15
16
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
22
import traceback, socket, fnmatch, difflib, time
23
from binascii import hexlify
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
26
from inventory import Inventory
27
from trace import mutter, note
28
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
29
from inventory import InventoryEntry, Inventory
30
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, chomp, \
31
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
32
joinpath, sha_string, file_kind, local_time_offset
33
from store import ImmutableStore
34
from revision import Revision
35
from errors import bailout
36
from textui import show_status
37
from diff import diff_trees
38
39
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
39
40
## 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)
103
def find_branch_root(f=None):
104
"""Find the branch root enclosing f, or pwd.
106
f may be a filename or a URL.
108
It is not necessary that f exists.
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.
115
elif hasattr(os.path, 'realpath'):
116
f = os.path.realpath(f)
118
f = os.path.abspath(f)
119
if not os.path.exists(f):
120
raise BzrError('%r does not exist' % f)
126
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
128
head, tail = os.path.split(f)
130
# reached the root, whatever that may be
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.")
144
46
######################################################################
147
class Branch(object):
148
50
"""Branch holding a history of revisions.
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.
52
:todo: Perhaps use different stores for different classes of object,
53
so that we can keep track of how much space each one uses,
54
or garbage-collect them.
56
:todo: Add a RemoteBranch subclass. For the basic case of read-only
57
HTTP access this should be very easy by,
58
just redirecting controlfile access into HTTP requests.
59
We would need a RemoteStore working similarly.
61
:todo: Keep the on-disk branch locked while the object exists.
63
:todo: mkdir() method.
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 = {}
173
def __init__(self, base, init=False, find_root=True):
65
def __init__(self, base, init=False):
174
66
"""Create new branch object at a particular location.
176
base -- Base directory for the branch. May be a file:// url.
178
init -- If True, create new control files in a previously
68
:param base: Base directory for the branch.
70
:param init: If True, create new control files in a previously
179
71
unversioned directory. If False, the branch must already
182
find_root -- If true and init is false, find the root of the
183
existing branch containing base.
185
74
In the test suite, creation of new trees is tested using the
186
75
`ScratchBranch` class.
188
from bzrlib.store import ImmutableStore
77
self.base = os.path.realpath(base)
190
self.base = os.path.realpath(base)
191
79
self._make_control()
193
self.base = find_branch_root(base)
195
if base.startswith("file://"):
197
self.base = os.path.realpath(base)
198
81
if not isdir(self.controlfilename('.')):
199
from errors import NotBranchError
200
raise NotBranchError("not a bzr branch: %s" % quotefn(base),
201
['use "bzr init" to initialize a new working tree',
202
'current bzr can only operate from top-of-tree'])
82
bailout("not a bzr branch: %s" % quotefn(base),
83
['use "bzr init" to initialize a new working tree',
84
'current bzr can only operate from top-of-tree'])
205
87
self.text_store = ImmutableStore(self.controlfilename('text-store'))
206
88
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
214
96
__repr__ = __str__
218
if self._lock_mode or self._lock:
219
from warnings import warn
220
warn("branch %r was not explicitly unlocked" % self)
223
def lock_write(self):
225
if self._lock_mode != 'w':
226
from errors import LockError
227
raise LockError("can't upgrade to a write lock from %r" %
229
self._lock_count += 1
231
from bzrlib.lock import WriteLock
233
self._lock = WriteLock(self.controlfilename('branch-lock'))
234
self._lock_mode = 'w'
240
assert self._lock_mode in ('r', 'w'), \
241
"invalid lock mode %r" % self._lock_mode
242
self._lock_count += 1
244
from bzrlib.lock import ReadLock
246
self._lock = ReadLock(self.controlfilename('branch-lock'))
247
self._lock_mode = 'r'
251
if not self._lock_mode:
252
from errors import LockError
253
raise LockError('branch %r is not locked' % (self))
255
if self._lock_count > 1:
256
self._lock_count -= 1
260
self._lock_mode = self._lock_count = None
262
def abspath(self, name):
263
"""Return absolute filename for something in the branch"""
100
"""Return filename relative to branch top"""
264
101
return os.path.join(self.base, name)
266
def relpath(self, path):
267
"""Return path relative to this branch of something inside it.
269
Raises an error if path is not in this branch."""
270
return _relpath(self.base, path)
272
104
def controlfilename(self, file_or_path):
273
105
"""Return location relative to branch."""
274
if isinstance(file_or_path, basestring):
106
if isinstance(file_or_path, types.StringTypes):
275
107
file_or_path = [file_or_path]
276
108
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
279
111
def controlfile(self, file_or_path, mode='r'):
280
"""Open a control file for this branch.
282
There are two classes of file in the control directory: text
283
and binary. binary files are untranslated byte streams. Text
284
control files are stored with Unix newlines and in UTF-8, even
285
if the platform or locale defaults are different.
287
Controlfiles should almost never be opened in write mode but
288
rather should be atomically copied and replaced using atomicfile.
291
fn = self.controlfilename(file_or_path)
293
if mode == 'rb' or mode == 'wb':
294
return file(fn, mode)
295
elif mode == 'r' or mode == 'w':
296
# open in binary mode anyhow so there's no newline translation;
297
# codecs uses line buffering by default; don't want that.
299
return codecs.open(fn, mode + 'b', 'utf-8',
302
raise BzrError("invalid controlfile mode %r" % mode)
112
"""Open a control file for this branch"""
113
return file(self.controlfilename(file_or_path), mode)
304
116
def _make_control(self):
305
from bzrlib.inventory import Inventory
306
from bzrlib.xml import pack_xml
308
117
os.mkdir(self.controlfilename([]))
309
118
self.controlfile('README', 'w').write(
310
119
"This is a Bazaar-NG control directory.\n"
311
"Do not change any files in this directory.\n")
120
"Do not change any files in this directory.")
