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
29
from bzrlib.textui import show_status
30
from bzrlib.revision import Revision
31
from bzrlib.xml import unpack_xml
32
from bzrlib.delta import compare_trees
33
from bzrlib.tree import EmptyTree, RevisionTree
38
38
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
39
39
## TODO: Maybe include checks for common corruption of newlines, etc?
42
# TODO: Some operations like log might retrieve the same revisions
43
# repeatedly to calculate deltas. We could perhaps have a weakref
44
# cache in memory to make this faster.
46
# TODO: please move the revision-string syntax stuff out of the branch
47
# object; it's clutter
50
def find_branch(f, **args):
51
if f and (f.startswith('http://') or f.startswith('https://')):
53
return remotebranch.RemoteBranch(f, **args)
55
return Branch(f, **args)
58
def find_cached_branch(f, cache_root, **args):
59
from remotebranch import RemoteBranch
60
br = find_branch(f, **args)
61
def cacheify(br, store_name):
62
from meta_store import CachedStore
63
cache_path = os.path.join(cache_root, store_name)
65
new_store = CachedStore(getattr(br, store_name), cache_path)
66
setattr(br, store_name, new_store)
68
if isinstance(br, RemoteBranch):
69
cacheify(br, 'inventory_store')
70
cacheify(br, 'text_store')
71
cacheify(br, 'revision_store')
75
def _relpath(base, path):
76
"""Return path relative to base, or raise exception.
78
The path may be either an absolute path or a path relative to the
79
current working directory.
81
Lifted out of Branch.relpath for ease of testing.
83
os.path.commonprefix (python2.4) has a bad bug that it works just
84
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
85
avoids that problem."""
86
rp = os.path.abspath(path)
90
while len(head) >= len(base):
93
head, tail = os.path.split(head)
97
from errors import NotBranchError
98
raise NotBranchError("path %r is not within branch %r" % (rp, base))
100
return os.sep.join(s)
43
103
def find_branch_root(f=None):
44
104
"""Find the branch root enclosing f, or pwd.
106
f may be a filename or a URL.
46
108
It is not necessary that f exists.
48
110
Basically we keep looking up until we find the control directory or
111
run into the root. If there isn't one, raises NotBranchError.
52
115
elif hasattr(os.path, 'realpath'):
53
116
f = os.path.realpath(f)
55
118
f = os.path.abspath(f)
119
if not os.path.exists(f):
120
raise BzrError('%r does not exist' % f)
61
126
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
63
128
head, tail = os.path.split(f)
65
130
# reached the root, whatever that may be
66
bailout('%r is not in a branch' % orig_f)
131
raise bzrlib.errors.NotBranchError('%s is not in a branch' % orig_f)
136
# XXX: move into bzrlib.errors; subclass BzrError
137
class DivergedBranches(Exception):
138
def __init__(self, branch1, branch2):
139
self.branch1 = branch1
140
self.branch2 = branch2
141
Exception.__init__(self, "These branches have diverged.")
71
144
######################################################################
147
class Branch(object):
75
148
"""Branch holding a history of revisions.
77
:todo: Perhaps use different stores for different classes of object,
78
so that we can keep track of how much space each one uses,
79
or garbage-collect them.
81
:todo: Add a RemoteBranch subclass. For the basic case of read-only
82
HTTP access this should be very easy by,
83
just redirecting controlfile access into HTTP requests.
84
We would need a RemoteStore working similarly.
86
:todo: Keep the on-disk branch locked while the object exists.
88
:todo: mkdir() method.
151
Base directory of the branch.
157
If _lock_mode is true, a positive count of the number of times the
161
Lock object from bzrlib.lock.
168
# Map some sort of prefix into a namespace
169
# stuff like "revno:10", "revid:", etc.
170
# This should match a prefix with a function which accepts
171
REVISION_NAMESPACES = {}
90
173
def __init__(self, base, init=False, find_root=True):
91
174
"""Create new branch object at a particular location.
93
:param base: Base directory for the branch.
176
base -- Base directory for the branch.
95
:param init: If True, create new control files in a previously
178
init -- If True, create new control files in a previously
96
179
unversioned directory. If False, the branch must already
99
:param find_root: If true and init is false, find the root of the
182
find_root -- If true and init is false, find the root of the
100
183
existing branch containing base.
