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
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, \
28
DivergedBranches, NotBranchError
29
from bzrlib.textui import show_status
30
from bzrlib.revision import Revision
31
from bzrlib.delta import compare_trees
32
from bzrlib.tree import EmptyTree, RevisionTree
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
def find_branch(*ignored, **ignored_too):
47
# XXX: leave this here for about one release, then remove it
48
raise NotImplementedError('find_branch() is not supported anymore, '
49
'please use one of the new branch constructors')
51
def _relpath(base, path):
52
"""Return path relative to base, or raise exception.
54
The path may be either an absolute path or a path relative to the
55
current working directory.
57
Lifted out of Branch.relpath for ease of testing.
59
os.path.commonprefix (python2.4) has a bad bug that it works just
60
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
61
avoids that problem."""
62
rp = os.path.abspath(path)
66
while len(head) >= len(base):
69
head, tail = os.path.split(head)
73
raise NotBranchError("path %r is not within branch %r" % (rp, base))
43
78
def find_branch_root(f=None):
44
79
"""Find the branch root enclosing f, or pwd.
81
f may be a filename or a URL.
46
83
It is not necessary that f exists.
48
85
Basically we keep looking up until we find the control directory or
86
run into the root. If there isn't one, raises NotBranchError.
52
90
elif hasattr(os.path, 'realpath'):
53
91
f = os.path.realpath(f)
55
93
f = os.path.abspath(f)
94
if not os.path.exists(f):
95
raise BzrError('%r does not exist' % f)
61
101
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
63
103
head, tail = os.path.split(f)
65
105
# reached the root, whatever that may be
66
bailout('%r is not in a branch' % orig_f)
106
raise NotBranchError('%s is not in a branch' % orig_f)
71
112
######################################################################
115
class Branch(object):
75
116
"""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.
119
Base directory/url of the branch.
123
def __init__(self, *ignored, **ignored_too):
124
raise NotImplementedError('The Branch class is abstract')
128
"""Open an existing branch, rooted at 'base' (url)"""
129
if base and (base.startswith('http://') or base.startswith('https://')):
130
from bzrlib.remotebranch import RemoteBranch
131
return RemoteBranch(base, find_root=False)
133
return LocalBranch(base, find_root=False)
136
def open_containing(url):
137
"""Open an existing branch, containing url (search upwards for the root)
139
if url and (url.startswith('http://') or url.startswith('https://')):
140
from bzrlib.remotebranch import RemoteBranch
141
return RemoteBranch(url)
143
return LocalBranch(url)
146
def initialize(base):
147
"""Create a new branch, rooted at 'base' (url)"""
148
if base and (base.startswith('http://') or base.startswith('https://')):
149
from bzrlib.remotebranch import RemoteBranch
150
return RemoteBranch(base, init=True)
152
return LocalBranch(base, init=True)
154
def setup_caching(self, cache_root):
155
"""Subclasses that care about caching should override this, and set
156
up cached stores located under cache_root.
160
class LocalBranch(Branch):
161
"""A branch stored in the actual filesystem.
163
Note that it's "local" in the context of the filesystem; it doesn't
164
really matter if it's on an nfs/smb/afs/coda/... share, as long as
165
it's writable, and can be accessed via the normal filesystem API.
171
If _lock_mode is true, a positive count of the number of times the
175
Lock object from bzrlib.lock.
177
# We actually expect this class to be somewhat short-lived; part of its
178
# purpose is to try to isolate what bits of the branch logic are tied to
179
# filesystem access, so that in a later step, we can extricate them to
180
# a separarte ("storage") class.
90
185
def __init__(self, base, init=False, find_root=True):
91
186
"""Create new branch object at a particular location.
93
:param base: Base directory for the branch.
188
base -- Base directory for the branch. May be a file:// url.
95
:param init: If True, create new control files in a previously
190
init -- If True, create new control files in a previously
96
191
unversioned directory. If False, the branch must already
99
:param find_root: If true and init is false, find the root of the
194
find_root -- If true and init is false, find the root of the
100
195
existing branch containing base.
102
197
In the test suite, creation of new trees is tested using the
103
198
`ScratchBranch` class.
