15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21
from warnings import warn
20
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
21
import traceback, socket, fnmatch, difflib, time
22
from binascii import hexlify
25
from bzrlib.trace import mutter, note
26
from bzrlib.osutils import (isdir, quotefn, compact_date, rand_bytes,
27
rename, splitpath, sha_file, appendpath,
29
from bzrlib.errors import (BzrError, InvalidRevisionNumber, InvalidRevisionId,
30
NoSuchRevision, HistoryMissing, NotBranchError,
31
DivergedBranches, LockError, UnlistableStore,
33
from bzrlib.textui import show_status
34
from bzrlib.revision import Revision, validate_revision_id, is_ancestor
35
from bzrlib.delta import compare_trees
36
from bzrlib.tree import EmptyTree, RevisionTree
37
from bzrlib.inventory import Inventory
38
from bzrlib.weavestore import WeaveStore
39
from bzrlib.store import copy_all, ImmutableStore
44
BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"
45
BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"
25
from inventory import Inventory
26
from trace import mutter, note
27
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
28
from inventory import InventoryEntry, Inventory
29
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, chomp, \
30
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
31
joinpath, sha_string, file_kind, local_time_offset, appendpath
32
from store import ImmutableStore
33
from revision import Revision
34
from errors import bailout
35
from textui import show_status
36
from diff import diff_trees
38
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
46
39
## TODO: Maybe include checks for common corruption of newlines, etc?
49
# TODO: Some operations like log might retrieve the same revisions
50
# repeatedly to calculate deltas. We could perhaps have a weakref
51
# cache in memory to make this faster. In general anything can be
52
# cached in memory between lock and unlock operations.
54
def find_branch(*ignored, **ignored_too):
55
# XXX: leave this here for about one release, then remove it
56
raise NotImplementedError('find_branch() is not supported anymore, '
57
'please use one of the new branch constructors')
59
def _relpath(base, path):
60
"""Return path relative to base, or raise exception.
62
The path may be either an absolute path or a path relative to the
63
current working directory.
65
Lifted out of Branch.relpath for ease of testing.
67
os.path.commonprefix (python2.4) has a bad bug that it works just
68
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
69
avoids that problem."""
70
rp = os.path.abspath(path)
74
while len(head) >= len(base):
77
head, tail = os.path.split(head)
81
raise NotBranchError("path %r is not within branch %r" % (rp, base))
86
43
def find_branch_root(f=None):
87
44
"""Find the branch root enclosing f, or pwd.
89
f may be a filename or a URL.
91
46
It is not necessary that f exists.
93
48
Basically we keep looking up until we find the control directory or
94
run into the root. If there isn't one, raises NotBranchError.
98
52
elif hasattr(os.path, 'realpath'):
99
53
f = os.path.realpath(f)
101
55
f = os.path.abspath(f)
102
if not os.path.exists(f):
103
raise BzrError('%r does not exist' % f)
109
61
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
111
63
head, tail = os.path.split(f)
113
65
# reached the root, whatever that may be
114
raise NotBranchError('%s is not in a branch' % orig_f)
66
bailout('%r is not in a branch' % orig_f)
120
71
######################################################################
123
class Branch(object):
124
75
"""Branch holding a history of revisions.
127
Base directory/url of the branch.
131
def __init__(self, *ignored, **ignored_too):
132
raise NotImplementedError('The Branch class is abstract')
135
def open_downlevel(base):
136
"""Open a branch which may be of an old format.
138
Only local branches are supported."""
139
return LocalBranch(base, find_root=False, relax_version_check=True)
143
"""Open an existing branch, rooted at 'base' (url)"""
144
if base and (base.startswith('http://') or base.startswith('https://')):
145
from bzrlib.remotebranch import RemoteBranch
146
return RemoteBranch(base, find_root=False)
148
return LocalBranch(base, find_root=False)
151
def open_containing(url):
152
"""Open an existing branch which contains url.
154
This probes for a branch at url, and searches upwards from there.
156
if url and (url.startswith('http://') or url.startswith('https://')):
157
from bzrlib.remotebranch import RemoteBranch
158
return RemoteBranch(url)
160
return LocalBranch(url)
163
def initialize(base):
164
"""Create a new branch, rooted at 'base' (url)"""
165
if base and (base.startswith('http://') or base.startswith('https://')):
166
from bzrlib.remotebranch import RemoteBranch
167
return RemoteBranch(base, init=True)
169
return LocalBranch(base, init=True)
171
def setup_caching(self, cache_root):
172
"""Subclasses that care about caching should override this, and set
173
up cached stores located under cache_root.