102
185
In the test suite, creation of new trees is tested using the
103
186
`ScratchBranch` class.
188
from bzrlib.store import ImmutableStore
106
190
self.base = os.path.realpath(base)
107
191
self._make_control()
127
212
__repr__ = __str__
216
if self._lock_mode or self._lock:
217
from warnings import warn
218
warn("branch %r was not explicitly unlocked" % self)
222
def lock_write(self):
224
if self._lock_mode != 'w':
225
from errors import LockError
226
raise LockError("can't upgrade to a write lock from %r" %
228
self._lock_count += 1
230
from bzrlib.lock import WriteLock
232
self._lock = WriteLock(self.controlfilename('branch-lock'))
233
self._lock_mode = 'w'
239
assert self._lock_mode in ('r', 'w'), \
240
"invalid lock mode %r" % self._lock_mode
241
self._lock_count += 1
243
from bzrlib.lock import ReadLock
245
self._lock = ReadLock(self.controlfilename('branch-lock'))
246
self._lock_mode = 'r'
250
if not self._lock_mode:
251
from errors import LockError
252
raise LockError('branch %r is not locked' % (self))
254
if self._lock_count > 1:
255
self._lock_count -= 1
259
self._lock_mode = self._lock_count = None
130
261
def abspath(self, name):
131
262
"""Return absolute filename for something in the branch"""
132
263
return os.path.join(self.base, name)
135
265
def relpath(self, path):
136
266
"""Return path relative to this branch of something inside it.
138
268
Raises an error if path is not in this branch."""
139
rp = os.path.realpath(path)
141
if not rp.startswith(self.base):
142
bailout("path %r is not within branch %r" % (rp, self.base))
143
rp = rp[len(self.base):]
144
rp = rp.lstrip(os.sep)
269
return _relpath(self.base, path)
148
271
def controlfilename(self, file_or_path):
149
272
"""Return location relative to branch."""
150
if isinstance(file_or_path, types.StringTypes):
273
if isinstance(file_or_path, basestring):
151
274
file_or_path = [file_or_path]
152
275
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
155
278
def controlfile(self, file_or_path, mode='r'):
156
"""Open a control file for this branch"""
157
return file(self.controlfilename(file_or_path), mode)
279
"""Open a control file for this branch.
281
There are two classes of file in the control directory: text
282
and binary. binary files are untranslated byte streams. Text
283
control files are stored with Unix newlines and in UTF-8, even
284
if the platform or locale defaults are different.
286
Controlfiles should almost never be opened in write mode but
287
rather should be atomically copied and replaced using atomicfile.
290
fn = self.controlfilename(file_or_path)
292
if mode == 'rb' or mode == 'wb':
293
return file(fn, mode)
294
elif mode == 'r' or mode == 'w':
295
# open in binary mode anyhow so there's no newline translation;
296
# codecs uses line buffering by default; don't want that.
298
return codecs.open(fn, mode + 'b', 'utf-8',
301
raise BzrError("invalid controlfile mode %r" % mode)
160
303
def _make_control(self):
304
from bzrlib.inventory import Inventory
305
from bzrlib.xml import pack_xml
161
307
os.mkdir(self.controlfilename([]))
162
308
self.controlfile('README', 'w').write(
163
309
"This is a Bazaar-NG control directory.\n"
164
"Do not change any files in this directory.")
310
"Do not change any files in this directory.\n")
165
311
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
166
312
for d in ('text-store', 'inventory-store', 'revision-store'):
167
313
os.mkdir(self.controlfilename(d))
168
314
for f in ('revision-history', 'merged-patches',
169
'pending-merged-patches', 'branch-name'):
315
'pending-merged-patches', 'branch-name',
170
318
self.controlfile(f, 'w').write('')
171
319
mutter('created control directory in ' + self.base)
172
Inventory().write_xml(self.controlfile('inventory','w'))
321
# if we want per-tree root ids then this is the place to set
322
# them; they're not needed for now and so ommitted for
324
pack_xml(Inventory(), self.controlfile('inventory','w'))
175
326
def _check_format(self):
176
327
"""Check this branch format is supported.