200
from bzrlib.store import ImmutableStore
106
202
self.base = os.path.realpath(base)
107
203
self._make_control()
109
205
self.base = find_branch_root(base)
207
if base.startswith("file://"):
111
209
self.base = os.path.realpath(base)
112
210
if not isdir(self.controlfilename('.')):
113
bailout("not a bzr branch: %s" % quotefn(base),
114
['use "bzr init" to initialize a new working tree',
115
'current bzr can only operate from top-of-tree'])
211
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
212
['use "bzr init" to initialize a new working tree',
213
'current bzr can only operate from top-of-tree'])
116
214
self._check_format()
118
216
self.text_store = ImmutableStore(self.controlfilename('text-store'))
127
225
__repr__ = __str__
229
if self._lock_mode or self._lock:
230
from bzrlib.warnings import warn
231
warn("branch %r was not explicitly unlocked" % self)
234
def lock_write(self):
236
if self._lock_mode != 'w':
237
from bzrlib.errors import LockError
238
raise LockError("can't upgrade to a write lock from %r" %
240
self._lock_count += 1
242
from bzrlib.lock import WriteLock
244
self._lock = WriteLock(self.controlfilename('branch-lock'))
245
self._lock_mode = 'w'
251
assert self._lock_mode in ('r', 'w'), \
252
"invalid lock mode %r" % self._lock_mode
253
self._lock_count += 1
255
from bzrlib.lock import ReadLock
257
self._lock = ReadLock(self.controlfilename('branch-lock'))
258
self._lock_mode = 'r'
262
if not self._lock_mode:
263
from bzrlib.errors import LockError
264
raise LockError('branch %r is not locked' % (self))
266
if self._lock_count > 1:
267
self._lock_count -= 1
271
self._lock_mode = self._lock_count = None
130
273
def abspath(self, name):
131
274
"""Return absolute filename for something in the branch"""
132
275
return os.path.join(self.base, name)
135
277
def relpath(self, path):
136
278
"""Return path relative to this branch of something inside it.
138
280
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)
281
return _relpath(self.base, path)
148
283
def controlfilename(self, file_or_path):
149
284
"""Return location relative to branch."""
150
if isinstance(file_or_path, types.StringTypes):
285
if isinstance(file_or_path, basestring):
151
286
file_or_path = [file_or_path]
152
287
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
155
290
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)
291
"""Open a control file for this branch.
293
There are two classes of file in the control directory: text
294
and binary. binary files are untranslated byte streams. Text
295
control files are stored with Unix newlines and in UTF-8, even
296
if the platform or locale defaults are different.
298
Controlfiles should almost never be opened in write mode but
299
rather should be atomically copied and replaced using atomicfile.
302
fn = self.controlfilename(file_or_path)
304
if mode == 'rb' or mode == 'wb':
305
return file(fn, mode)
306
elif mode == 'r' or mode == 'w':
307
# open in binary mode anyhow so there's no newline translation;
308
# codecs uses line buffering by default; don't want that.
310
return codecs.open(fn, mode + 'b', 'utf-8',
313
raise BzrError("invalid controlfile mode %r" % mode)
160
315
def _make_control(self):
316
from bzrlib.inventory import Inventory
161
318
os.mkdir(self.controlfilename([]))
162
319
self.controlfile('README', 'w').write(
163
320
"This is a Bazaar-NG control directory.\n"
164
"Do not change any files in this directory.")
321
"Do not change any files in this directory.\n")
165
322
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
166
323
for d in ('text-store', 'inventory-store', 'revision-store'):
167
324
os.mkdir(self.controlfilename(d))
168
325
for f in ('revision-history', 'merged-patches',
169
'pending-merged-patches', 'branch-name'):
326
'pending-merged-patches', 'branch-name',
170
329
self.controlfile(f, 'w').write('')
171
330
mutter('created control directory in ' + self.base)
172
Inventory().write_xml(self.controlfile('inventory','w'))
332
# if we want per-tree root ids then this is the place to set
333
# them; they're not needed for now and so ommitted for
335
f = self.controlfile('inventory','w')
336
bzrlib.xml.serializer_v4.write_inventory(Inventory(), f)
175
339
def _check_format(self):
180
344
In the future, we might need different in-memory Branch
181
345
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()