177
class LocalBranch(Branch):
178
"""A branch stored in the actual filesystem.
180
Note that it's "local" in the context of the filesystem; it doesn't
181
really matter if it's on an nfs/smb/afs/coda/... share, as long as
182
it's writable, and can be accessed via the normal filesystem API.
188
If _lock_mode is true, a positive count of the number of times the
192
Lock object from bzrlib.lock.
194
# We actually expect this class to be somewhat short-lived; part of its
195
# purpose is to try to isolate what bits of the branch logic are tied to
196
# filesystem access, so that in a later step, we can extricate them to
197
# a separarte ("storage") class.
201
_inventory_weave = None
203
# Map some sort of prefix into a namespace
204
# stuff like "revno:10", "revid:", etc.
205
# This should match a prefix with a function which accepts
206
REVISION_NAMESPACES = {}
208
def push_stores(self, branch_to):
209
"""Copy the content of this branches store to branch_to."""
210
if (self._branch_format != branch_to._branch_format
211
or self._branch_format != 4):
212
from bzrlib.fetch import greedy_fetch
213
mutter("falling back to fetch logic to push between %s(%s) and %s(%s)",
214
self, self._branch_format, branch_to, branch_to._branch_format)
215
greedy_fetch(to_branch=branch_to, from_branch=self,
216
revision=self.last_revision())
219
store_pairs = ((self.text_store, branch_to.text_store),
220
(self.inventory_store, branch_to.inventory_store),
221
(self.revision_store, branch_to.revision_store))
223
for from_store, to_store in store_pairs:
224
copy_all(from_store, to_store)
225
except UnlistableStore:
226
raise UnlistableBranch(from_store)
228
def __init__(self, base, init=False, find_root=True,
229
relax_version_check=False):
77
:todo: Perhaps use different stores for different classes of object,
78
so that we can keep track of how much space each one uses,
79
or garbage-collect them.
81
:todo: Add a RemoteBranch subclass. For the basic case of read-only
82
HTTP access this should be very easy by,
83
just redirecting controlfile access into HTTP requests.
84
We would need a RemoteStore working similarly.
86
:todo: Keep the on-disk branch locked while the object exists.
88
:todo: mkdir() method.
90
def __init__(self, base, init=False, find_root=True):
230
91
"""Create new branch object at a particular location.
232
base -- Base directory for the branch. May be a file:// url.
93
:param base: Base directory for the branch.
234
init -- If True, create new control files in a previously
95
:param init: If True, create new control files in a previously
235
96
unversioned directory. If False, the branch must already
238
find_root -- If true and init is false, find the root of the
99
:param find_root: If true and init is false, find the root of the
239
100
existing branch containing base.
241
relax_version_check -- If true, the usual check for the branch
242
version is not applied. This is intended only for
243
upgrade/recovery type use; it's not guaranteed that
244
all operations will work on old format branches.
246
102
In the test suite, creation of new trees is tested using the
247
103
`ScratchBranch` class.
281
127
__repr__ = __str__
285
if self._lock_mode or self._lock:
286
# XXX: This should show something every time, and be suitable for
287
# headless operation and embedding
288
warn("branch %r was not explicitly unlocked" % self)
291
def lock_write(self):
293
if self._lock_mode != 'w':
294
raise LockError("can't upgrade to a write lock from %r" %
296
self._lock_count += 1
298
from bzrlib.lock import WriteLock
300
self._lock = WriteLock(self.controlfilename('branch-lock'))
301
self._lock_mode = 'w'
307
assert self._lock_mode in ('r', 'w'), \
308
"invalid lock mode %r" % self._lock_mode
309
self._lock_count += 1
311
from bzrlib.lock import ReadLock
313
self._lock = ReadLock(self.controlfilename('branch-lock'))
314
self._lock_mode = 'r'
318
if not self._lock_mode:
319
raise LockError('branch %r is not locked' % (self))
321
if self._lock_count > 1:
322
self._lock_count -= 1
326
self._lock_mode = self._lock_count = None
328
130
def abspath(self, name):
329
131
"""Return absolute filename for something in the branch"""
330
132
return os.path.join(self.base, name)
332
135
def relpath(self, path):
333
136
"""Return path relative to this branch of something inside it.
335
138
Raises an error if path is not in this branch."""
336
return _relpath(self.base, path)
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)
338
148
def controlfilename(self, file_or_path):
339
149
"""Return location relative to branch."""