180
331
In the future, we might need different in-memory Branch
181
332
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()
334
# This ignores newlines so that we can open branches created
335
# on Windows from Linux and so on. I think it might be better
336
# to always make all internal files in unix format.
337
fmt = self.controlfile('branch-format', 'r').read()
338
fmt.replace('\r\n', '')
185
339
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'])
340
raise BzrError('sorry, branch format %r not supported' % fmt,
341
['use a different bzr version',
342
'or remove the .bzr directory and "bzr init" again'])
344
def get_root_id(self):
345
"""Return the id of this branches root"""
346
inv = self.read_working_inventory()
347
return inv.root.file_id
349
def set_root_id(self, file_id):
350
inv = self.read_working_inventory()
351
orig_root_id = inv.root.file_id
352
del inv._byid[inv.root.file_id]
353
inv.root.file_id = file_id
354
inv._byid[inv.root.file_id] = inv.root
357
if entry.parent_id in (None, orig_root_id):
358
entry.parent_id = inv.root.file_id
359
self._write_inventory(inv)
191
361
def read_working_inventory(self):
192
362
"""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))
363
from bzrlib.inventory import Inventory
364
from bzrlib.xml import unpack_xml
365
from time import time
369
# ElementTree does its own conversion from UTF-8, so open in
371
inv = unpack_xml(Inventory,
372
self.controlfile('inventory', 'rb'))
373
mutter("loaded inventory of %d items in %f"
374
% (len(inv), time() - before))
200
380
def _write_inventory(self, inv):
201
381
"""Update the working inventory.
203
383
That is to say, the inventory describing changes underway, that
204
384
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'))
386
from bzrlib.atomicfile import AtomicFile
387
from bzrlib.xml import pack_xml
391
f = AtomicFile(self.controlfilename('inventory'), 'wb')
213
400
mutter('wrote working inventory')
216
403
inventory = property(read_working_inventory, _write_inventory, None,
217
404
"""Inventory for the working copy.""")
220
def add(self, files, verbose=False):
407
def add(self, files, ids=None):
221
408
"""Make files versioned.
410
Note that the command line normally calls smart_add instead,
411
which can automatically recurse.
223
413
This puts the files in the Added state, so that they will be
224
414
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', [])
417
List of paths to add, relative to the base of the tree.
420
If set, use these instead of automatically generated ids.
421
Must be the same length as the list of files, but may
422
contain None for ids that are to be autogenerated.
424
TODO: Perhaps have an option to add the ids even if the files do
427
TODO: Perhaps yield the ids and paths as they're added.
261
429
# TODO: Re-adding a file that is removed in the working copy
262
430
# should probably put it back with the previous ID.
263
if isinstance(files, types.StringTypes):
431
if isinstance(files, basestring):
432
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)
438
ids = [None] * len(files)
440
assert(len(ids) == len(files))
444
inv = self.read_working_inventory()
445
for f,file_id in zip(files, ids):
446
if is_control_file(f):
447
raise BzrError("cannot add control file %s" % quotefn(f))
452
raise BzrError("cannot add top-level %r" % f)
454
fullpath = os.path.normpath(self.abspath(f))
457
kind = file_kind(fullpath)
459
# maybe something better?
460
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
462
if kind != 'file' and kind != 'directory':
463
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
466
file_id = gen_file_id(f)
467
inv.add_path(f, kind=kind, file_id=file_id)
469
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
471
self._write_inventory(inv)
476
def print_file(self, file, revno):
477
"""Print `file` to stdout."""
480
tree = self.revision_tree(self.lookup_revision(revno))
481
# use inventory as it was in that revision
482
file_id = tree.inventory.path2id(file)
484
raise BzrError("%r is not present in revision %s" % (file, revno))
485
tree.print_file(file_id)
299
490
def remove(self, files, verbose=False):
302
493
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
495
TODO: Refuse to remove modified files unless --force is given?
497
TODO: Do something useful with directories.
499
TODO: Should this remove the text or not? Tough call; not
331
500
removing may be useful and the user can just use use rm, and
332
501
is the opposite of add. Removing it is consistent with most
333
502
other tools. Maybe an option.