347
# This ignores newlines so that we can open branches created
348
# on Windows from Linux and so on. I think it might be better
349
# to always make all internal files in unix format.
350
fmt = self.controlfile('branch-format', 'r').read()
351
fmt = fmt.replace('\r\n', '\n')
185
352
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'])
353
raise BzrError('sorry, branch format %r not supported' % fmt,
354
['use a different bzr version',
355
'or remove the .bzr directory and "bzr init" again'])
357
def get_root_id(self):
358
"""Return the id of this branches root"""
359
inv = self.read_working_inventory()
360
return inv.root.file_id
362
def set_root_id(self, file_id):
363
inv = self.read_working_inventory()
364
orig_root_id = inv.root.file_id
365
del inv._byid[inv.root.file_id]
366
inv.root.file_id = file_id
367
inv._byid[inv.root.file_id] = inv.root
370
if entry.parent_id in (None, orig_root_id):
371
entry.parent_id = inv.root.file_id
372
self._write_inventory(inv)
191
374
def read_working_inventory(self):
192
375
"""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))
376
from bzrlib.inventory import Inventory
379
# ElementTree does its own conversion from UTF-8, so open in
381
f = self.controlfile('inventory', 'rb')
382
return bzrlib.xml.serializer_v4.read_inventory(f)
200
387
def _write_inventory(self, inv):
201
388
"""Update the working inventory.
203
390
That is to say, the inventory describing changes underway, that
204
391
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'))
393
from bzrlib.atomicfile import AtomicFile
397
f = AtomicFile(self.controlfilename('inventory'), 'wb')
399
bzrlib.xml.serializer_v4.write_inventory(inv, f)
213
406
mutter('wrote working inventory')
216
409
inventory = property(read_working_inventory, _write_inventory, None,
217
410
"""Inventory for the working copy.""")
220
def add(self, files, verbose=False):
413
def add(self, files, ids=None):
221
414
"""Make files versioned.
416
Note that the command line normally calls smart_add instead,
417
which can automatically recurse.
223
419
This puts the files in the Added state, so that they will be
224
420
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', [])
423
List of paths to add, relative to the base of the tree.
426
If set, use these instead of automatically generated ids.
427
Must be the same length as the list of files, but may
428
contain None for ids that are to be autogenerated.
430
TODO: Perhaps have an option to add the ids even if the files do
433
TODO: Perhaps yield the ids and paths as they're added.
261
435
# TODO: Re-adding a file that is removed in the working copy
262
436
# should probably put it back with the previous ID.
263
if isinstance(files, types.StringTypes):
437
if isinstance(files, basestring):
438
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)
444
ids = [None] * len(files)
446
assert(len(ids) == len(files))
450
inv = self.read_working_inventory()
451
for f,file_id in zip(files, ids):
452
if is_control_file(f):
453
raise BzrError("cannot add control file %s" % quotefn(f))
458
raise BzrError("cannot add top-level %r" % f)
460
fullpath = os.path.normpath(self.abspath(f))
463
kind = file_kind(fullpath)
465
# maybe something better?
466
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
468
if kind != 'file' and kind != 'directory':
469
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
472
file_id = gen_file_id(f)
473
inv.add_path(f, kind=kind, file_id=file_id)
475
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
477
self._write_inventory(inv)
482
def print_file(self, file, revno):
483
"""Print `file` to stdout."""
486
tree = self.revision_tree(self.get_rev_id(revno))
487
# use inventory as it was in that revision
488
file_id = tree.inventory.path2id(file)
490
raise BzrError("%r is not present in revision %s" % (file, revno))
491
tree.print_file(file_id)
299
496
def remove(self, files, verbose=False):
302
499
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
501
TODO: Refuse to remove modified files unless --force is given?
503
TODO: Do something useful with directories.
505
TODO: Should this remove the text or not? Tough call; not
331
506
removing may be useful and the user can just use use rm, and
332
507
is the opposite of add. Removing it is consistent with most
333
508
other tools. Maybe an option.