312
121
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
313
122
for d in ('text-store', 'inventory-store', 'revision-store'):
314
123
os.mkdir(self.controlfilename(d))
315
124
for f in ('revision-history', 'merged-patches',
316
'pending-merged-patches', 'branch-name',
125
'pending-merged-patches', 'branch-name'):
319
126
self.controlfile(f, 'w').write('')
320
127
mutter('created control directory in ' + self.base)
128
Inventory().write_xml(self.controlfile('inventory','w'))
322
# if we want per-tree root ids then this is the place to set
323
# them; they're not needed for now and so ommitted for
325
pack_xml(Inventory(), self.controlfile('inventory','w'))
327
131
def _check_format(self):
328
132
"""Check this branch format is supported.
332
136
In the future, we might need different in-memory Branch
333
137
classes to support downlevel branches. But not yet.
335
# This ignores newlines so that we can open branches created
336
# on Windows from Linux and so on. I think it might be better
337
# to always make all internal files in unix format.
338
fmt = self.controlfile('branch-format', 'r').read()
339
fmt.replace('\r\n', '')
139
# read in binary mode to detect newline wierdness.
140
fmt = self.controlfile('branch-format', 'rb').read()
340
141
if fmt != BZR_BRANCH_FORMAT:
341
raise BzrError('sorry, branch format %r not supported' % fmt,
342
['use a different bzr version',
343
'or remove the .bzr directory and "bzr init" again'])
345
def get_root_id(self):
346
"""Return the id of this branches root"""
347
inv = self.read_working_inventory()
348
return inv.root.file_id
350
def set_root_id(self, file_id):
351
inv = self.read_working_inventory()
352
orig_root_id = inv.root.file_id
353
del inv._byid[inv.root.file_id]
354
inv.root.file_id = file_id
355
inv._byid[inv.root.file_id] = inv.root
358
if entry.parent_id in (None, orig_root_id):
359
entry.parent_id = inv.root.file_id
360
self._write_inventory(inv)
142
bailout('sorry, branch format %r not supported' % fmt,
143
['use a different bzr version',
144
'or remove the .bzr directory and "bzr init" again'])
362
147
def read_working_inventory(self):
363
148
"""Read the working inventory."""
364
from bzrlib.inventory import Inventory
365
from bzrlib.xml import unpack_xml
366
from time import time
370
# ElementTree does its own conversion from UTF-8, so open in
372
inv = unpack_xml(Inventory,
373
self.controlfile('inventory', 'rb'))
374
mutter("loaded inventory of %d items in %f"
375
% (len(inv), time() - before))
150
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
151
mutter("loaded inventory of %d items in %f"
152
% (len(inv), time.time() - before))
381
156
def _write_inventory(self, inv):
382
157
"""Update the working inventory.
384
159
That is to say, the inventory describing changes underway, that
385
160
will be committed to the next revision.
387
from bzrlib.atomicfile import AtomicFile
388
from bzrlib.xml import pack_xml
392
f = AtomicFile(self.controlfilename('inventory'), 'wb')
401
mutter('wrote working inventory')
162
inv.write_xml(self.controlfile('inventory', 'w'))
163
mutter('wrote inventory to %s' % quotefn(self.controlfilename('inventory')))
404
166
inventory = property(read_working_inventory, _write_inventory, None,
405
167
"""Inventory for the working copy.""")
408
def add(self, files, ids=None):
170
def add(self, files, verbose=False):
409
171
"""Make files versioned.
411
Note that the command line normally calls smart_add instead,
412
which can automatically recurse.
414
173
This puts the files in the Added state, so that they will be
415
174
recorded by the next commit.
418
List of paths to add, relative to the base of the tree.
421
If set, use these instead of automatically generated ids.
422
Must be the same length as the list of files, but may
423
contain None for ids that are to be autogenerated.
425
TODO: Perhaps have an option to add the ids even if the files do
428
TODO: Perhaps yield the ids and paths as they're added.
176
:todo: Perhaps have an option to add the ids even if the files do
179
:todo: Perhaps return the ids of the files? But then again it
180
is easy to retrieve them if they're needed.
182
:todo: Option to specify file id.
184
:todo: Adding a directory should optionally recurse down and
185
add all non-ignored children. Perhaps do that in a
188
>>> b = ScratchBranch(files=['foo'])
189
>>> 'foo' in b.unknowns()
194
>>> 'foo' in b.unknowns()
196
>>> bool(b.inventory.path2id('foo'))
202
Traceback (most recent call last):
204
BzrError: ('foo is already versioned', [])
206
>>> b.add(['nothere'])
207
Traceback (most recent call last):
208
BzrError: ('cannot add: not a regular file or directory: nothere', [])
430
211
# TODO: Re-adding a file that is removed in the working copy
431
212
# should probably put it back with the previous ID.
432
if isinstance(files, basestring):
433
assert(ids is None or isinstance(ids, basestring))
213
if isinstance(files, types.StringTypes):
439
ids = [None] * len(files)
441
assert(len(ids) == len(files))
445
inv = self.read_working_inventory()
446
for f,file_id in zip(files, ids):
447
if is_control_file(f):
448
raise BzrError("cannot add control file %s" % quotefn(f))
453
raise BzrError("cannot add top-level %r" % f)
455
fullpath = os.path.normpath(self.abspath(f))
458
kind = file_kind(fullpath)
460
# maybe something better?
461
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
463
if kind != 'file' and kind != 'directory':
464
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
467
file_id = gen_file_id(f)
468
inv.add_path(f, kind=kind, file_id=file_id)
470
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
472
self._write_inventory(inv)
477
def print_file(self, file, revno):
478
"""Print `file` to stdout."""