340
if isinstance(file_or_path, basestring):
150
if isinstance(file_or_path, types.StringTypes):
341
151
file_or_path = [file_or_path]
342
152
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
345
155
def controlfile(self, file_or_path, mode='r'):
346
"""Open a control file for this branch.
348
There are two classes of file in the control directory: text
349
and binary. binary files are untranslated byte streams. Text
350
control files are stored with Unix newlines and in UTF-8, even
351
if the platform or locale defaults are different.
353
Controlfiles should almost never be opened in write mode but
354
rather should be atomically copied and replaced using atomicfile.
357
fn = self.controlfilename(file_or_path)
359
if mode == 'rb' or mode == 'wb':
360
return file(fn, mode)
361
elif mode == 'r' or mode == 'w':
362
# open in binary mode anyhow so there's no newline translation;
363
# codecs uses line buffering by default; don't want that.
365
return codecs.open(fn, mode + 'b', 'utf-8',
368
raise BzrError("invalid controlfile mode %r" % mode)
156
"""Open a control file for this branch"""
157
return file(self.controlfilename(file_or_path), mode)
370
160
def _make_control(self):
371
161
os.mkdir(self.controlfilename([]))
372
162
self.controlfile('README', 'w').write(
373
163
"This is a Bazaar-NG control directory.\n"
374
"Do not change any files in this directory.\n")
375
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT_5)
376
for d in ('text-store', 'revision-store',
164
"Do not change any files in this directory.")
165
self.controlfile('branch-format', 'wb').write(BZR_BRANCH_FORMAT)
166
for d in ('text-store', 'inventory-store', 'revision-store'):
378
167
os.mkdir(self.controlfilename(d))
379
for f in ('revision-history',
168
for f in ('revision-history', 'merged-patches',
169
'pending-merged-patches', 'branch-name'):
383
170
self.controlfile(f, 'w').write('')
384
171
mutter('created control directory in ' + self.base)
386
# if we want per-tree root ids then this is the place to set
387
# them; they're not needed for now and so ommitted for
389
f = self.controlfile('inventory','w')
390
bzrlib.xml5.serializer_v5.write_inventory(Inventory(), f)
393
def _check_format(self, relax_version_check):
172
Inventory().write_xml(self.controlfile('inventory','w'))
175
def _check_format(self):
394
176
"""Check this branch format is supported.
396
The format level is stored, as an integer, in
397
self._branch_format for code that needs to check it later.
178
The current tool only supports the current unstable format.
399
180
In the future, we might need different in-memory Branch
400
181
classes to support downlevel branches. But not yet.
403
fmt = self.controlfile('branch-format', 'r').read()
405
if e.errno == errno.ENOENT:
406
raise NotBranchError(self.base)
410
if fmt == BZR_BRANCH_FORMAT_5:
411
self._branch_format = 5
412
elif fmt == BZR_BRANCH_FORMAT_4:
413
self._branch_format = 4
415
if (not relax_version_check
416
and self._branch_format != 5):
417
raise BzrError('sorry, branch format %r not supported' % fmt,
418
['use a different bzr version',
419
'or remove the .bzr directory and "bzr init" again'])
421
def get_root_id(self):
422
"""Return the id of this branches root"""
423
inv = self.read_working_inventory()
424
return inv.root.file_id
426
def set_root_id(self, file_id):
427
inv = self.read_working_inventory()
428
orig_root_id = inv.root.file_id
429
del inv._byid[inv.root.file_id]
430
inv.root.file_id = file_id
431
inv._byid[inv.root.file_id] = inv.root
434
if entry.parent_id in (None, orig_root_id):
435
entry.parent_id = inv.root.file_id
436
self._write_inventory(inv)
183
# This ignores newlines so that we can open branches created
184
# on Windows from Linux and so on. I think it might be better
185
# to always make all internal files in unix format.
186
fmt = self.controlfile('branch-format', 'rb').read()
187
fmt.replace('\r\n', '')
188
if fmt != BZR_BRANCH_FORMAT:
189
bailout('sorry, branch format %r not supported' % fmt,
190
['use a different bzr version',
191
'or remove the .bzr directory and "bzr init" again'])
438
194
def read_working_inventory(self):
439
195
"""Read the working inventory."""
442
# ElementTree does its own conversion from UTF-8, so open in
444
f = self.controlfile('inventory', 'rb')
445
return bzrlib.xml5.serializer_v5.read_inventory(f)
197
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
198
mutter("loaded inventory of %d items in %f"
199
% (len(inv), time.time() - before))
450
203
def _write_inventory(self, inv):
451
204
"""Update the working inventory.