335
510
## TODO: Normalize names
336
511
## TODO: Remove nested loops; better scalability
338
if isinstance(files, types.StringTypes):
512
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))
518
tree = self.working_tree()
521
# do this before any modifications
525
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
526
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
528
# having remove it, it must be either ignored or unknown
529
if tree.is_ignored(f):
533
show_status(new_status, inv[fid].kind, quotefn(f))
536
self._write_inventory(inv)
541
# FIXME: this doesn't need to be a branch method
542
def set_inventory(self, new_inventory_list):
543
from bzrlib.inventory import Inventory, InventoryEntry
544
inv = Inventory(self.get_root_id())
545
for path, file_id, parent, kind in new_inventory_list:
546
name = os.path.basename(path)
549
inv.add(InventoryEntry(file_id, name, kind, parent))
359
550
self._write_inventory(inv)
378
569
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
f = self.controlfile('revision-history', 'at')
546
f.write(rev_id + '\n')
550
note("commited r%d" % self.revno())
572
def append_revision(self, *revision_ids):
573
from bzrlib.atomicfile import AtomicFile
575
for revision_id in revision_ids:
576
mutter("add {%s} to revision-history" % revision_id)
578
rev_history = self.revision_history()
579
rev_history.extend(revision_ids)
581
f = AtomicFile(self.controlfilename('revision-history'))
583
for rev_id in rev_history:
590
def get_revision_xml_file(self, revision_id):
591
"""Return XML file object for revision object."""
592
if not revision_id or not isinstance(revision_id, basestring):
593
raise InvalidRevisionId(revision_id)
598
return self.revision_store[revision_id]
599
except (IndexError, KeyError):
600
raise bzrlib.errors.NoSuchRevision(self, revision_id)
606
get_revision_xml = get_revision_xml_file
553
609
def get_revision(self, revision_id):
554
610
"""Return the Revision object for a named revision"""
555
r = Revision.read_xml(self.revision_store[revision_id])
611
xml_file = self.get_revision_xml_file(revision_id)
614
r = bzrlib.xml.serializer_v4.read_revision(xml_file)
615
except SyntaxError, e:
616
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
556
620
assert r.revision_id == revision_id
624
def get_revision_delta(self, revno):
625
"""Return the delta for one revision.
627
The delta is relative to its mainline predecessor, or the
628
empty tree for revision 1.
630
assert isinstance(revno, int)
631
rh = self.revision_history()
632
if not (1 <= revno <= len(rh)):
633
raise InvalidRevisionNumber(revno)
635
# revno is 1-based; list is 0-based
637
new_tree = self.revision_tree(rh[revno-1])
639
old_tree = EmptyTree()
641
old_tree = self.revision_tree(rh[revno-2])
643
return compare_trees(old_tree, new_tree)
647
def get_revision_sha1(self, revision_id):
648
"""Hash the stored value of a revision, and return it."""
649
# In the future, revision entries will be signed. At that
650
# point, it is probably best *not* to include the signature
651
# in the revision hash. Because that lets you re-sign
652
# the revision, (add signatures/remove signatures) and still
653
# have all hash pointers stay consistent.
654
# But for now, just hash the contents.
655
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
560
658
def get_inventory(self, inventory_id):
561
659
"""Get Inventory object by hash.
563
:todo: Perhaps for this and similar methods, take a revision
661
TODO: Perhaps for this and similar methods, take a revision
564
662
parameter which can be either an integer revno or a
566
i = Inventory.read_xml(self.inventory_store[inventory_id])
664
from bzrlib.inventory import Inventory
666
f = self.get_inventory_xml_file(inventory_id)
667
return bzrlib.xml.serializer_v4.read_inventory(f)
670
def get_inventory_xml(self, inventory_id):
671
"""Get inventory XML as a file object."""
672
return self.inventory_store[inventory_id]
674
get_inventory_xml_file = get_inventory_xml
677
def get_inventory_sha1(self, inventory_id):
678
"""Return the sha1 hash of the inventory entry
680
return sha_file(self.get_inventory_xml(inventory_id))
570
683
def get_revision_inventory(self, revision_id):
571
684
"""Return inventory of a past revision."""
