, relative to the base of the tree.
461
If set, use these instead of automatically generated ids.
462
Must be the same length as the list of files, but may
463
contain None for ids that are to be autogenerated.
465
TODO: Perhaps have an option to add the ids even if the files do
468
TODO: Perhaps yield the ids and paths as they're added.
261
470
# TODO: Re-adding a file that is removed in the working copy
262
471
# should probably put it back with the previous ID.
263
if isinstance(files, types.StringTypes):
472
if isinstance(files, basestring):
473
assert(ids is None or isinstance(ids, basestring))
266
inv = self.read_working_inventory()
268
if is_control_file(f):
269
bailout("cannot add control file %s" % quotefn(f))
274
bailout("cannot add top-level %r" % f)
276
fullpath = os.path.normpath(self.abspath(f))
279
kind = file_kind(fullpath)
281
# maybe something better?
282
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
284
if kind != 'file' and kind != 'directory':
285
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
287
file_id = gen_file_id(f)
288
inv.add_path(f, kind=kind, file_id=file_id)
291
show_status('A', kind, quotefn(f))
293
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
295
self._write_inventory(inv)
479
ids = [None] * len(files)
481
assert(len(ids) == len(files))
485
inv = self.read_working_inventory()
486
for f,file_id in zip(files, ids):
487
if is_control_file(f):
488
raise BzrError("cannot add control file %s" % quotefn(f))
493
raise BzrError("cannot add top-level %r" % f)
495
fullpath = os.path.normpath(self.abspath(f))
498
kind = file_kind(fullpath)
500
# maybe something better?
501
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
503
if kind != 'file' and kind != 'directory':
504
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
507
file_id = gen_file_id(f)
508
inv.add_path(f, kind=kind, file_id=file_id)
510
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
512
self._write_inventory(inv)
517
def print_file(self, file, revno):
518
"""Print `file` to stdout."""
521
tree = self.revision_tree(self.get_rev_id(revno))
522
# use inventory as it was in that revision
523
file_id = tree.inventory.path2id(file)
525
raise BzrError("%r is not present in revision %s" % (file, revno))
526
tree.print_file(file_id)
299
531
def remove(self, files, verbose=False):
302
534
This does not remove their text. This does not run on
304
:todo: Refuse to remove modified files unless --force is given?
306
>>> b = ScratchBranch(files=['foo'])
308
>>> b.inventory.has_filename('foo')
311
>>> b.working_tree().has_filename('foo')
313
>>> b.inventory.has_filename('foo')
316
>>> b = ScratchBranch(files=['foo'])
321
>>> b.inventory.has_filename('foo')
323
>>> b.basis_tree().has_filename('foo')
325
>>> b.working_tree().has_filename('foo')
328
:todo: Do something useful with directories.
330
:todo: Should this remove the text or not? Tough call; not
536
TODO: Refuse to remove modified files unless --force is given?
538
TODO: Do something useful with directories.
540
TODO: Should this remove the text or not? Tough call; not
331
541
removing may be useful and the user can just use use rm, and
332
542
is the opposite of add. Removing it is consistent with most
333
543
other tools. Maybe an option.
335
545
## TODO: Normalize names
336
546
## TODO: Remove nested loops; better scalability
338
if isinstance(files, types.StringTypes):
547
if isinstance(files, basestring):
341
tree = self.working_tree()
344
# do this before any modifications
348
bailout("cannot remove unversioned file %s" % quotefn(f))
349
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
351
# having remove it, it must be either ignored or unknown
352
if tree.is_ignored(f):
356
show_status(new_status, inv[fid].kind, quotefn(f))
553
tree = self.working_tree()
556
# do this before any modifications
560
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
561
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
563
# having remove it, it must be either ignored or unknown
564
if tree.is_ignored(f):
568
show_status(new_status, inv[fid].kind, quotefn(f))
571
self._write_inventory(inv)
576
# FIXME: this doesn't need to be a branch method
577
def set_inventory(self, new_inventory_list):
578
from bzrlib.inventory import Inventory, InventoryEntry
579
inv = Inventory(self.get_root_id())
580
for path, file_id, parent, kind in new_inventory_list:
581
name = os.path.basename(path)
584
inv.add(InventoryEntry(file_id, name, kind, parent))
359
585
self._write_inventory(inv)
378
604
return self.working_tree().unknowns()
381
def commit(self, message, timestamp=None, timezone=None,
384
"""Commit working copy as a new revision.