335
504
## TODO: Normalize names
336
505
## TODO: Remove nested loops; better scalability
338
if isinstance(files, types.StringTypes):
506
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))
512
tree = self.working_tree()
515
# do this before any modifications
519
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
520
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
522
# having remove it, it must be either ignored or unknown
523
if tree.is_ignored(f):
527
show_status(new_status, inv[fid].kind, quotefn(f))
530
self._write_inventory(inv)
535
# FIXME: this doesn't need to be a branch method
536
def set_inventory(self, new_inventory_list):
537
from bzrlib.inventory import Inventory, InventoryEntry
538
inv = Inventory(self.get_root_id())
539
for path, file_id, parent, kind in new_inventory_list:
540
name = os.path.basename(path)
543
inv.add(InventoryEntry(file_id, name, kind, parent))
359
544
self._write_inventory(inv)
378
563
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')
566
def append_revision(self, *revision_ids):
567
from bzrlib.atomicfile import AtomicFile
569
for revision_id in revision_ids:
570
mutter("add {%s} to revision-history" % revision_id)
572
rev_history = self.revision_history()
573
rev_history.extend(revision_ids)
575
f = AtomicFile(self.controlfilename('revision-history'))
577
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)
550
599
def get_revision(self, revision_id):
551
600
"""Return the Revision object for a named revision"""
552
r = Revision.read_xml(self.revision_store[revision_id])
601
xml_file = self.get_revision_xml(revision_id)
604
r = unpack_xml(Revision, xml_file)
605
except SyntaxError, e:
606
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
553
610
assert r.revision_id == revision_id
614
def get_revision_delta(self, revno):
615
"""Return the delta for one revision.
617
The delta is relative to its mainline predecessor, or the
618
empty tree for revision 1.
620
assert isinstance(revno, int)
621
rh = self.revision_history()
622
if not (1 <= revno <= len(rh)):
623
raise InvalidRevisionNumber(revno)
625
# revno is 1-based; list is 0-based
627
new_tree = self.revision_tree(rh[revno-1])
629
old_tree = EmptyTree()
631
old_tree = self.revision_tree(rh[revno-2])
633
return compare_trees(old_tree, new_tree)
637
def get_revision_sha1(self, revision_id):
638
"""Hash the stored value of a revision, and return it."""
639
# In the future, revision entries will be signed. At that
640
# point, it is probably best *not* to include the signature
641
# in the revision hash. Because that lets you re-sign
642
# the revision, (add signatures/remove signatures) and still
643
# have all hash pointers stay consistent.
644
# But for now, just hash the contents.
645
return bzrlib.osutils.sha_file(self.get_revision_xml(revision_id))
557
648
def get_inventory(self, inventory_id):
558
649
"""Get Inventory object by hash.
560
:todo: Perhaps for this and similar methods, take a revision
651
TODO: Perhaps for this and similar methods, take a revision
561
652
parameter which can be either an integer revno or a
563
i = Inventory.read_xml(self.inventory_store[inventory_id])
654
from bzrlib.inventory import Inventory
655
from bzrlib.xml import unpack_xml
657
return unpack_xml(Inventory, self.get_inventory_xml(inventory_id))
660
def get_inventory_xml(self, inventory_id):
661
"""Get inventory XML as a file object."""
662
return self.inventory_store[inventory_id]
665
def get_inventory_sha1(self, inventory_id):
666
"""Return the sha1 hash of the inventory entry
668
return sha_file(self.get_inventory_xml(inventory_id))
567
671
def get_revision_inventory(self, revision_id):
568
672
"""Return inventory of a past revision."""
673
# bzr 0.0.6 imposes the constraint that the inventory_id
674
# must be the same as its revision, so this is trivial.
569
675
if revision_id == None:
676
from bzrlib.inventory import Inventory
677
return Inventory(self.get_root_id())
572
return self.get_inventory(self.get_revision(revision_id).inventory_id)
679
return self.get_inventory(revision_id)
575
682
def revision_history(self):
587
744
That is equivalent to the number of revisions committed to
590
>>> b = ScratchBranch()
593
>>> b.commit('no foo')
597
747
return len(self.revision_history())
600
750
def last_patch(self):
601
751
"""Return last patch hash, or None if no history.
603
>>> ScratchBranch().last_patch() == None
606
753
ph = self.revision_history()
611
def lookup_revision(self, revno):
612
"""Return revision hash for revision number."""
617
# list is 0-based; revisions are 1-based
618
return self.revision_history()[revno-1]
620
bailout("no such revision %s" % revno)
760
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
762
If self and other have not diverged, return a list of the revisions
763
present in other, but missing from self.