453
206
That is to say, the inventory describing changes underway, that
454
207
will be committed to the next revision.
456
from bzrlib.atomicfile import AtomicFile
460
f = AtomicFile(self.controlfilename('inventory'), 'wb')
462
bzrlib.xml5.serializer_v5.write_inventory(inv, f)
209
## TODO: factor out to atomicfile? is rename safe on windows?
210
## TODO: Maybe some kind of clean/dirty marker on inventory?
211
tmpfname = self.controlfilename('inventory.tmp')
212
tmpf = file(tmpfname, 'w')
215
inv_fname = self.controlfilename('inventory')
216
if sys.platform == 'win32':
218
os.rename(tmpfname, inv_fname)
469
219
mutter('wrote working inventory')
472
222
inventory = property(read_working_inventory, _write_inventory, None,
473
223
"""Inventory for the working copy.""")
476
def add(self, files, ids=None):
226
def add(self, files, verbose=False):
477
227
"""Make files versioned.
479
Note that the command line normally calls smart_add instead,
480
which can automatically recurse.
482
229
This puts the files in the Added state, so that they will be
483
230
recorded by the next commit.
486
List of paths to add, relative to the base of the tree.
489
If set, use these instead of automatically generated ids.
490
Must be the same length as the list of files, but may
491
contain None for ids that are to be autogenerated.
493
TODO: Perhaps have an option to add the ids even if the files do
496
TODO: Perhaps yield the ids and paths as they're added.
232
:todo: Perhaps have an option to add the ids even if the files do
235
:todo: Perhaps return the ids of the files? But then again it
236
is easy to retrieve them if they're needed.
238
:todo: Option to specify file id.
240
:todo: Adding a directory should optionally recurse down and
241
add all non-ignored children. Perhaps do that in a
244
>>> b = ScratchBranch(files=['foo'])
245
>>> 'foo' in b.unknowns()
250
>>> 'foo' in b.unknowns()
252
>>> bool(b.inventory.path2id('foo'))
258
Traceback (most recent call last):
260
BzrError: ('foo is already versioned', [])
262
>>> b.add(['nothere'])
263
Traceback (most recent call last):
264
BzrError: ('cannot add: not a regular file or directory: nothere', [])
498
267
# TODO: Re-adding a file that is removed in the working copy
499
268
# should probably put it back with the previous ID.
500
if isinstance(files, basestring):
501
assert(ids is None or isinstance(ids, basestring))
269
if isinstance(files, types.StringTypes):
507
ids = [None] * len(files)
509
assert(len(ids) == len(files))
513
inv = self.read_working_inventory()
514
for f,file_id in zip(files, ids):
515
if is_control_file(f):
516
raise BzrError("cannot add control file %s" % quotefn(f))
521
raise BzrError("cannot add top-level %r" % f)
523
fullpath = os.path.normpath(self.abspath(f))
526
kind = file_kind(fullpath)
528
# maybe something better?
529
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
531
if kind != 'file' and kind != 'directory':
532
raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))
535
file_id = gen_file_id(f)
536
inv.add_path(f, kind=kind, file_id=file_id)
538
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
540
self._write_inventory(inv)
545
def print_file(self, file, revno):
546
"""Print `file` to stdout."""
549
tree = self.revision_tree(self.get_rev_id(revno))
550
# use inventory as it was in that revision
551
file_id = tree.inventory.path2id(file)
553
raise BzrError("%r is not present in revision %s" % (file, revno))
554
tree.print_file(file_id)
272
inv = self.read_working_inventory()
274
if is_control_file(f):
275
bailout("cannot add control file %s" % quotefn(f))
280
bailout("cannot add top-level %r" % f)
282
fullpath = os.path.normpath(self.abspath(f))
285
kind = file_kind(fullpath)
287
# maybe something better?
288
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
290
if kind != 'file' and kind != 'directory':
291
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
293
file_id = gen_file_id(f)
294
inv.add_path(f, kind=kind, file_id=file_id)
297
show_status('A', kind, quotefn(f))
299
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
301
self._write_inventory(inv)
559
305
def remove(self, files, verbose=False):
562
308
This does not remove their text. This does not run on
564
TODO: Refuse to remove modified files unless --force is given?
566
TODO: Do something useful with directories.
568
TODO: Should this remove the text or not? Tough call; not
310
:todo: Refuse to remove modified files unless --force is given?