386
The basic approach is to add all the file texts into the
387
store, then the inventory, then make a new revision pointing
388
to that inventory and store that.
390
This is not quite safe if the working copy changes during the
391
commit; for the moment that is simply not allowed. A better
392
approach is to make a temporary copy of the files before
393
computing their hashes, and then add those hashes in turn to
394
the inventory. This should mean at least that there are no
395
broken hash pointers. There is no way we can get a snapshot
396
of the whole directory at an instant. This would also have to
397
be robust against files disappearing, moving, etc. So the
398
whole thing is a bit hard.
400
:param timestamp: if not None, seconds-since-epoch for a
401
postdated/predated commit.
404
## TODO: Show branch names
406
# TODO: Don't commit if there are no changes, unless forced?
408
# First walk over the working inventory; and both update that
409
# and also build a new revision inventory. The revision
410
# inventory needs to hold the text-id, sha1 and size of the
411
# actual file versions committed in the revision. (These are
412
# not present in the working inventory.) We also need to
413
# detect missing/deleted files, and remove them from the
416
work_inv = self.read_working_inventory()
418
basis = self.basis_tree()
419
basis_inv = basis.inventory
421
for path, entry in work_inv.iter_entries():
422
## TODO: Cope with files that have gone missing.
424
## TODO: Check that the file kind has not changed from the previous
425
## revision of this file (if any).
429
p = self.abspath(path)
430
file_id = entry.file_id
431
mutter('commit prep file %s, id %r ' % (p, file_id))
433
if not os.path.exists(p):
434
mutter(" file is missing, removing from inventory")
436
show_status('D', entry.kind, quotefn(path))
437
missing_ids.append(file_id)
440
# TODO: Handle files that have been deleted
442
# TODO: Maybe a special case for empty files? Seems a
443
# waste to store them many times.
447
if basis_inv.has_id(file_id):
448
old_kind = basis_inv[file_id].kind
449
if old_kind != entry.kind:
450
bailout("entry %r changed kind from %r to %r"
451
% (file_id, old_kind, entry.kind))
453
if entry.kind == 'directory':
455
bailout("%s is entered as directory but not a directory" % quotefn(p))
456
elif entry.kind == 'file':
458
bailout("%s is entered as file but is not a file" % quotefn(p))
460
content = file(p, 'rb').read()
462
entry.text_sha1 = sha_string(content)
463
entry.text_size = len(content)
465
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
467
and (old_ie.text_size == entry.text_size)
468
and (old_ie.text_sha1 == entry.text_sha1)):
469
## assert content == basis.get_file(file_id).read()
470
entry.text_id = basis_inv[file_id].text_id
471
mutter(' unchanged from previous text_id {%s}' %
475
entry.text_id = gen_file_id(entry.name)
476
self.text_store.add(content, entry.text_id)
477
mutter(' stored with text_id {%s}' % entry.text_id)
481
elif (old_ie.name == entry.name
482
and old_ie.parent_id == entry.parent_id):
487
show_status(state, entry.kind, quotefn(path))
489
for file_id in missing_ids:
490
# have to do this later so we don't mess up the iterator.
491
# since parents may be removed before their children we
494
# FIXME: There's probably a better way to do this; perhaps
495
# the workingtree should know how to filter itself.
496
if work_inv.has_id(file_id):
497
del work_inv[file_id]
500
inv_id = rev_id = _gen_revision_id(time.time())
502
inv_tmp = tempfile.TemporaryFile()
503
inv.write_xml(inv_tmp)
505
self.inventory_store.add(inv_tmp, inv_id)
506
mutter('new inventory_id is {%s}' % inv_id)
508
self._write_inventory(work_inv)
510
if timestamp == None:
511
timestamp = time.time()
513
if committer == None:
514
committer = username()
517
timezone = local_time_offset()
519
mutter("building commit log message")
520
rev = Revision(timestamp=timestamp,
523
precursor = self.last_patch(),
528
rev_tmp = tempfile.TemporaryFile()
529
rev.write_xml(rev_tmp)
531
self.revision_store.add(rev_tmp, rev_id)
532
mutter("new revision_id is {%s}" % rev_id)