765
>>> from bzrlib.commit import commit
766
>>> bzrlib.trace.silent = True
767
>>> br1 = ScratchBranch()
768
>>> br2 = ScratchBranch()
769
>>> br1.missing_revisions(br2)
771
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
772
>>> br1.missing_revisions(br2)
774
>>> br2.missing_revisions(br1)
776
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
777
>>> br1.missing_revisions(br2)
779
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
780
>>> br1.missing_revisions(br2)
782
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
783
>>> br1.missing_revisions(br2)
784
Traceback (most recent call last):
785
DivergedBranches: These branches have diverged.
787
self_history = self.revision_history()
788
self_len = len(self_history)
789
other_history = other.revision_history()
790
other_len = len(other_history)
791
common_index = min(self_len, other_len) -1
792
if common_index >= 0 and \
793
self_history[common_index] != other_history[common_index]:
794
raise DivergedBranches(self, other)
796
if stop_revision is None:
797
stop_revision = other_len
798
elif stop_revision > other_len:
799
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
801
return other_history[self_len:stop_revision]
804
def update_revisions(self, other, stop_revision=None):
805
"""Pull in all new revisions from other branch.
807
from bzrlib.fetch import greedy_fetch
809
pb = bzrlib.ui.ui_factory.progress_bar()
810
pb.update('comparing histories')
812
revision_ids = self.missing_revisions(other, stop_revision)
814
if len(revision_ids) > 0:
815
count = greedy_fetch(self, other, revision_ids[-1], pb)[0]
818
self.append_revision(*revision_ids)
819
## note("Added %d revisions." % count)
822
def install_revisions(self, other, revision_ids, pb):
823
if hasattr(other.revision_store, "prefetch"):
824
other.revision_store.prefetch(revision_ids)
825
if hasattr(other.inventory_store, "prefetch"):
826
inventory_ids = [other.get_revision(r).inventory_id
827
for r in revision_ids]
828
other.inventory_store.prefetch(inventory_ids)
831
pb = bzrlib.ui.ui_factory.progress_bar()
838
for i, rev_id in enumerate(revision_ids):
839
pb.update('fetching revision', i+1, len(revision_ids))
841
rev = other.get_revision(rev_id)
842
except bzrlib.errors.NoSuchRevision:
846
revisions.append(rev)
847
inv = other.get_inventory(str(rev.inventory_id))
848
for key, entry in inv.iter_entries():
849
if entry.text_id is None:
851
if entry.text_id not in self.text_store:
852
needed_texts.add(entry.text_id)
856
count, cp_fail = self.text_store.copy_multi(other.text_store,
858
#print "Added %d texts." % count
859
inventory_ids = [ f.inventory_id for f in revisions ]
860
count, cp_fail = self.inventory_store.copy_multi(other.inventory_store,
862
#print "Added %d inventories." % count
863
revision_ids = [ f.revision_id for f in revisions]
865
count, cp_fail = self.revision_store.copy_multi(other.revision_store,
868
assert len(cp_fail) == 0
869
return count, failures
872
def commit(self, *args, **kw):
873
from bzrlib.commit import commit
874
commit(self, *args, **kw)
877
def lookup_revision(self, revision):
878
"""Return the revision identifier for a given revision information."""
879
revno, info = self.get_revision_info(revision)
883
def revision_id_to_revno(self, revision_id):
884
"""Given a revision id, return its revno"""
885
history = self.revision_history()
887
return history.index(revision_id) + 1
889
raise bzrlib.errors.NoSuchRevision(self, revision_id)
892
def get_revision_info(self, revision):
893
"""Return (revno, revision id) for revision identifier.
895
revision can be an integer, in which case it is assumed to be revno (though
896
this will translate negative values into positive ones)
897
revision can also be a string, in which case it is parsed for something like
898
'date:' or 'revid:' etc.