685
# bzr 0.0.6 imposes the constraint that the inventory_id
686
# must be the same as its revision, so this is trivial.
572
687
if revision_id == None:
688
from bzrlib.inventory import Inventory
689
return Inventory(self.get_root_id())
575
return self.get_inventory(self.get_revision(revision_id).inventory_id)
691
return self.get_inventory(revision_id)
578
694
def revision_history(self):
590
756
That is equivalent to the number of revisions committed to
593
>>> b = ScratchBranch()
596
>>> b.commit('no foo')
600
759
return len(self.revision_history())
603
762
def last_patch(self):
604
763
"""Return last patch hash, or None if no history.
606
>>> ScratchBranch().last_patch() == None
609
765
ph = self.revision_history()
614
def lookup_revision(self, revno):
615
"""Return revision hash for revision number."""
772
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
774
If self and other have not diverged, return a list of the revisions
775
present in other, but missing from self.
777
>>> from bzrlib.commit import commit
778
>>> bzrlib.trace.silent = True
779
>>> br1 = ScratchBranch()
780
>>> br2 = ScratchBranch()
781
>>> br1.missing_revisions(br2)
783
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
784
>>> br1.missing_revisions(br2)
786
>>> br2.missing_revisions(br1)
788
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
789
>>> br1.missing_revisions(br2)
791
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
792
>>> br1.missing_revisions(br2)
794
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
795
>>> br1.missing_revisions(br2)
796
Traceback (most recent call last):
797
DivergedBranches: These branches have diverged.
799
self_history = self.revision_history()
800
self_len = len(self_history)
801
other_history = other.revision_history()
802
other_len = len(other_history)
803
common_index = min(self_len, other_len) -1
804
if common_index >= 0 and \
805
self_history[common_index] != other_history[common_index]:
806
raise DivergedBranches(self, other)
808
if stop_revision is None:
809
stop_revision = other_len
810
elif stop_revision > other_len:
811
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
813
return other_history[self_len:stop_revision]
816
def update_revisions(self, other, stop_revision=None):
817
"""Pull in all new revisions from other branch.
819
from bzrlib.fetch import greedy_fetch
820
from bzrlib.revision import get_intervening_revisions
822
pb = bzrlib.ui.ui_factory.progress_bar()
823
pb.update('comparing histories')
824
if stop_revision is None:
825
other_revision = other.last_patch()
827
other_revision = other.get_rev_id(stop_revision)
828
count = greedy_fetch(self, other, other_revision, pb)[0]
830
revision_ids = self.missing_revisions(other, stop_revision)
831
except DivergedBranches, e:
833
revision_ids = get_intervening_revisions(self.last_patch(),
834
other_revision, self)
835
assert self.last_patch() not in revision_ids
836
except bzrlib.errors.NotAncestor:
839
self.append_revision(*revision_ids)
842
def install_revisions(self, other, revision_ids, pb):
843
if hasattr(other.revision_store, "prefetch"):
844
other.revision_store.prefetch(revision_ids)
845
if hasattr(other.inventory_store, "prefetch"):
847
for rev_id in revision_ids:
849
revision = other.get_revision(rev_id).inventory_id
850
inventory_ids.append(revision)
851
except bzrlib.errors.NoSuchRevision:
853
other.inventory_store.prefetch(inventory_ids)
856
pb = bzrlib.ui.ui_factory.progress_bar()
863
for i, rev_id in enumerate(revision_ids):
864
pb.update('fetching revision', i+1, len(revision_ids))
866
rev = other.get_revision(rev_id)
867
except bzrlib.errors.NoSuchRevision:
871
revisions.append(rev)
872
inv = other.get_inventory(str(rev.inventory_id))
873
for key, entry in inv.iter_entries():
874
if entry.text_id is None:
876
if entry.text_id not in self.text_store:
877
needed_texts.add(entry.text_id)
881
count, cp_fail = self.text_store.copy_multi(other.text_store,
883
#print "Added %d texts." % count
884
inventory_ids = [ f.inventory_id for f in revisions ]
885
count, cp_fail = self.inventory_store.copy_multi(other.inventory_store,
887
#print "Added %d inventories." % count
888
revision_ids = [ f.revision_id for f in revisions]
890
count, cp_fail = self.revision_store.copy_multi(other.revision_store,
893
assert len(cp_fail) == 0
894
return count, failures
897
def commit(self, *args, **kw):
898
from bzrlib.commit import commit
899
commit(self, *args, **kw)
902
def revision_id_to_revno(self, revision_id):
903
"""Given a revision id, return its revno"""
904
history = self.revision_history()
906
return history.index(revision_id) + 1
908
raise bzrlib.errors.NoSuchRevision(self, revision_id)
911
def get_rev_id(self, revno, history=None):
912
"""Find the revision id of the specified revno."""