534
## XXX: Everything up to here can simply be orphaned if we abort
535
## the commit; it will leave junk files behind but that doesn't
538
## TODO: Read back the just-generated changeset, and make sure it
539
## applies and recreates the right state.
541
## TODO: Also calculate and store the inventory SHA1
542
mutter("committing patch r%d" % (self.revno() + 1))
544
mutter("append to revision-history")
545
self.controlfile('revision-history', 'at').write(rev_id + '\n')
607
def append_revision(self, *revision_ids):
608
from bzrlib.atomicfile import AtomicFile
610
for revision_id in revision_ids:
611
mutter("add {%s} to revision-history" % revision_id)
613
rev_history = self.revision_history()
614
rev_history.extend(revision_ids)
616
f = AtomicFile(self.controlfilename('revision-history'))
618
for rev_id in rev_history:
625
def has_revision(self, revision_id):
626
"""True if this branch has a copy of the revision.
628
This does not necessarily imply the revision is merge
629
or on the mainline."""
630
return (revision_id is None
631
or revision_id in self.revision_store)
634
def get_revision_xml_file(self, revision_id):
635
"""Return XML file object for revision object."""
636
if not revision_id or not isinstance(revision_id, basestring):
637
raise InvalidRevisionId(revision_id)
642
return self.revision_store[revision_id]
643
except (IndexError, KeyError):
644
raise bzrlib.errors.NoSuchRevision(self, revision_id)
649
def get_revision_xml(self, revision_id):
650
return self.get_revision_xml_file(revision_id).read()
550
653
def get_revision(self, revision_id):
551
654
"""Return the Revision object for a named revision"""
552
r = Revision.read_xml(self.revision_store[revision_id])
655
xml_file = self.get_revision_xml_file(revision_id)
658
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
659
except SyntaxError, e:
660
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
553
664
assert r.revision_id == revision_id
557
def get_inventory(self, inventory_id):
558
"""Get Inventory object by hash.
560
:todo: Perhaps for this and similar methods, take a revision
561
parameter which can be either an integer revno or a
563
i = Inventory.read_xml(self.inventory_store[inventory_id])
668
def get_revision_delta(self, revno):
669
"""Return the delta for one revision.
671
The delta is relative to its mainline predecessor, or the
672
empty tree for revision 1.
674
assert isinstance(revno, int)
675
rh = self.revision_history()
676
if not (1 <= revno <= len(rh)):
677
raise InvalidRevisionNumber(revno)
679
# revno is 1-based; list is 0-based
681
new_tree = self.revision_tree(rh[revno-1])
683
old_tree = EmptyTree()
685
old_tree = self.revision_tree(rh[revno-2])
687
return compare_trees(old_tree, new_tree)
690
def get_revision_sha1(self, revision_id):
691
"""Hash the stored value of a revision, and return it."""
692
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
695
def _get_ancestry_weave(self):
696
return self.control_weaves.get_weave('ancestry')
699
def get_ancestry(self, revision_id):
700
"""Return a list of revision-ids integrated by a revision.
703
if revision_id is None:
705
w = self._get_ancestry_weave()
706
return [None] + [l[:-1] for l in w.get_iter(w.lookup(revision_id))]
709
def get_inventory_weave(self):
710
return self.control_weaves.get_weave('inventory')
713
def get_inventory(self, revision_id):
714
"""Get Inventory object by hash."""
715
xml = self.get_inventory_xml(revision_id)
716
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
719
def get_inventory_xml(self, revision_id):
720
"""Get inventory XML as a file object."""
722
assert isinstance(revision_id, basestring), type(revision_id)
723
iw = self.get_inventory_weave()
724
return iw.get_text(iw.lookup(revision_id))
726
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
729
def get_inventory_sha1(self, revision_id):
730
"""Return the sha1 hash of the inventory entry
732
return self.get_revision(revision_id).inventory_sha1
567
735
def get_revision_inventory(self, revision_id):
568
736
"""Return inventory of a past revision."""