903
try:# Convert to int if possible
904
revision = int(revision)
907
revs = self.revision_history()
908
if isinstance(revision, int):
911
# Mabye we should do this first, but we don't need it if revision == 0
913
revno = len(revs) + revision + 1
916
elif isinstance(revision, basestring):
917
for prefix, func in Branch.REVISION_NAMESPACES.iteritems():
918
if revision.startswith(prefix):
919
revno = func(self, revs, revision)
922
raise BzrError('No namespace registered for string: %r' % revision)
924
if revno is None or revno <= 0 or revno > len(revs):
925
raise BzrError("no such revision %s" % revision)
926
return revno, revs[revno-1]
928
def _namespace_revno(self, revs, revision):
929
"""Lookup a revision by revision number"""
930
assert revision.startswith('revno:')
932
return int(revision[6:])
935
REVISION_NAMESPACES['revno:'] = _namespace_revno
937
def _namespace_revid(self, revs, revision):
938
assert revision.startswith('revid:')
940
return revs.index(revision[6:]) + 1
943
REVISION_NAMESPACES['revid:'] = _namespace_revid
945
def _namespace_last(self, revs, revision):
946
assert revision.startswith('last:')
948
offset = int(revision[5:])
953
raise BzrError('You must supply a positive value for --revision last:XXX')
954
return len(revs) - offset + 1
955
REVISION_NAMESPACES['last:'] = _namespace_last
957
def _namespace_tag(self, revs, revision):
958
assert revision.startswith('tag:')
959
raise BzrError('tag: namespace registered, but not implemented.')
960
REVISION_NAMESPACES['tag:'] = _namespace_tag
962
def _namespace_date(self, revs, revision):
963
assert revision.startswith('date:')
965
# Spec for date revisions:
967
# value can be 'yesterday', 'today', 'tomorrow' or a YYYY-MM-DD string.
968
# it can also start with a '+/-/='. '+' says match the first
969
# entry after the given date. '-' is match the first entry before the date
970
# '=' is match the first entry after, but still on the given date.
972
# +2005-05-12 says find the first matching entry after May 12th, 2005 at 0:00
973
# -2005-05-12 says find the first matching entry before May 12th, 2005 at 0:00
974
# =2005-05-12 says find the first match after May 12th, 2005 at 0:00 but before
975
# May 13th, 2005 at 0:00
977
# So the proper way of saying 'give me all entries for today' is:
978
# -r {date:+today}:{date:-tomorrow}
979
# The default is '=' when not supplied
982
if val[:1] in ('+', '-', '='):
983
match_style = val[:1]
986
today = datetime.datetime.today().replace(hour=0,minute=0,second=0,microsecond=0)
987
if val.lower() == 'yesterday':
988
dt = today - datetime.timedelta(days=1)
989
elif val.lower() == 'today':
991
elif val.lower() == 'tomorrow':
992
dt = today + datetime.timedelta(days=1)
995
# This should be done outside the function to avoid recompiling it.
996
_date_re = re.compile(
997
r'(?P<date>(?P<year>\d\d\d\d)-(?P<month>\d\d)-(?P<day>\d\d))?'
999
r'(?P<time>(?P<hour>\d\d):(?P<minute>\d\d)(:(?P<second>\d\d))?)?'
1001
m = _date_re.match(val)
1002
if not m or (not m.group('date') and not m.group('time')):
1003
raise BzrError('Invalid revision date %r' % revision)
1006
year, month, day = int(m.group('year')), int(m.group('month')), int(m.group('day'))
1008
year, month, day = today.year, today.month, today.day
1010
hour = int(m.group('hour'))
1011
minute = int(m.group('minute'))
1012
if m.group('second'):
1013
second = int(m.group('second'))
1017
hour, minute, second = 0,0,0
1019
dt = datetime.datetime(year=year, month=month, day=day,
1020
hour=hour, minute=minute, second=second)
1024
if match_style == '-':
1026
elif match_style == '=':
1027
last = dt + datetime.timedelta(days=1)
1030
for i in range(len(revs)-1, -1, -1):
1031
r = self.get_revision(revs[i])
1032
# TODO: Handle timezone.
1033
dt = datetime.datetime.fromtimestamp(r.timestamp)
1034
if first >= dt and (last is None or dt >= last):
1037
for i in range(len(revs)):
1038
r = self.get_revision(revs[i])
1039
# TODO: Handle timezone.
1040
dt = datetime.datetime.fromtimestamp(r.timestamp)
1041
if first <= dt and (last is None or dt <= last):
1043
REVISION_NAMESPACES['date:'] = _namespace_date
623
1045
def revision_tree(self, revision_id):
624
1046
"""Return Tree for a revision on this branch.
626
1048
`revision_id` may be None for the null revision, in which case
627
1049
an `EmptyTree` is returned."""
1050
# TODO: refactor this to use an existing revision object
1051
# so we don't need to read it in twice.
629
1052
if revision_id == None:
630
1053
return EmptyTree()
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),))
1078
def rename_one(self, from_rel, to_rel):
1081
This can change the directory or the filename or both.