620
# list is 0-based; revisions are 1-based
621
return self.revision_history()[revno-1]
623
bailout("no such revision %s" % revno)
916
history = self.revision_history()
917
elif revno <= 0 or revno > len(history):
918
raise bzrlib.errors.NoSuchRevision(self, revno)
919
return history[revno - 1]
626
922
def revision_tree(self, revision_id):
667
def write_log(self, show_timezone='original'):
668
"""Write out human-readable log of commits to this branch
670
:param utc: If true, show dates in universal time, not local time."""
671
## TODO: Option to choose either original, utc or local timezone
674
for p in self.revision_history():
676
print 'revno:', revno
677
## TODO: Show hash if --id is given.
678
##print 'revision-hash:', p
679
rev = self.get_revision(p)
680
print 'committer:', rev.committer
681
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
684
## opportunistic consistency check, same as check_patch_chaining
685
if rev.precursor != precursor:
686
bailout("mismatched precursor!")
690
print ' (no message)'
692
for l in rev.message.split('\n'):
700
def show_status(branch, show_all=False):
701
"""Display single-line status for non-ignored working files.
703
The list is show sorted in order by file name.
705
>>> b = ScratchBranch(files=['foo', 'foo~'])
711
>>> b.commit("add foo")
713
>>> os.unlink(b.abspath('foo'))
718
:todo: Get state for single files.
720
:todo: Perhaps show a slash at the end of directory names.
724
# We have to build everything into a list first so that it can
725
# sorted by name, incorporating all the different sources.
727
# FIXME: Rather than getting things in random order and then sorting,
728
# just step through in order.
730
# Interesting case: the old ID for a file has been removed,
731
# but a new file has been created under that name.
733
old = branch.basis_tree()
734
old_inv = old.inventory
735
new = branch.working_tree()
736
new_inv = new.inventory
738
for fs, fid, oldname, newname, kind in diff_trees(old, new):
740
show_status(fs, kind,
741
oldname + ' => ' + newname)
742
elif fs == 'A' or fs == 'M':
743
show_status(fs, kind, newname)
745
show_status(fs, kind, oldname)
748
show_status(fs, kind, newname)
751
show_status(fs, kind, newname)
753
show_status(fs, kind, newname)
755
bailout("wierd file state %r" % ((fs, fid),))
759
class ScratchBranch(Branch):
955
def rename_one(self, from_rel, to_rel):
958
This can change the directory or the filename or both.
962
tree = self.working_tree()
964
if not tree.has_filename(from_rel):
965
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
966
if tree.has_filename(to_rel):
967
raise BzrError("can't rename: new working file %r already exists" % to_rel)
969
file_id = inv.path2id(from_rel)
971
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
973
if inv.path2id(to_rel):
974
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
976
to_dir, to_tail = os.path.split(to_rel)
977
to_dir_id = inv.path2id(to_dir)
978
if to_dir_id == None and to_dir != '':
979
raise BzrError("can't determine destination directory id for %r" % to_dir)
981
mutter("rename_one:")
982
mutter(" file_id {%s}" % file_id)
983
mutter(" from_rel %r" % from_rel)
984
mutter(" to_rel %r" % to_rel)
985
mutter(" to_dir %r" % to_dir)
986
mutter(" to_dir_id {%s}" % to_dir_id)
988
inv.rename(file_id, to_dir_id, to_tail)
990
from_abs = self.abspath(from_rel)
991
to_abs = self.abspath(to_rel)
993
os.rename(from_abs, to_abs)
995
raise BzrError("failed to rename %r to %r: %s"
996
% (from_abs, to_abs, e[1]),
997
["rename rolled back"])
999
self._write_inventory(inv)
1004
def move(self, from_paths, to_name):
1007
to_name must exist as a versioned directory.