481
tree = self.revision_tree(self.lookup_revision(revno))
482
# use inventory as it was in that revision
483
file_id = tree.inventory.path2id(file)
485
raise BzrError("%r is not present in revision %s" % (file, revno))
486
tree.print_file(file_id)
216
inv = self.read_working_inventory()
218
if is_control_file(f):
219
bailout("cannot add control file %s" % quotefn(f))
224
bailout("cannot add top-level %r" % f)
226
fullpath = os.path.normpath(self._rel(f))
230
elif isdir(fullpath):
233
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
236
parent_name = joinpath(fp[:-1])
237
mutter("lookup parent %r" % parent_name)
238
parent_id = inv.path2id(parent_name)
239
if parent_id == None:
240
bailout("cannot add: parent %r is not versioned"
245
file_id = _gen_file_id(fp[-1])
246
inv.add(InventoryEntry(file_id, fp[-1], kind=kind, parent_id=parent_id))
248
show_status('A', kind, quotefn(f))
250
mutter("add file %s file_id:{%s} kind=%r parent_id={%s}"
251
% (f, file_id, kind, parent_id))
252
self._write_inventory(inv)
491
256
def remove(self, files, verbose=False):
494
259
This does not remove their text. This does not run on
496
TODO: Refuse to remove modified files unless --force is given?
498
TODO: Do something useful with directories.
500
TODO: Should this remove the text or not? Tough call; not
261
:todo: Refuse to remove modified files unless --force is given?
263
>>> b = ScratchBranch(files=['foo'])
265
>>> b.inventory.has_filename('foo')
268
>>> b.working_tree().has_filename('foo')
270
>>> b.inventory.has_filename('foo')
273
>>> b = ScratchBranch(files=['foo'])
278
>>> b.inventory.has_filename('foo')
280
>>> b.basis_tree().has_filename('foo')
282
>>> b.working_tree().has_filename('foo')
285
:todo: Do something useful with directories.
287
:todo: Should this remove the text or not? Tough call; not
501
288
removing may be useful and the user can just use use rm, and
502
289
is the opposite of add. Removing it is consistent with most
503
290
other tools. Maybe an option.
505
292
## TODO: Normalize names
506
293
## TODO: Remove nested loops; better scalability
507
if isinstance(files, basestring):
295
if isinstance(files, types.StringTypes):
513
tree = self.working_tree()
516
# do this before any modifications
520
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
521
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
523
# having remove it, it must be either ignored or unknown
524
if tree.is_ignored(f):
528
show_status(new_status, inv[fid].kind, quotefn(f))
531
self._write_inventory(inv)
536
# FIXME: this doesn't need to be a branch method
537
def set_inventory(self, new_inventory_list):
538
from bzrlib.inventory import Inventory, InventoryEntry
539
inv = Inventory(self.get_root_id())
540
for path, file_id, parent, kind in new_inventory_list:
541
name = os.path.basename(path)
544
inv.add(InventoryEntry(file_id, name, kind, parent))
298
inv = self.read_working_inventory()
300
# do this before any modifications
304
bailout("cannot remove unversioned file %s" % quotefn(f))
305
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
307
show_status('D', inv[fid].kind, quotefn(f))
545
310
self._write_inventory(inv)
564
329
return self.working_tree().unknowns()
567
def append_revision(self, *revision_ids):
568
from bzrlib.atomicfile import AtomicFile
570
for revision_id in revision_ids:
571
mutter("add {%s} to revision-history" % revision_id)
573
rev_history = self.revision_history()
574
rev_history.extend(revision_ids)
576
f = AtomicFile(self.controlfilename('revision-history'))
578
for rev_id in rev_history:
584
def get_revision_xml(self, revision_id):
585
"""Return XML file object for revision object."""
586
if not revision_id or not isinstance(revision_id, basestring):
587
raise InvalidRevisionId(revision_id)
592
return self.revision_store[revision_id]
594
raise bzrlib.errors.NoSuchRevision(self, revision_id)
332
def commit(self, message, timestamp=None, timezone=None,
335
"""Commit working copy as a new revision.
337
The basic approach is to add all the file texts into the
338
store, then the inventory, then make a new revision pointing
339
to that inventory and store that.
341
This is not quite safe if the working copy changes during the
342
commit; for the moment that is simply not allowed. A better
343
approach is to make a temporary copy of the files before
344
computing their hashes, and then add those hashes in turn to
345
the inventory. This should mean at least that there are no
346
broken hash pointers. There is no way we can get a snapshot
347
of the whole directory at an instant. This would also have to
348
be robust against files disappearing, moving, etc. So the
349
whole thing is a bit hard.
351
:param timestamp: if not None, seconds-since-epoch for a
352
postdated/predated commit.