312
>>> b = ScratchBranch(files=['foo'])
314
>>> b.inventory.has_filename('foo')
317
>>> b.working_tree().has_filename('foo')
319
>>> b.inventory.has_filename('foo')
322
>>> b = ScratchBranch(files=['foo'])
327
>>> b.inventory.has_filename('foo')
329
>>> b.basis_tree().has_filename('foo')
331
>>> b.working_tree().has_filename('foo')
334
:todo: Do something useful with directories.
336
:todo: Should this remove the text or not? Tough call; not
569
337
removing may be useful and the user can just use use rm, and
570
338
is the opposite of add. Removing it is consistent with most
571
339
other tools. Maybe an option.
573
341
## TODO: Normalize names
574
342
## TODO: Remove nested loops; better scalability
575
if isinstance(files, basestring):
344
if isinstance(files, types.StringTypes):
581
tree = self.working_tree()
584
# do this before any modifications
588
raise BzrError("cannot remove unversioned file %s" % quotefn(f))
589
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
591
# having remove it, it must be either ignored or unknown
592
if tree.is_ignored(f):
596
show_status(new_status, inv[fid].kind, quotefn(f))
599
self._write_inventory(inv)
604
# FIXME: this doesn't need to be a branch method
605
def set_inventory(self, new_inventory_list):
606
from bzrlib.inventory import Inventory, InventoryEntry
607
inv = Inventory(self.get_root_id())
608
for path, file_id, parent, kind in new_inventory_list:
609
name = os.path.basename(path)
612
inv.add(InventoryEntry(file_id, name, kind, parent))
347
tree = self.working_tree()
350
# do this before any modifications
354
bailout("cannot remove unversioned file %s" % quotefn(f))
355
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
357
# having remove it, it must be either ignored or unknown
358
if tree.is_ignored(f):
362
show_status(new_status, inv[fid].kind, quotefn(f))
613
365
self._write_inventory(inv)
632
384
return self.working_tree().unknowns()
635
def append_revision(self, *revision_ids):
636
from bzrlib.atomicfile import AtomicFile
638
for revision_id in revision_ids:
639
mutter("add {%s} to revision-history" % revision_id)
641
rev_history = self.revision_history()
642
rev_history.extend(revision_ids)
644
f = AtomicFile(self.controlfilename('revision-history'))
646
for rev_id in rev_history:
653
def has_revision(self, revision_id):
654
"""True if this branch has a copy of the revision.
656
This does not necessarily imply the revision is merge
657
or on the mainline."""
658
return (revision_id is None
659
or revision_id in self.revision_store)
662
def get_revision_xml_file(self, revision_id):
663
"""Return XML file object for revision object."""
664
if not revision_id or not isinstance(revision_id, basestring):
665
raise InvalidRevisionId(revision_id)
670
return self.revision_store[revision_id]
671
except (IndexError, KeyError):
672
raise bzrlib.errors.NoSuchRevision(self, revision_id)
677
def get_revision_xml(self, revision_id):
678
return self.get_revision_xml_file(revision_id).read()
387
def commit(self, message, timestamp=None, timezone=None,
390
"""Commit working copy as a new revision.
392
The basic approach is to add all the file texts into the
393
store, then the inventory, then make a new revision pointing
394
to that inventory and store that.
396
This is not quite safe if the working copy changes during the
397
commit; for the moment that is simply not allowed. A better
398
approach is to make a temporary copy of the files before
399
computing their hashes, and then add those hashes in turn to
400
the inventory. This should mean at least that there are no
401
broken hash pointers. There is no way we can get a snapshot
402
of the whole directory at an instant. This would also have to
403
be robust against files disappearing, moving, etc. So the
404
whole thing is a bit hard.
406
:param timestamp: if not None, seconds-since-epoch for a
407
postdated/predated commit.
410
## TODO: Show branch names
412
# TODO: Don't commit if there are no changes, unless forced?
414
# First walk over the working inventory; and both update that
415
# and also build a new revision inventory. The revision
416
# inventory needs to hold the text-id, sha1 and size of the
417
# actual file versions committed in the revision. (These are
418
# not present in the working inventory.) We also need to
419
# detect missing/deleted files, and remove them from the
422
work_inv = self.read_working_inventory()
424
basis = self.basis_tree()
425
basis_inv = basis.inventory
427
for path, entry in work_inv.iter_entries():