1009
If to_name exists and is a directory, the files are moved into
1010
it, keeping their old names. If it is a directory,
1012
Note that to_name is only the last component of the new name;
1013
this doesn't change the directory.
1015
This returns a list of (from_path, to_path) pairs for each
1016
entry that is moved.
1021
## TODO: Option to move IDs only
1022
assert not isinstance(from_paths, basestring)
1023
tree = self.working_tree()
1024
inv = tree.inventory
1025
to_abs = self.abspath(to_name)
1026
if not isdir(to_abs):
1027
raise BzrError("destination %r is not a directory" % to_abs)
1028
if not tree.has_filename(to_name):
1029
raise BzrError("destination %r not in working directory" % to_abs)
1030
to_dir_id = inv.path2id(to_name)
1031
if to_dir_id == None and to_name != '':
1032
raise BzrError("destination %r is not a versioned directory" % to_name)
1033
to_dir_ie = inv[to_dir_id]
1034
if to_dir_ie.kind not in ('directory', 'root_directory'):
1035
raise BzrError("destination %r is not a directory" % to_abs)
1037
to_idpath = inv.get_idpath(to_dir_id)
1039
for f in from_paths:
1040
if not tree.has_filename(f):
1041
raise BzrError("%r does not exist in working tree" % f)
1042
f_id = inv.path2id(f)
1044
raise BzrError("%r is not versioned" % f)
1045
name_tail = splitpath(f)[-1]
1046
dest_path = appendpath(to_name, name_tail)
1047
if tree.has_filename(dest_path):
1048
raise BzrError("destination %r already exists" % dest_path)
1049
if f_id in to_idpath:
1050
raise BzrError("can't move %r to a subdirectory of itself" % f)
1052
# OK, so there's a race here, it's possible that someone will
1053
# create a file in this interval and then the rename might be
1054
# left half-done. But we should have caught most problems.
1056
for f in from_paths:
1057
name_tail = splitpath(f)[-1]
1058
dest_path = appendpath(to_name, name_tail)
1059
result.append((f, dest_path))
1060
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1062
os.rename(self.abspath(f), self.abspath(dest_path))
1064
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1065
["rename rolled back"])
1067
self._write_inventory(inv)
1074
def revert(self, filenames, old_tree=None, backups=True):
1075
"""Restore selected files to the versions from a previous tree.
1078
If true (default) backups are made of files before
1081
from bzrlib.errors import NotVersionedError, BzrError
1082
from bzrlib.atomicfile import AtomicFile
1083
from bzrlib.osutils import backup_file
1085
inv = self.read_working_inventory()
1086
if old_tree is None:
1087
old_tree = self.basis_tree()
1088
old_inv = old_tree.inventory
1091
for fn in filenames:
1092
file_id = inv.path2id(fn)
1094
raise NotVersionedError("not a versioned file", fn)
1095
if not old_inv.has_id(file_id):
1096
raise BzrError("file not present in old tree", fn, file_id)
1097
nids.append((fn, file_id))
1099
# TODO: Rename back if it was previously at a different location
1101
# TODO: If given a directory, restore the entire contents from
1102
# the previous version.
1104
# TODO: Make a backup to a temporary file.
1106
# TODO: If the file previously didn't exist, delete it?
1107
for fn, file_id in nids:
1110
f = AtomicFile(fn, 'wb')
1112
f.write(old_tree.get_file(file_id).read())
1118
def pending_merges(self):
1119
"""Return a list of pending merges.
1121
These are revisions that have been merged into the working
1122
directory but not yet committed.