355
## TODO: Show branch names
357
# TODO: Don't commit if there are no changes, unless forced?
359
# First walk over the working inventory; and both update that
360
# and also build a new revision inventory. The revision
361
# inventory needs to hold the text-id, sha1 and size of the
362
# actual file versions committed in the revision. (These are
363
# not present in the working inventory.) We also need to
364
# detect missing/deleted files, and remove them from the
367
work_inv = self.read_working_inventory()
369
basis = self.basis_tree()
370
basis_inv = basis.inventory
372
for path, entry in work_inv.iter_entries():
373
## TODO: Cope with files that have gone missing.
375
## TODO: Check that the file kind has not changed from the previous
376
## revision of this file (if any).
381
file_id = entry.file_id
382
mutter('commit prep file %s, id %r ' % (p, file_id))
384
if not os.path.exists(p):
385
mutter(" file is missing, removing from inventory")
387
show_status('D', entry.kind, quotefn(path))
388
missing_ids.append(file_id)
391
# TODO: Handle files that have been deleted
393
# TODO: Maybe a special case for empty files? Seems a
394
# waste to store them many times.
398
if basis_inv.has_id(file_id):
399
old_kind = basis_inv[file_id].kind
400
if old_kind != entry.kind:
401
bailout("entry %r changed kind from %r to %r"
402
% (file_id, old_kind, entry.kind))
404
if entry.kind == 'directory':
406
bailout("%s is entered as directory but not a directory" % quotefn(p))
407
elif entry.kind == 'file':
409
bailout("%s is entered as file but is not a file" % quotefn(p))
411
content = file(p, 'rb').read()
413
entry.text_sha1 = sha_string(content)
414
entry.text_size = len(content)
416
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
418
and (old_ie.text_size == entry.text_size)
419
and (old_ie.text_sha1 == entry.text_sha1)):
420
## assert content == basis.get_file(file_id).read()
421
entry.text_id = basis_inv[file_id].text_id
422
mutter(' unchanged from previous text_id {%s}' %
426
entry.text_id = _gen_file_id(entry.name)
427
self.text_store.add(content, entry.text_id)
428
mutter(' stored with text_id {%s}' % entry.text_id)
432
elif (old_ie.name == entry.name
433
and old_ie.parent_id == entry.parent_id):
438
show_status(state, entry.kind, quotefn(path))
440
for file_id in missing_ids:
441
# have to do this later so we don't mess up the iterator.
442
# since parents may be removed before their children we
445
# FIXME: There's probably a better way to do this; perhaps
446
# the workingtree should know how to filter itself.
447
if work_inv.has_id(file_id):
448
del work_inv[file_id]
451
inv_id = rev_id = _gen_revision_id(time.time())
453
inv_tmp = tempfile.TemporaryFile()
454
inv.write_xml(inv_tmp)
456
self.inventory_store.add(inv_tmp, inv_id)
457
mutter('new inventory_id is {%s}' % inv_id)
459
self._write_inventory(work_inv)
461
if timestamp == None:
462
timestamp = time.time()
464
if committer == None:
465
committer = username()
468
timezone = local_time_offset()
470
mutter("building commit log message")
471
rev = Revision(timestamp=timestamp,
474
precursor = self.last_patch(),
479
rev_tmp = tempfile.TemporaryFile()
480
rev.write_xml(rev_tmp)
482
self.revision_store.add(rev_tmp, rev_id)
483
mutter("new revision_id is {%s}" % rev_id)
485
## XXX: Everything up to here can simply be orphaned if we abort
486
## the commit; it will leave junk files behind but that doesn't
489
## TODO: Read back the just-generated changeset, and make sure it
490
## applies and recreates the right state.
492
## TODO: Also calculate and store the inventory SHA1
493
mutter("committing patch r%d" % (self.revno() + 1))
495
mutter("append to revision-history")
496
self.controlfile('revision-history', 'at').write(rev_id + '\n')
598
501
def get_revision(self, revision_id):
599
502
"""Return the Revision object for a named revision"""
600
xml_file = self.get_revision_xml(revision_id)
603
r = unpack_xml(Revision, xml_file)
604
except SyntaxError, e:
605
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
503
r = Revision.read_xml(self.revision_store[revision_id])
609
504
assert r.revision_id == revision_id
612
def get_revision_delta(self, revno):
613
"""Return the delta for one revision.
615
The delta is relative to its mainline predecessor, or the
616
empty tree for revision 1.
618
assert isinstance(revno, int)
619
rh = self.revision_history()
620
if not (1 <= revno <= len(rh)):
621
raise InvalidRevisionNumber(revno)
623
# revno is 1-based; list is 0-based
625
new_tree = self.revision_tree(rh[revno-1])
627
old_tree = EmptyTree()
629
old_tree = self.revision_tree(rh[revno-2])
631
return compare_trees(old_tree, new_tree)
635
def get_revision_sha1(self, revision_id):
636
"""Hash the stored value of a revision, and return it."""
637
# In the future, revision entries will be signed. At that
638
# point, it is probably best *not* to include the signature
639
# in the revision hash. Because that lets you re-sign
640
# the revision, (add signatures/remove signatures) and still
641
# have all hash pointers stay consistent.
642
# But for now, just hash the contents.
643
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
646
508
def get_inventory(self, inventory_id):
647
509
"""Get Inventory object by hash.
649
TODO: Perhaps for this and similar methods, take a revision
511
:todo: Perhaps for this and similar methods, take a revision
650
512
parameter which can be either an integer revno or a
652
from bzrlib.inventory import Inventory
653
from bzrlib.xml import unpack_xml
655
return unpack_xml(Inventory, self.get_inventory_xml(inventory_id))
658
def get_inventory_xml(self, inventory_id):
659
"""Get inventory XML as a file object."""