428
## TODO: Cope with files that have gone missing.
430
## TODO: Check that the file kind has not changed from the previous
431
## revision of this file (if any).
435
p = self.abspath(path)
436
file_id = entry.file_id
437
mutter('commit prep file %s, id %r ' % (p, file_id))
439
if not os.path.exists(p):
440
mutter(" file is missing, removing from inventory")
442
show_status('D', entry.kind, quotefn(path))
443
missing_ids.append(file_id)
446
# TODO: Handle files that have been deleted
448
# TODO: Maybe a special case for empty files? Seems a
449
# waste to store them many times.
453
if basis_inv.has_id(file_id):
454
old_kind = basis_inv[file_id].kind
455
if old_kind != entry.kind:
456
bailout("entry %r changed kind from %r to %r"
457
% (file_id, old_kind, entry.kind))
459
if entry.kind == 'directory':
461
bailout("%s is entered as directory but not a directory" % quotefn(p))
462
elif entry.kind == 'file':
464
bailout("%s is entered as file but is not a file" % quotefn(p))
466
content = file(p, 'rb').read()
468
entry.text_sha1 = sha_string(content)
469
entry.text_size = len(content)
471
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
473
and (old_ie.text_size == entry.text_size)
474
and (old_ie.text_sha1 == entry.text_sha1)):
475
## assert content == basis.get_file(file_id).read()
476
entry.text_id = basis_inv[file_id].text_id
477
mutter(' unchanged from previous text_id {%s}' %
481
entry.text_id = gen_file_id(entry.name)
482
self.text_store.add(content, entry.text_id)
483
mutter(' stored with text_id {%s}' % entry.text_id)
487
elif (old_ie.name == entry.name
488
and old_ie.parent_id == entry.parent_id):
493
show_status(state, entry.kind, quotefn(path))
495
for file_id in missing_ids:
496
# have to do this later so we don't mess up the iterator.
497
# since parents may be removed before their children we
500
# FIXME: There's probably a better way to do this; perhaps
501
# the workingtree should know how to filter itself.
502
if work_inv.has_id(file_id):
503
del work_inv[file_id]
506
inv_id = rev_id = _gen_revision_id(time.time())
508
inv_tmp = tempfile.TemporaryFile()
509
inv.write_xml(inv_tmp)
511
self.inventory_store.add(inv_tmp, inv_id)
512
mutter('new inventory_id is {%s}' % inv_id)
514
self._write_inventory(work_inv)
516
if timestamp == None:
517
timestamp = time.time()
519
if committer == None:
520
committer = username()
523
timezone = local_time_offset()
525
mutter("building commit log message")
526
rev = Revision(timestamp=timestamp,
529
precursor = self.last_patch(),
534
rev_tmp = tempfile.TemporaryFile()
535
rev.write_xml(rev_tmp)
537
self.revision_store.add(rev_tmp, rev_id)
538
mutter("new revision_id is {%s}" % rev_id)
540
## XXX: Everything up to here can simply be orphaned if we abort
541
## the commit; it will leave junk files behind but that doesn't
544
## TODO: Read back the just-generated changeset, and make sure it
545
## applies and recreates the right state.
547
## TODO: Also calculate and store the inventory SHA1
548
mutter("committing patch r%d" % (self.revno() + 1))
550
mutter("append to revision-history")
551
f = self.controlfile('revision-history', 'at')
552
f.write(rev_id + '\n')
556
note("commited r%d" % self.revno())
681
559
def get_revision(self, revision_id):
682
560
"""Return the Revision object for a named revision"""
683
xml_file = self.get_revision_xml_file(revision_id)
686
r = bzrlib.xml5.serializer_v5.read_revision(xml_file)
687
except SyntaxError, e:
688
raise bzrlib.errors.BzrError('failed to unpack revision_xml',
561
r = Revision.read_xml(self.revision_store[revision_id])
692
562
assert r.revision_id == revision_id
696
def get_revision_delta(self, revno):
697
"""Return the delta for one revision.
699
The delta is relative to its mainline predecessor, or the
700
empty tree for revision 1.
702
assert isinstance(revno, int)
703
rh = self.revision_history()
704
if not (1 <= revno <= len(rh)):
705
raise InvalidRevisionNumber(revno)
707
# revno is 1-based; list is 0-based
709
new_tree = self.revision_tree(rh[revno-1])
711
old_tree = EmptyTree()
713
old_tree = self.revision_tree(rh[revno-2])
715
return compare_trees(old_tree, new_tree)
718
def get_revision_sha1(self, revision_id):
719
"""Hash the stored value of a revision, and return it."""