737
# TODO: Unify this with get_inventory()
738
# bzr 0.0.6 and later imposes the constraint that the inventory_id
739
# must be the same as its revision, so this is trivial.
569
740
if revision_id == None:
741
return Inventory(self.get_root_id())
572
return self.get_inventory(self.get_revision(revision_id).inventory_id)
743
return self.get_inventory(revision_id)
575
746
def revision_history(self):
576
"""Return sequence of revision hashes on to this branch.
578
>>> ScratchBranch().revision_history()
581
return [chomp(l) for l in self.controlfile('revision-history').readlines()]
747
"""Return sequence of revision hashes on to this branch."""
750
return [l.rstrip('\r\n') for l in
751
self.controlfile('revision-history', 'r').readlines()]
756
def common_ancestor(self, other, self_revno=None, other_revno=None):
758
>>> from bzrlib.commit import commit
759
>>> sb = ScratchBranch(files=['foo', 'foo~'])
760
>>> sb.common_ancestor(sb) == (None, None)
762
>>> commit(sb, "Committing first revision", verbose=False)
763
>>> sb.common_ancestor(sb)[0]
765
>>> clone = sb.clone()
766
>>> commit(sb, "Committing second revision", verbose=False)
767
>>> sb.common_ancestor(sb)[0]
769
>>> sb.common_ancestor(clone)[0]
771
>>> commit(clone, "Committing divergent second revision",
773
>>> sb.common_ancestor(clone)[0]
775
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
777
>>> sb.common_ancestor(sb) != clone.common_ancestor(clone)
779
>>> clone2 = sb.clone()
780
>>> sb.common_ancestor(clone2)[0]
782
>>> sb.common_ancestor(clone2, self_revno=1)[0]
784
>>> sb.common_ancestor(clone2, other_revno=1)[0]
787
my_history = self.revision_history()
788
other_history = other.revision_history()
789
if self_revno is None:
790
self_revno = len(my_history)
791
if other_revno is None:
792
other_revno = len(other_history)
793
indices = range(min((self_revno, other_revno)))
796
if my_history[r] == other_history[r]:
797
return r+1, my_history[r]
587
804
That is equivalent to the number of revisions committed to
590
>>> b = ScratchBranch()
593
>>> b.commit('no foo')
597
807
return len(self.revision_history())
600
def last_patch(self):
810
def last_revision(self):
601
811
"""Return last patch hash, or None if no history.
603
>>> ScratchBranch().last_patch() == None
606
813
ph = self.revision_history()
611
def lookup_revision(self, revno):
612
"""Return revision hash for revision number."""
820
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
821
"""Return a list of new revisions that would perfectly fit.
823
If self and other have not diverged, return a list of the revisions
824
present in other, but missing from self.
826
>>> from bzrlib.commit import commit
827
>>> bzrlib.trace.silent = True
828
>>> br1 = ScratchBranch()
829
>>> br2 = ScratchBranch()
830
>>> br1.missing_revisions(br2)
832
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
833
>>> br1.missing_revisions(br2)
835
>>> br2.missing_revisions(br1)
837
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
838
>>> br1.missing_revisions(br2)
840
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
841
>>> br1.missing_revisions(br2)
843
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
844
>>> br1.missing_revisions(br2)
845
Traceback (most recent call last):
846
DivergedBranches: These branches have diverged.
848
# FIXME: If the branches have diverged, but the latest
849
# revision in this branch is completely merged into the other,
850
# then we should still be able to pull.
851
self_history = self.revision_history()
852
self_len = len(self_history)
853
other_history = other.revision_history()
854
other_len = len(other_history)
855
common_index = min(self_len, other_len) -1
856
if common_index >= 0 and \
857
self_history[common_index] != other_history[common_index]:
858
raise DivergedBranches(self, other)
860
if stop_revision is None:
861
stop_revision = other_len
863
assert isinstance(stop_revision, int)
864
if stop_revision > other_len:
865
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
866
return other_history[self_len:stop_revision]
868
def update_revisions(self, other, stop_revision=None):
869
"""Pull in new perfect-fit revisions."""