1124
cfn = self.controlfilename('pending-merges')
1125
if not os.path.exists(cfn):
1128
for l in self.controlfile('pending-merges', 'r').readlines():
1129
p.append(l.rstrip('\n'))
1133
def add_pending_merge(self, revision_id):
1134
from bzrlib.revision import validate_revision_id
1136
validate_revision_id(revision_id)
1138
p = self.pending_merges()
1139
if revision_id in p:
1141
p.append(revision_id)
1142
self.set_pending_merges(p)
1145
def set_pending_merges(self, rev_list):
1146
from bzrlib.atomicfile import AtomicFile
1149
f = AtomicFile(self.controlfilename('pending-merges'))
1160
def get_parent(self):
1161
"""Return the parent location of the branch.
1163
This is the default location for push/pull/missing. The usual
1164
pattern is that the user can override it by specifying a
1168
_locs = ['parent', 'pull', 'x-pull']
1171
return self.controlfile(l, 'r').read().strip('\n')
1173
if e.errno != errno.ENOENT:
1178
def set_parent(self, url):
1179
# TODO: Maybe delete old location files?
1180
from bzrlib.atomicfile import AtomicFile
1183
f = AtomicFile(self.controlfilename('parent'))
1192
def check_revno(self, revno):
1194
Check whether a revno corresponds to any revision.
1195
Zero (the NULL revision) is considered valid.
1198
self.check_real_revno(revno)
1200
def check_real_revno(self, revno):
1202
Check whether a revno corresponds to a real revision.
1203
Zero (the NULL revision) is considered invalid
1205
if revno < 1 or revno > self.revno():
1206
raise InvalidRevisionNumber(revno)
1212
class ScratchBranch(LocalBranch):
760
1213
"""Special test class: a branch that cleans up after itself.
762
1215
>>> b = ScratchBranch()
763
1216
>>> isdir(b.base)
770
def __init__(self, files=[], dirs=[]):
1223
def __init__(self, files=[], dirs=[], base=None):
771
1224
"""Make a test branch.
773
1226
This creates a temporary directory and runs init-tree in it.
775
1228
If any files are listed, they are created in the working copy.
777
Branch.__init__(self, tempfile.mkdtemp(), init=True)
1230
from tempfile import mkdtemp
1235
LocalBranch.__init__(self, base, init=init)
779
1237
os.mkdir(self.abspath(d))
810
def _gen_revision_id(when):
811
"""Return new revision-id."""
812
s = '%s-%s-' % (user_email(), compact_date(when))
813
s += hexlify(rand_bytes(8))
817
1302
def gen_file_id(name):
818
1303
"""Return new file id.
820
1305
This should probably generate proper UUIDs, but for the moment we
821
1306
cope with just randomness because running uuidgen every time is
1309
from binascii import hexlify
1310
from time import time
1312
# get last component
823
1313
idx = name.rfind('/')
825
1315
name = name[idx+1 : ]
1316
idx = name.rfind('\\')
1318
name = name[idx+1 : ]
1320
# make it not a hidden file
827
1321
name = name.lstrip('.')
1323
# remove any wierd characters; we don't escape them but rather
1324
# just pull them out
1325
name = re.sub(r'[^\w.]', '', name)
829
1327
s = hexlify(rand_bytes(8))
830
return '-'.join((name, compact_date(time.time()), s))
1328
return '-'.join((name, compact_date(time()), s))
1332
"""Return a new tree-root file id."""
1333
return gen_file_id('TREE_ROOT')
1336
def copy_branch(branch_from, to_location, revision=None):
1337
"""Copy branch_from into the existing directory to_location.
1340
If not None, only revisions up to this point will be copied.
1341
The head of the new branch will be that revision.
1344
The name of a local directory that exists but is empty.
1346
from bzrlib.merge import merge
1347
from bzrlib.revisionspec import RevisionSpec
1349
assert isinstance(branch_from, Branch)
1350
assert isinstance(to_location, basestring)
1352
br_to = Branch.initialize(to_location)
1353
br_to.set_root_id(branch_from.get_root_id())
1354
if revision is None:
1355
revno = branch_from.revno()
1357
revno, rev_id = RevisionSpec(revision).in_history(branch_from)
1358
br_to.update_revisions(branch_from, stop_revision=revno)
1359
merge((to_location, -1), (to_location, 0), this_dir=to_location,
1360
check_clean=False, ignore_zero=True)
1361
br_to.set_parent(branch_from.base)