720
return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))
723
def _get_ancestry_weave(self):
724
return self.control_weaves.get_weave('ancestry')
727
def get_ancestry(self, revision_id):
728
"""Return a list of revision-ids integrated by a revision.
731
if revision_id is None:
733
w = self._get_ancestry_weave()
734
return [None] + [l[:-1] for l in w.get_iter(w.lookup(revision_id))]
737
def get_inventory_weave(self):
738
return self.control_weaves.get_weave('inventory')
741
def get_inventory(self, revision_id):
742
"""Get Inventory object by hash."""
743
xml = self.get_inventory_xml(revision_id)
744
return bzrlib.xml5.serializer_v5.read_inventory_from_string(xml)
747
def get_inventory_xml(self, revision_id):
748
"""Get inventory XML as a file object."""
750
assert isinstance(revision_id, basestring), type(revision_id)
751
iw = self.get_inventory_weave()
752
return iw.get_text(iw.lookup(revision_id))
754
raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)
757
def get_inventory_sha1(self, revision_id):
758
"""Return the sha1 hash of the inventory entry
760
return self.get_revision(revision_id).inventory_sha1
566
def get_inventory(self, inventory_id):
567
"""Get Inventory object by hash.
569
:todo: Perhaps for this and similar methods, take a revision
570
parameter which can be either an integer revno or a
572
i = Inventory.read_xml(self.inventory_store[inventory_id])
763
576
def get_revision_inventory(self, revision_id):
764
577
"""Return inventory of a past revision."""
765
# TODO: Unify this with get_inventory()
766
# bzr 0.0.6 and later imposes the constraint that the inventory_id
767
# must be the same as its revision, so this is trivial.
768
578
if revision_id == None:
769
return Inventory(self.get_root_id())
771
return self.get_inventory(revision_id)
581
return self.get_inventory(self.get_revision(revision_id).inventory_id)
774
584
def revision_history(self):
775
"""Return sequence of revision hashes on to this branch."""
778
return [l.rstrip('\r\n') for l in
779
self.controlfile('revision-history', 'r').readlines()]
784
def common_ancestor(self, other, self_revno=None, other_revno=None):
786
>>> from bzrlib.commit import commit
787
>>> sb = ScratchBranch(files=['foo', 'foo~'])
788
>>> sb.common_ancestor(sb) == (None, None)
790
>>> commit(sb, "Committing first revision", verbose=False)
791
>>> sb.common_ancestor(sb)[0]
793
>>> clone = sb.clone()
794
>>> commit(sb, "Committing second revision", verbose=False)
795
>>> sb.common_ancestor(sb)[0]
797
>>> sb.common_ancestor(clone)[0]
799
>>> commit(clone, "Committing divergent second revision",
801
>>> sb.common_ancestor(clone)[0]
803
>>> sb.common_ancestor(clone) == clone.common_ancestor(sb)
805
>>> sb.common_ancestor(sb) != clone.common_ancestor(clone)
807
>>> clone2 = sb.clone()
808
>>> sb.common_ancestor(clone2)[0]
810
>>> sb.common_ancestor(clone2, self_revno=1)[0]
812
>>> sb.common_ancestor(clone2, other_revno=1)[0]
815
my_history = self.revision_history()
816
other_history = other.revision_history()
817
if self_revno is None:
818
self_revno = len(my_history)
819
if other_revno is None:
820
other_revno = len(other_history)
821
indices = range(min((self_revno, other_revno)))
824
if my_history[r] == other_history[r]:
825
return r+1, my_history[r]
585
"""Return sequence of revision hashes on to this branch.
587
>>> ScratchBranch().revision_history()
590
return [chomp(l) for l in self.controlfile('revision-history').readlines()]
832
596
That is equivalent to the number of revisions committed to
599
>>> b = ScratchBranch()
602
>>> b.commit('no foo')
835
606
return len(self.revision_history())
838
def last_revision(self):
609
def last_patch(self):
839
610
"""Return last patch hash, or None if no history.
612
>>> ScratchBranch().last_patch() == None
841
615
ph = self.revision_history()
848
def missing_revisions(self, other, stop_revision=None, diverged_ok=False):
849
"""Return a list of new revisions that would perfectly fit.
851
If self and other have not diverged, return a list of the revisions
852
present in other, but missing from self.