660
return self.inventory_store[inventory_id]
663
def get_inventory_sha1(self, inventory_id):
664
"""Return the sha1 hash of the inventory entry
666
return sha_file(self.get_inventory_xml(inventory_id))
514
i = Inventory.read_xml(self.inventory_store[inventory_id])
669
518
def get_revision_inventory(self, revision_id):
670
519
"""Return inventory of a past revision."""
671
# bzr 0.0.6 imposes the constraint that the inventory_id
672
# must be the same as its revision, so this is trivial.
673
520
if revision_id == None:
674
from bzrlib.inventory import Inventory
675
return Inventory(self.get_root_id())
677
return self.get_inventory(revision_id)
523
return self.get_inventory(self.get_revision(revision_id).inventory_id)
680
526
def revision_history(self):
742
538
That is equivalent to the number of revisions committed to
541
>>> b = ScratchBranch()
544
>>> b.commit('no foo')
745
548
return len(self.revision_history())
748
551
def last_patch(self):
749
552
"""Return last patch hash, or None if no history.
554
>>> ScratchBranch().last_patch() == None
751
557
ph = self.revision_history()
758
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
760
If self and other have not diverged, return a list of the revisions
761
present in other, but missing from self.
763
>>> from bzrlib.commit import commit
764
>>> bzrlib.trace.silent = True
765
>>> br1 = ScratchBranch()
766
>>> br2 = ScratchBranch()
767
>>> br1.missing_revisions(br2)
769
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
770
>>> br1.missing_revisions(br2)
772
>>> br2.missing_revisions(br1)
774
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
775
>>> br1.missing_revisions(br2)
777
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
778
>>> br1.missing_revisions(br2)
780
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
781
>>> br1.missing_revisions(br2)
782
Traceback (most recent call last):
783
DivergedBranches: These branches have diverged.
785
self_history = self.revision_history()
786
self_len = len(self_history)
787
other_history = other.revision_history()
788
other_len = len(other_history)
789
common_index = min(self_len, other_len) -1
790
if common_index >= 0 and \
791
self_history[common_index] != other_history[common_index]:
792
raise DivergedBranches(self, other)
794
if stop_revision is None:
795
stop_revision = other_len
796
elif stop_revision > other_len:
797
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
799
return other_history[self_len:stop_revision]
802
def update_revisions(self, other, stop_revision=None):
803
"""Pull in all new revisions from other branch.
805
from bzrlib.fetch import greedy_fetch
807
pb = bzrlib.ui.ui_factory.progress_bar()
808
pb.update('comparing histories')
810
revision_ids = self.missing_revisions(other, stop_revision)
812
if len(revision_ids) > 0:
813
count = greedy_fetch(self, other, revision_ids[-1], pb)[0]
816
self.append_revision(*revision_ids)
817
## note("Added %d revisions." % count)
820
def install_revisions(self, other, revision_ids, pb):
821
if hasattr(other.revision_store, "prefetch"):
822
other.revision_store.prefetch(revision_ids)
823
if hasattr(other.inventory_store, "prefetch"):
824
inventory_ids = [other.get_revision(r).inventory_id
825
for r in revision_ids]
826
other.inventory_store.prefetch(inventory_ids)
829
pb = bzrlib.ui.ui_factory.progress_bar()
836
for i, rev_id in enumerate(revision_ids):
837
pb.update('fetching revision', i+1, len(revision_ids))
839
rev = other.get_revision(rev_id)
840
except bzrlib.errors.NoSuchRevision:
844
revisions.append(rev)
845
inv = other.get_inventory(str(rev.inventory_id))
846
for key, entry in inv.iter_entries():
847
if entry.text_id is None:
849
if entry.text_id not in self.text_store:
850
needed_texts.add(entry.text_id)
854
count, cp_fail = self.text_store.copy_multi(other.text_store,
856
#print "Added %d texts." % count
857
inventory_ids = [ f.inventory_id for f in revisions ]
858
count, cp_fail = self.inventory_store.copy_multi(other.inventory_store,
860
#print "Added %d inventories." % count
861
revision_ids = [ f.revision_id for f in revisions]
863
count, cp_fail = self.revision_store.copy_multi(other.revision_store,
866
assert len(cp_fail) == 0
867
return count, failures
870
def commit(self, *args, **kw):
871
from bzrlib.commit import commit
872
commit(self, *args, **kw)
875
def lookup_revision(self, revision):
876
"""Return the revision identifier for a given revision information."""
877
revno, info = self._get_revision_info(revision)
881
def revision_id_to_revno(self, revision_id):
882
"""Given a revision id, return its revno"""
883
history = self.revision_history()
885
return history.index(revision_id) + 1
887
raise bzrlib.errors.NoSuchRevision(self, revision_id)
890
def get_revision_info(self, revision):
891
"""Return (revno, revision id) for revision identifier.
893
revision can be an integer, in which case it is assumed to be revno (though
894
this will translate negative values into positive ones)
895
revision can also be a string, in which case it is parsed for something like
896
'date:' or 'revid:' etc.
898
revno, rev_id = self._get_revision_info(revision)
900
raise bzrlib.errors.NoSuchRevision(self, revision)
903
def get_rev_id(self, revno, history=None):
904
"""Find the revision id of the specified revno."""
562
def lookup_revision(self, revno):
563
"""Return revision hash for revision number."""
908
history = self.revision_history()
909
elif revno <= 0 or revno > len(history):
910
raise bzrlib.errors.NoSuchRevision(self, revno)
911
return history[revno - 1]
913
def _get_revision_info(self, revision):
914
"""Return (revno, revision id) for revision specifier.