1085
tree = self.working_tree()
1086
inv = tree.inventory
1087
if not tree.has_filename(from_rel):
1088
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
1089
if tree.has_filename(to_rel):
1090
raise BzrError("can't rename: new working file %r already exists" % to_rel)
1092
file_id = inv.path2id(from_rel)
1094
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
1096
if inv.path2id(to_rel):
1097
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
1099
to_dir, to_tail = os.path.split(to_rel)
1100
to_dir_id = inv.path2id(to_dir)
1101
if to_dir_id == None and to_dir != '':
1102
raise BzrError("can't determine destination directory id for %r" % to_dir)
1104
mutter("rename_one:")
1105
mutter(" file_id {%s}" % file_id)
1106
mutter(" from_rel %r" % from_rel)
1107
mutter(" to_rel %r" % to_rel)
1108
mutter(" to_dir %r" % to_dir)
1109
mutter(" to_dir_id {%s}" % to_dir_id)
1111
inv.rename(file_id, to_dir_id, to_tail)
1113
from_abs = self.abspath(from_rel)
1114
to_abs = self.abspath(to_rel)
1116
os.rename(from_abs, to_abs)
1118
raise BzrError("failed to rename %r to %r: %s"
1119
% (from_abs, to_abs, e[1]),
1120
["rename rolled back"])
1122
self._write_inventory(inv)
1127
def move(self, from_paths, to_name):
1130
to_name must exist as a versioned directory.
1132
If to_name exists and is a directory, the files are moved into
1133
it, keeping their old names. If it is a directory,
1135
Note that to_name is only the last component of the new name;
1136
this doesn't change the directory.
1138
This returns a list of (from_path, to_path) pairs for each
1139
entry that is moved.
1144
## TODO: Option to move IDs only
1145
assert not isinstance(from_paths, basestring)
1146
tree = self.working_tree()
1147
inv = tree.inventory
1148
to_abs = self.abspath(to_name)
1149
if not isdir(to_abs):
1150
raise BzrError("destination %r is not a directory" % to_abs)
1151
if not tree.has_filename(to_name):
1152
raise BzrError("destination %r not in working directory" % to_abs)
1153
to_dir_id = inv.path2id(to_name)
1154
if to_dir_id == None and to_name != '':
1155
raise BzrError("destination %r is not a versioned directory" % to_name)
1156
to_dir_ie = inv[to_dir_id]
1157
if to_dir_ie.kind not in ('directory', 'root_directory'):
1158
raise BzrError("destination %r is not a directory" % to_abs)
1160
to_idpath = inv.get_idpath(to_dir_id)
1162
for f in from_paths:
1163
if not tree.has_filename(f):
1164
raise BzrError("%r does not exist in working tree" % f)
1165
f_id = inv.path2id(f)
1167
raise BzrError("%r is not versioned" % f)
1168
name_tail = splitpath(f)[-1]
1169
dest_path = appendpath(to_name, name_tail)
1170
if tree.has_filename(dest_path):
1171
raise BzrError("destination %r already exists" % dest_path)
1172
if f_id in to_idpath:
1173
raise BzrError("can't move %r to a subdirectory of itself" % f)
1175
# OK, so there's a race here, it's possible that someone will
1176
# create a file in this interval and then the rename might be
1177
# left half-done. But we should have caught most problems.
1179
for f in from_paths:
1180
name_tail = splitpath(f)[-1]
1181
dest_path = appendpath(to_name, name_tail)
1182
result.append((f, dest_path))
1183
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1185
os.rename(self.abspath(f), self.abspath(dest_path))
1187
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1188
["rename rolled back"])
1190
self._write_inventory(inv)
1197
def revert(self, filenames, old_tree=None, backups=True):
1198
"""Restore selected files to the versions from a previous tree.