870
from bzrlib.fetch import greedy_fetch
871
from bzrlib.revision import get_intervening_revisions
872
if stop_revision is None:
873
stop_revision = other.last_revision()
874
greedy_fetch(to_branch=self, from_branch=other,
875
revision=stop_revision)
876
pullable_revs = self.missing_revisions(
877
other, other.revision_id_to_revno(stop_revision))
879
greedy_fetch(to_branch=self,
881
revision=pullable_revs[-1])
882
self.append_revision(*pullable_revs)
884
def commit(self, *args, **kw):
885
from bzrlib.commit import Commit
886
Commit().commit(self, *args, **kw)
888
def revision_id_to_revno(self, revision_id):
889
"""Given a revision id, return its revno"""
890
if revision_id is None:
892
history = self.revision_history()
894
return history.index(revision_id) + 1
896
raise bzrlib.errors.NoSuchRevision(self, revision_id)
898
def get_rev_id(self, revno, history=None):
899
"""Find the revision id of the specified revno."""
617
# list is 0-based; revisions are 1-based
618
return self.revision_history()[revno-1]
620
bailout("no such revision %s" % revno)
903
history = self.revision_history()
904
elif revno <= 0 or revno > len(history):
905
raise bzrlib.errors.NoSuchRevision(self, revno)
906
return history[revno - 1]
623
908
def revision_tree(self, revision_id):
624
909
"""Return Tree for a revision on this branch.
626
911
`revision_id` may be None for the null revision, in which case
627
912
an `EmptyTree` is returned."""
913
# TODO: refactor this to use an existing revision object
914
# so we don't need to read it in twice.
629
915
if revision_id == None:
630
916
return EmptyTree()
632
918
inv = self.get_revision_inventory(revision_id)
633
return RevisionTree(self.text_store, inv)
919
return RevisionTree(self.weave_store, inv, revision_id)
636
922
def working_tree(self):
637
923
"""Return a `Tree` for the working copy."""
924
from bzrlib.workingtree import WorkingTree
638
925
return WorkingTree(self.base, self.read_working_inventory())
642
929
"""Return `Tree` object for last revision.
644
931
If there are no revisions yet, return an `EmptyTree`.
646
>>> b = ScratchBranch(files=['foo'])
647
>>> b.basis_tree().has_filename('foo')
649
>>> b.working_tree().has_filename('foo')
652
>>> b.commit('add foo')
653
>>> b.basis_tree().has_filename('foo')
656
r = self.last_patch()
660
return RevisionTree(self.text_store, self.get_revision_inventory(r))
664
def write_log(self, show_timezone='original'):
665
"""Write out human-readable log of commits to this branch
667
:param utc: If true, show dates in universal time, not local time."""
668
## TODO: Option to choose either original, utc or local timezone
671
for p in self.revision_history():
673
print 'revno:', revno
674
## TODO: Show hash if --id is given.
675
##print 'revision-hash:', p
676
rev = self.get_revision(p)
677
print 'committer:', rev.committer
678
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
681
## opportunistic consistency check, same as check_patch_chaining
682
if rev.precursor != precursor:
683
bailout("mismatched precursor!")
687
print ' (no message)'
689
for l in rev.message.split('\n'):
697
def show_status(branch, show_all=False):
698
"""Display single-line status for non-ignored working files.
700
The list is show sorted in order by file name.
702
>>> b = ScratchBranch(files=['foo', 'foo~'])
708
>>> b.commit("add foo")
710
>>> os.unlink(b.abspath('foo'))
715
:todo: Get state for single files.
717
:todo: Perhaps show a slash at the end of directory names.
721
# We have to build everything into a list first so that it can
722
# sorted by name, incorporating all the different sources.
724
# FIXME: Rather than getting things in random order and then sorting,
725
# just step through in order.
727
# Interesting case: the old ID for a file has been removed,
728
# but a new file has been created under that name.
730
old = branch.basis_tree()
731
old_inv = old.inventory
732
new = branch.working_tree()
733
new_inv = new.inventory
735
for fs, fid, oldname, newname, kind in diff_trees(old, new):
737
show_status(fs, kind,
738
oldname + ' => ' + newname)
739
elif fs == 'A' or fs == 'M':
740
show_status(fs, kind, newname)
742
show_status(fs, kind, oldname)
745
show_status(fs, kind, newname)
748
show_status(fs, kind, newname)
750
show_status(fs, kind, newname)
752
bailout("wierd file state %r" % ((fs, fid),))
756
class ScratchBranch(Branch):
933
return self.revision_tree(self.last_revision())
936
def rename_one(self, from_rel, to_rel):
939
This can change the directory or the filename or both.