916
revision can be an integer, in which case it is assumed to be revno
917
(though this will translate negative values into positive ones)
918
revision can also be a string, in which case it is parsed for something
919
like 'date:' or 'revid:' etc.
921
A revid is always returned. If it is None, the specifier referred to
922
the null revision. If the revid does not occur in the revision
923
history, revno will be None.
929
try:# Convert to int if possible
930
revision = int(revision)
933
revs = self.revision_history()
934
if isinstance(revision, int):
936
revno = len(revs) + revision + 1
939
rev_id = self.get_rev_id(revno, revs)
940
elif isinstance(revision, basestring):
941
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
942
if revision.startswith(prefix):
943
result = func(self, revs, revision)
945
revno, rev_id = result
948
rev_id = self.get_rev_id(revno, revs)
951
raise BzrError('No namespace registered for string: %r' %
954
raise TypeError('Unhandled revision type %s' % revision)
958
raise bzrlib.errors.NoSuchRevision(self, revision)
961
def _namespace_revno(self, revs, revision):
962
"""Lookup a revision by revision number"""
963
assert revision.startswith('revno:')
965
return (int(revision[6:]),)
968
REVISION_NAMESPACES['revno:'] = _namespace_revno
970
def _namespace_revid(self, revs, revision):
971
assert revision.startswith('revid:')
972
rev_id = revision[len('revid:'):]
974
return revs.index(rev_id) + 1, rev_id
977
REVISION_NAMESPACES['revid:'] = _namespace_revid
979
def _namespace_last(self, revs, revision):
980
assert revision.startswith('last:')
982
offset = int(revision[5:])
987
raise BzrError('You must supply a positive value for --revision last:XXX')
988
return (len(revs) - offset + 1,)
989
REVISION_NAMESPACES['last:'] = _namespace_last
991
def _namespace_tag(self, revs, revision):
992
assert revision.startswith('tag:')
993
raise BzrError('tag: namespace registered, but not implemented.')
994
REVISION_NAMESPACES['tag:'] = _namespace_tag
996
def _namespace_date(self, revs, revision):
997
assert revision.startswith('date:')
999
# Spec for date revisions:
1001
# value can be 'yesterday', 'today', 'tomorrow' or a YYYY-MM-DD string.
1002
# it can also start with a '+/-/='. '+' says match the first
1003
# entry after the given date. '-' is match the first entry before the date
1004
# '=' is match the first entry after, but still on the given date.
1006
# +2005-05-12 says find the first matching entry after May 12th, 2005 at 0:00
1007
# -2005-05-12 says find the first matching entry before May 12th, 2005 at 0:00
1008
# =2005-05-12 says find the first match after May 12th, 2005 at 0:00 but before
1009
# May 13th, 2005 at 0:00
1011
# So the proper way of saying 'give me all entries for today' is:
1012
# -r {date:+today}:{date:-tomorrow}
1013
# The default is '=' when not supplied
1016
if val[:1] in ('+', '-', '='):
1017
match_style = val[:1]
1020
today = datetime.datetime.today().replace(hour=0,minute=0,second=0,microsecond=0)
1021
if val.lower() == 'yesterday':
1022
dt = today - datetime.timedelta(days=1)
1023
elif val.lower() == 'today':
1025
elif val.lower() == 'tomorrow':
1026
dt = today + datetime.timedelta(days=1)
1029
# This should be done outside the function to avoid recompiling it.
1030
_date_re = re.compile(
1031
r'(?P<date>(?P<year>\d\d\d\d)-(?P<month>\d\d)-(?P<day>\d\d))?'
1033
r'(?P<time>(?P<hour>\d\d):(?P<minute>\d\d)(:(?P<second>\d\d))?)?'
1035
m = _date_re.match(val)
1036
if not m or (not m.group('date') and not m.group('time')):
1037
raise BzrError('Invalid revision date %r' % revision)
1040
year, month, day = int(m.group('year')), int(m.group('month')), int(m.group('day'))
1042
year, month, day = today.year, today.month, today.day
1044
hour = int(m.group('hour'))
1045
minute = int(m.group('minute'))
1046
if m.group('second'):
1047
second = int(m.group('second'))
1051
hour, minute, second = 0,0,0
1053
dt = datetime.datetime(year=year, month=month, day=day,
1054
hour=hour, minute=minute, second=second)
1058
if match_style == '-':
1060
elif match_style == '=':
1061
last = dt + datetime.timedelta(days=1)
1064
for i in range(len(revs)-1, -1, -1):
1065
r = self.get_revision(revs[i])
1066
# TODO: Handle timezone.
1067
dt = datetime.datetime.fromtimestamp(r.timestamp)
1068
if first >= dt and (last is None or dt >= last):
1071
for i in range(len(revs)):
1072
r = self.get_revision(revs[i])
1073
# TODO: Handle timezone.
1074
dt = datetime.datetime.fromtimestamp(r.timestamp)
1075
if first <= dt and (last is None or dt <= last):
1077
REVISION_NAMESPACES['date:'] = _namespace_date
568
# list is 0-based; revisions are 1-based
569
return self.revision_history()[revno-1]
571
bailout("no such revision %s" % revno)
1079
574
def revision_tree(self, revision_id):
1080
575
"""Return Tree for a revision on this branch.
1082
577
`revision_id` may be None for the null revision, in which case
1083
578
an `EmptyTree` is returned."""
1084
# TODO: refactor this to use an existing revision object
1085
# so we don't need to read it in twice.
1086
580
if revision_id == None:
1087
581
return EmptyTree()
1112
def rename_one(self, from_rel, to_rel):
1115
This can change the directory or the filename or both.