1201
If true (default) backups are made of files before
1204
from bzrlib.errors import NotVersionedError, BzrError
1205
from bzrlib.atomicfile import AtomicFile
1206
from bzrlib.osutils import backup_file
1208
inv = self.read_working_inventory()
1209
if old_tree is None:
1210
old_tree = self.basis_tree()
1211
old_inv = old_tree.inventory
1214
for fn in filenames:
1215
file_id = inv.path2id(fn)
1217
raise NotVersionedError("not a versioned file", fn)
1218
if not old_inv.has_id(file_id):
1219
raise BzrError("file not present in old tree", fn, file_id)
1220
nids.append((fn, file_id))
1222
# TODO: Rename back if it was previously at a different location
1224
# TODO: If given a directory, restore the entire contents from
1225
# the previous version.
1227
# TODO: Make a backup to a temporary file.
1229
# TODO: If the file previously didn't exist, delete it?
1230
for fn, file_id in nids:
1233
f = AtomicFile(fn, 'wb')
1235
f.write(old_tree.get_file(file_id).read())
1241
def pending_merges(self):
1242
"""Return a list of pending merges.
1244
These are revisions that have been merged into the working
1245
directory but not yet committed.
1247
cfn = self.controlfilename('pending-merges')
1248
if not os.path.exists(cfn):
1251
for l in self.controlfile('pending-merges', 'r').readlines():
1252
p.append(l.rstrip('\n'))
1256
def add_pending_merge(self, revision_id):
1257
from bzrlib.revision import validate_revision_id
1259
validate_revision_id(revision_id)
1261
p = self.pending_merges()
1262
if revision_id in p:
1264
p.append(revision_id)
1265
self.set_pending_merges(p)
1268
def set_pending_merges(self, rev_list):
1269
from bzrlib.atomicfile import AtomicFile
1272
f = AtomicFile(self.controlfilename('pending-merges'))
1283
def get_parent(self):
1284
"""Return the parent location of the branch.
1286
This is the default location for push/pull/missing. The usual
1287
pattern is that the user can override it by specifying a
1291
_locs = ['parent', 'pull', 'x-pull']
1294
return self.controlfile(l, 'r').read().strip('\n')
1296
if e.errno != errno.ENOENT:
1301
def set_parent(self, url):
1302
# TODO: Maybe delete old location files?
1303
from bzrlib.atomicfile import AtomicFile
1306
f = AtomicFile(self.controlfilename('parent'))
756
1318
class ScratchBranch(Branch):
804
def _gen_revision_id(when):
805
"""Return new revision-id."""
806
s = '%s-%s-' % (user_email(), compact_date(when))
807
s += hexlify(rand_bytes(8))
811
1408
def gen_file_id(name):
812
1409
"""Return new file id.
814
1411
This should probably generate proper UUIDs, but for the moment we
815
1412
cope with just randomness because running uuidgen every time is
1415
from binascii import hexlify
1416
from time import time
1418
# get last component
817
1419
idx = name.rfind('/')
819
1421
name = name[idx+1 : ]
1422
idx = name.rfind('\\')
1424
name = name[idx+1 : ]
1426
# make it not a hidden file
821
1427
name = name.lstrip('.')
1429
# remove any wierd characters; we don't escape them but rather
1430
# just pull them out
1431
name = re.sub(r'[^\w.]', '', name)
823
1433
s = hexlify(rand_bytes(8))
824
return '-'.join((name, compact_date(time.time()), s))
1434
return '-'.join((name, compact_date(time()), s))
1438
"""Return a new tree-root file id."""
1439
return gen_file_id('TREE_ROOT')
1442
def pull_loc(branch):
1443
# TODO: Should perhaps just make attribute be 'base' in
1444
# RemoteBranch and Branch?
1445
if hasattr(branch, "baseurl"):
1446
return branch.baseurl
1451
def copy_branch(branch_from, to_location, revision=None):
1452
"""Copy branch_from into the existing directory to_location.
1455
If not None, only revisions up to this point will be copied.
1456
The head of the new branch will be that revision.
1459
The name of a local directory that exists but is empty.
1461
from bzrlib.merge import merge
1462
from bzrlib.branch import Branch
1464
assert isinstance(branch_from, Branch)
1465
assert isinstance(to_location, basestring)
1467
br_to = Branch(to_location, init=True)
1468
br_to.set_root_id(branch_from.get_root_id())
1469
if revision is None:
1470
revno = branch_from.revno()
1472
revno, rev_id = branch_from.get_revision_info(revision)
1473
br_to.update_revisions(branch_from, stop_revision=revno)
1474
merge((to_location, -1), (to_location, 0), this_dir=to_location,
1475
check_clean=False, ignore_zero=True)
1477
from_location = pull_loc(branch_from)
1478
br_to.set_parent(pull_loc(branch_from))