943
tree = self.working_tree()
945
if not tree.has_filename(from_rel):
946
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
947
if tree.has_filename(to_rel):
948
raise BzrError("can't rename: new working file %r already exists" % to_rel)
950
file_id = inv.path2id(from_rel)
952
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
954
if inv.path2id(to_rel):
955
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
957
to_dir, to_tail = os.path.split(to_rel)
958
to_dir_id = inv.path2id(to_dir)
959
if to_dir_id == None and to_dir != '':
960
raise BzrError("can't determine destination directory id for %r" % to_dir)
962
mutter("rename_one:")
963
mutter(" file_id {%s}" % file_id)
964
mutter(" from_rel %r" % from_rel)
965
mutter(" to_rel %r" % to_rel)
966
mutter(" to_dir %r" % to_dir)
967
mutter(" to_dir_id {%s}" % to_dir_id)
969
inv.rename(file_id, to_dir_id, to_tail)
971
from_abs = self.abspath(from_rel)
972
to_abs = self.abspath(to_rel)
974
rename(from_abs, to_abs)
976
raise BzrError("failed to rename %r to %r: %s"
977
% (from_abs, to_abs, e[1]),
978
["rename rolled back"])
980
self._write_inventory(inv)
985
def move(self, from_paths, to_name):
988
to_name must exist as a versioned directory.
990
If to_name exists and is a directory, the files are moved into
991
it, keeping their old names. If it is a directory,
993
Note that to_name is only the last component of the new name;
994
this doesn't change the directory.
996
This returns a list of (from_path, to_path) pairs for each
1002
## TODO: Option to move IDs only
1003
assert not isinstance(from_paths, basestring)
1004
tree = self.working_tree()
1005
inv = tree.inventory
1006
to_abs = self.abspath(to_name)
1007
if not isdir(to_abs):
1008
raise BzrError("destination %r is not a directory" % to_abs)
1009
if not tree.has_filename(to_name):
1010
raise BzrError("destination %r not in working directory" % to_abs)
1011
to_dir_id = inv.path2id(to_name)
1012
if to_dir_id == None and to_name != '':
1013
raise BzrError("destination %r is not a versioned directory" % to_name)
1014
to_dir_ie = inv[to_dir_id]
1015
if to_dir_ie.kind not in ('directory', 'root_directory'):
1016
raise BzrError("destination %r is not a directory" % to_abs)
1018
to_idpath = inv.get_idpath(to_dir_id)
1020
for f in from_paths:
1021
if not tree.has_filename(f):
1022
raise BzrError("%r does not exist in working tree" % f)
1023
f_id = inv.path2id(f)
1025
raise BzrError("%r is not versioned" % f)
1026
name_tail = splitpath(f)[-1]
1027
dest_path = appendpath(to_name, name_tail)
1028
if tree.has_filename(dest_path):
1029
raise BzrError("destination %r already exists" % dest_path)
1030
if f_id in to_idpath:
1031
raise BzrError("can't move %r to a subdirectory of itself" % f)
1033
# OK, so there's a race here, it's possible that someone will
1034
# create a file in this interval and then the rename might be
1035
# left half-done. But we should have caught most problems.
1037
for f in from_paths:
1038
name_tail = splitpath(f)[-1]
1039
dest_path = appendpath(to_name, name_tail)
1040
result.append((f, dest_path))
1041
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1043
rename(self.abspath(f), self.abspath(dest_path))
1045
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1046
["rename rolled back"])
1048
self._write_inventory(inv)
1055
def revert(self, filenames, old_tree=None, backups=True):
1056
"""Restore selected files to the versions from a previous tree.