1119
tree = self.working_tree()
1120
inv = tree.inventory
1121
if not tree.has_filename(from_rel):
1122
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1123
if tree.has_filename(to_rel):
1124
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1126
file_id = inv.path2id(from_rel)
1128
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1130
if inv.path2id(to_rel):
1131
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1133
to_dir, to_tail = os.path.split(to_rel)
1134
to_dir_id = inv.path2id(to_dir)
1135
if to_dir_id == None and to_dir != '':
1136
raise BzrError("can't determine destination directory id for %r" % to_dir)
1138
mutter("rename_one:")
1139
mutter(" file_id {%s}" % file_id)
1140
mutter(" from_rel %r" % from_rel)
1141
mutter(" to_rel %r" % to_rel)
1142
mutter(" to_dir %r" % to_dir)
1143
mutter(" to_dir_id {%s}" % to_dir_id)
1145
inv.rename(file_id, to_dir_id, to_tail)
1147
from_abs = self.abspath(from_rel)
1148
to_abs = self.abspath(to_rel)
1150
os.rename(from_abs, to_abs)
1152
raise BzrError("failed to rename %r to %r: %s"
1153
% (from_abs, to_abs, e[1]),
1154
["rename rolled back"])
1156
self._write_inventory(inv)
1161
def move(self, from_paths, to_name):
1164
to_name must exist as a versioned directory.
1166
If to_name exists and is a directory, the files are moved into
1167
it, keeping their old names. If it is a directory,
1169
Note that to_name is only the last component of the new name;
1170
this doesn't change the directory.
1172
This returns a list of (from_path, to_path) pairs for each
1173
entry that is moved.
1178
## TODO: Option to move IDs only
1179
assert not isinstance(from_paths, basestring)
1180
tree = self.working_tree()
1181
inv = tree.inventory
1182
to_abs = self.abspath(to_name)
1183
if not isdir(to_abs):
1184
raise BzrError("destination %r is not a directory" % to_abs)
1185
if not tree.has_filename(to_name):
1186
raise BzrError("destination %r not in working directory" % to_abs)
1187
to_dir_id = inv.path2id(to_name)
1188
if to_dir_id == None and to_name != '':
1189
raise BzrError("destination %r is not a versioned directory" % to_name)
1190
to_dir_ie = inv[to_dir_id]
1191
if to_dir_ie.kind not in ('directory', 'root_directory'):
1192
raise BzrError("destination %r is not a directory" % to_abs)
1194
to_idpath = inv.get_idpath(to_dir_id)
1196
for f in from_paths:
1197
if not tree.has_filename(f):
1198
raise BzrError("%r does not exist in working tree" % f)
1199
f_id = inv.path2id(f)
1201
raise BzrError("%r is not versioned" % f)
1202
name_tail = splitpath(f)[-1]
1203
dest_path = appendpath(to_name, name_tail)
1204
if tree.has_filename(dest_path):
1205
raise BzrError("destination %r already exists" % dest_path)
1206
if f_id in to_idpath:
1207
raise BzrError("can't move %r to a subdirectory of itself" % f)
1209
# OK, so there's a race here, it's possible that someone will
1210
# create a file in this interval and then the rename might be
1211
# left half-done. But we should have caught most problems.
1213
for f in from_paths:
1214
name_tail = splitpath(f)[-1]
1215
dest_path = appendpath(to_name, name_tail)
1216
result.append((f, dest_path))
1217
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1219
os.rename(self.abspath(f), self.abspath(dest_path))
1221
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1222
["rename rolled back"])
1224
self._write_inventory(inv)
1231
def revert(self, filenames, old_tree=None, backups=True):
1232
"""Restore selected files to the versions from a previous tree.
1235
If true (default) backups are made of files before
1238
from bzrlib.errors import NotVersionedError, BzrError
1239
from bzrlib.atomicfile import AtomicFile
1240
from bzrlib.osutils import backup_file
1242
inv = self.read_working_inventory()
1243
if old_tree is None:
1244
old_tree = self.basis_tree()
1245
old_inv = old_tree.inventory
1248
for fn in filenames:
1249
file_id = inv.path2id(fn)
1251
raise NotVersionedError("not a versioned file", fn)
1252
if not old_inv.has_id(file_id):
1253
raise BzrError("file not present in old tree", fn, file_id)
1254
nids.append((fn, file_id))
1256
# TODO: Rename back if it was previously at a different location
1258
# TODO: If given a directory, restore the entire contents from
1259
# the previous version.
1261
# TODO: Make a backup to a temporary file.
1263
# TODO: If the file previously didn't exist, delete it?
1264
for fn, file_id in nids:
1267
f = AtomicFile(fn, 'wb')
1269
f.write(old_tree.get_file(file_id).read())
1275
def pending_merges(self):
1276
"""Return a list of pending merges.
1278
These are revisions that have been merged into the working
1279
directory but not yet committed.
1281
cfn = self.controlfilename('pending-merges')
1282
if not os.path.exists(cfn):
1285
for l in self.controlfile('pending-merges', 'r').readlines():
1286
p.append(l.rstrip('\n'))
1290
def add_pending_merge(self, revision_id):
1291
from bzrlib.revision import validate_revision_id
1293
validate_revision_id(revision_id)
1295
p = self.pending_merges()
1296
if revision_id in p:
1298
p.append(revision_id)
1299
self.set_pending_merges(p)
1302
def set_pending_merges(self, rev_list):
1303
from bzrlib.atomicfile import AtomicFile
1306
f = AtomicFile(self.controlfilename('pending-merges'))
1317
def get_parent(self):
1318
"""Return the parent location of the branch.