1059
If true (default) backups are made of files before
1062
from bzrlib.errors import NotVersionedError, BzrError
1063
from bzrlib.atomicfile import AtomicFile
1064
from bzrlib.osutils import backup_file
1066
inv = self.read_working_inventory()
1067
if old_tree is None:
1068
old_tree = self.basis_tree()
1069
old_inv = old_tree.inventory
1072
for fn in filenames:
1073
file_id = inv.path2id(fn)
1075
raise NotVersionedError("not a versioned file", fn)
1076
if not old_inv.has_id(file_id):
1077
raise BzrError("file not present in old tree", fn, file_id)
1078
nids.append((fn, file_id))
1080
# TODO: Rename back if it was previously at a different location
1082
# TODO: If given a directory, restore the entire contents from
1083
# the previous version.
1085
# TODO: Make a backup to a temporary file.
1087
# TODO: If the file previously didn't exist, delete it?
1088
for fn, file_id in nids:
1091
f = AtomicFile(fn, 'wb')
1093
f.write(old_tree.get_file(file_id).read())
1099
def pending_merges(self):
1100
"""Return a list of pending merges.
1102
These are revisions that have been merged into the working
1103
directory but not yet committed.
1105
cfn = self.controlfilename('pending-merges')
1106
if not os.path.exists(cfn):
1109
for l in self.controlfile('pending-merges', 'r').readlines():
1110
p.append(l.rstrip('\n'))
1114
def add_pending_merge(self, revision_id):
1115
validate_revision_id(revision_id)
1116
# TODO: Perhaps should check at this point that the
1117
# history of the revision is actually present?
1118
p = self.pending_merges()
1119
if revision_id in p:
1121
p.append(revision_id)
1122
self.set_pending_merges(p)
1125
def set_pending_merges(self, rev_list):
1126
from bzrlib.atomicfile import AtomicFile
1129
f = AtomicFile(self.controlfilename('pending-merges'))
1140
def get_parent(self):
1141
"""Return the parent location of the branch.
1143
This is the default location for push/pull/missing. The usual
1144
pattern is that the user can override it by specifying a
1148
_locs = ['parent', 'pull', 'x-pull']
1151
return self.controlfile(l, 'r').read().strip('\n')
1153
if e.errno != errno.ENOENT:
1158
def set_parent(self, url):
1159
# TODO: Maybe delete old location files?
1160
from bzrlib.atomicfile import AtomicFile
1163
f = AtomicFile(self.controlfilename('parent'))
1172
def check_revno(self, revno):
1174
Check whether a revno corresponds to any revision.
1175
Zero (the NULL revision) is considered valid.
1178
self.check_real_revno(revno)
1180
def check_real_revno(self, revno):
1182
Check whether a revno corresponds to a real revision.
1183
Zero (the NULL revision) is considered invalid
1185
if revno < 1 or revno > self.revno():
1186
raise InvalidRevisionNumber(revno)
1192
class ScratchBranch(LocalBranch):
757
1193
"""Special test class: a branch that cleans up after itself.
759
1195
>>> b = ScratchBranch()
760
1196
>>> isdir(b.base)
767
def __init__(self, files = []):
1203
def __init__(self, files=[], dirs=[], base=None):
768
1204
"""Make a test branch.
770
1206
This creates a temporary directory and runs init-tree in it.
772
1208
If any files are listed, they are created in the working copy.
774
Branch.__init__(self, tempfile.mkdtemp(), init=True)
1210
from tempfile import mkdtemp
1215
LocalBranch.__init__(self, base, init=init)
1217
os.mkdir(self.abspath(d))
776
1220
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
1225
>>> orig = ScratchBranch(files=["file1", "file2"])
1226
>>> clone = orig.clone()
1227
>>> if os.name != 'nt':
1228
... os.path.samefile(orig.base, clone.base)
1230
... orig.base == clone.base
1233
>>> os.path.isfile(os.path.join(clone.base, "file1"))
1236
from shutil import copytree
1237
from tempfile import mkdtemp
1240
copytree(self.base, base, symlinks=True)
1241
return ScratchBranch(base=base)
779
1245
def __del__(self):
780
1249
"""Destroy the test branch, removing the scratch directory."""
781
shutil.rmtree(self.base)
1250
from shutil import rmtree
1253
mutter("delete ScratchBranch %s" % self.base)
1256
# Work around for shutil.rmtree failing on Windows when
1257
# readonly files are encountered
1258
mutter("hit exception in destroying ScratchBranch: %s" % e)
1259
for root, dirs, files in os.walk(self.base, topdown=False):
1261
os.chmod(os.path.join(root, name), 0700)