1320
This is the default location for push/pull/missing. The usual
1321
pattern is that the user can override it by specifying a
1325
_locs = ['parent', 'pull', 'x-pull']
1328
return self.controlfile(l, 'r').read().strip('\n')
1330
if e.errno != errno.ENOENT:
1335
def set_parent(self, url):
1336
# TODO: Maybe delete old location files?
1337
from bzrlib.atomicfile import AtomicFile
1340
f = AtomicFile(self.controlfilename('parent'))
1349
def check_revno(self, revno):
1351
Check whether a revno corresponds to any revision.
1352
Zero (the NULL revision) is considered valid.
1355
self.check_real_revno(revno)
1357
def check_real_revno(self, revno):
1359
Check whether a revno corresponds to a real revision.
1360
Zero (the NULL revision) is considered invalid
1362
if revno < 1 or revno > self.revno():
1363
raise InvalidRevisionNumber(revno)
615
def write_log(self, utc=False):
616
"""Write out human-readable log of commits to this branch
618
:param utc: If true, show dates in universal time, not local time."""
619
## TODO: Option to choose either original, utc or local timezone
622
for p in self.revision_history():
624
print 'revno:', revno
625
## TODO: Show hash if --id is given.
626
##print 'revision-hash:', p
627
rev = self.get_revision(p)
628
print 'committer:', rev.committer
629
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0))
631
## opportunistic consistency check, same as check_patch_chaining
632
if rev.precursor != precursor:
633
bailout("mismatched precursor!")
637
print ' (no message)'
639
for l in rev.message.split('\n'):
647
def show_status(branch, show_all=False):
648
"""Display single-line status for non-ignored working files.
650
The list is show sorted in order by file name.
652
>>> b = ScratchBranch(files=['foo', 'foo~'])
658
>>> b.commit("add foo")
661
:todo: Get state for single files.
663
:todo: Perhaps show a slash at the end of directory names.
667
# We have to build everything into a list first so that it can
668
# sorted by name, incorporating all the different sources.
670
# FIXME: Rather than getting things in random order and then sorting,
671
# just step through in order.
673
# Interesting case: the old ID for a file has been removed,
674
# but a new file has been created under that name.
676
old = branch.basis_tree()
677
old_inv = old.inventory
678
new = branch.working_tree()
679
new_inv = new.inventory
681
for fs, fid, oldname, newname, kind in diff_trees(old, new):
683
show_status(fs, kind,
684
oldname + ' => ' + newname)
685
elif fs == 'A' or fs == 'M':
686
show_status(fs, kind, newname)
688
show_status(fs, kind, oldname)
691
show_status(fs, kind, newname)
694
show_status(fs, kind, newname)
696
show_status(fs, kind, newname)
698
bailout("wierd file state %r" % ((fs, fid),))
1368
702
class ScratchBranch(Branch):
1448
740
## mutter('check %r for control file' % ((head, tail), ))
1449
741
if tail == bzrlib.BZRDIR:
1451
if filename == head:
1458
def gen_file_id(name):
748
def _gen_revision_id(when):
749
"""Return new revision-id."""
750
s = '%s-%s-' % (user_email(), compact_date(when))
751
s += hexlify(rand_bytes(8))
755
def _gen_file_id(name):
1459
756
"""Return new file id.
1461
758
This should probably generate proper UUIDs, but for the moment we
1462
759
cope with just randomness because running uuidgen every time is
1465
from binascii import hexlify
1466
from time import time
1468
# get last component
1469
idx = name.rfind('/')
1471
name = name[idx+1 : ]
1472
idx = name.rfind('\\')
1474
name = name[idx+1 : ]
1476
# make it not a hidden file
1477
name = name.lstrip('.')
1479
# remove any wierd characters; we don't escape them but rather
1480
# just pull them out
1481
name = re.sub(r'[^\w.]', '', name)
761
assert '/' not in name
762
while name[0] == '.':
1483
764
s = hexlify(rand_bytes(8))
1484
return '-'.join((name, compact_date(time()), s))
1488
"""Return a new tree-root file id."""
1489
return gen_file_id('TREE_ROOT')
1492
def pull_loc(branch):
1493
# TODO: Should perhaps just make attribute be 'base' in
1494
# RemoteBranch and Branch?
1495
if hasattr(branch, "baseurl"):
1496
return branch.baseurl
1501
def copy_branch(branch_from, to_location, revision=None):
1502
"""Copy branch_from into the existing directory to_location.
1505
If not None, only revisions up to this point will be copied.
1506
The head of the new branch will be that revision.
1509
The name of a local directory that exists but is empty.
1511
from bzrlib.merge import merge
1512
from bzrlib.branch import Branch
1514
assert isinstance(branch_from, Branch)
1515
assert isinstance(to_location, basestring)
1517
br_to = Branch(to_location, init=True)
1518
br_to.set_root_id(branch_from.get_root_id())
1519
if revision is None:
1520
revno = branch_from.revno()
1522
revno, rev_id = branch_from.get_revision_info(revision)
1523
br_to.update_revisions(branch_from, stop_revision=revno)
1524
merge((to_location, -1), (to_location, 0), this_dir=to_location,
1525
check_clean=False, ignore_zero=True)
1527
from_location = pull_loc(branch_from)
1528
br_to.set_parent(pull_loc(branch_from))
765
return '-'.join((name, compact_date(time.time()), s))