854
>>> from bzrlib.commit import commit
855
>>> bzrlib.trace.silent = True
856
>>> br1 = ScratchBranch()
857
>>> br2 = ScratchBranch()
858
>>> br1.missing_revisions(br2)
860
>>> commit(br2, "lala!", rev_id="REVISION-ID-1")
861
>>> br1.missing_revisions(br2)
863
>>> br2.missing_revisions(br1)
865
>>> commit(br1, "lala!", rev_id="REVISION-ID-1")
866
>>> br1.missing_revisions(br2)
868
>>> commit(br2, "lala!", rev_id="REVISION-ID-2A")
869
>>> br1.missing_revisions(br2)
871
>>> commit(br1, "lala!", rev_id="REVISION-ID-2B")
872
>>> br1.missing_revisions(br2)
873
Traceback (most recent call last):
874
DivergedBranches: These branches have diverged.
876
# FIXME: If the branches have diverged, but the latest
877
# revision in this branch is completely merged into the other,
878
# then we should still be able to pull.
879
self_history = self.revision_history()
880
self_len = len(self_history)
881
other_history = other.revision_history()
882
other_len = len(other_history)
883
common_index = min(self_len, other_len) -1
884
if common_index >= 0 and \
885
self_history[common_index] != other_history[common_index]:
886
raise DivergedBranches(self, other)
888
if stop_revision is None:
889
stop_revision = other_len
891
assert isinstance(stop_revision, int)
892
if stop_revision > other_len:
893
raise bzrlib.errors.NoSuchRevision(self, stop_revision)
894
return other_history[self_len:stop_revision]
896
def update_revisions(self, other, stop_revision=None):
897
"""Pull in new perfect-fit revisions."""
898
from bzrlib.fetch import greedy_fetch
899
from bzrlib.revision import get_intervening_revisions
900
if stop_revision is None:
901
stop_revision = other.last_revision()
902
greedy_fetch(to_branch=self, from_branch=other,
903
revision=stop_revision)
904
pullable_revs = self.missing_revisions(
905
other, other.revision_id_to_revno(stop_revision))
907
greedy_fetch(to_branch=self,
909
revision=pullable_revs[-1])
910
self.append_revision(*pullable_revs)
913
def commit(self, *args, **kw):
914
from bzrlib.commit import Commit
915
Commit().commit(self, *args, **kw)
917
def revision_id_to_revno(self, revision_id):
918
"""Given a revision id, return its revno"""
919
if revision_id is None:
921
history = self.revision_history()
923
return history.index(revision_id) + 1
925
raise bzrlib.errors.NoSuchRevision(self, revision_id)
927
def get_rev_id(self, revno, history=None):
928
"""Find the revision id of the specified revno."""
620
def lookup_revision(self, revno):
621
"""Return revision hash for revision number."""
932
history = self.revision_history()
933
elif revno <= 0 or revno > len(history):
934
raise bzrlib.errors.NoSuchRevision(self, revno)
935
return history[revno - 1]
626
# list is 0-based; revisions are 1-based
627
return self.revision_history()[revno-1]
629
bailout("no such revision %s" % revno)
937
632
def revision_tree(self, revision_id):
938
633
"""Return Tree for a revision on this branch.
940
635
`revision_id` may be None for the null revision, in which case
941
636
an `EmptyTree` is returned."""
942
# TODO: refactor this to use an existing revision object
943
# so we don't need to read it in twice.
944
638
if revision_id == None:
945
639
return EmptyTree()
947
641
inv = self.get_revision_inventory(revision_id)
948
return RevisionTree(self.weave_store, inv, revision_id)
642
return RevisionTree(self.text_store, inv)
951
645
def working_tree(self):
952
646
"""Return a `Tree` for the working copy."""
953
from bzrlib.workingtree import WorkingTree
954
647
return WorkingTree(self.base, self.read_working_inventory())
958
651
"""Return `Tree` object for last revision.
960
653
If there are no revisions yet, return an `EmptyTree`.
655
>>> b = ScratchBranch(files=['foo'])
656
>>> b.basis_tree().has_filename('foo')
658
>>> b.working_tree().has_filename('foo')
661
>>> b.commit('add foo')
662
>>> b.basis_tree().has_filename('foo')
962
return self.revision_tree(self.last_revision())
665
r = self.last_patch()
669
return RevisionTree(self.text_store, self.get_revision_inventory(r))
673
def write_log(self, show_timezone='original'):
674
"""Write out human-readable log of commits to this branch
676
:param utc: If true, show dates in universal time, not local time."""
677
## TODO: Option to choose either original, utc or local timezone
680
for p in self.revision_history():
682
print 'revno:', revno
683
## TODO: Show hash if --id is given.
684
##print 'revision-hash:', p
685
rev = self.get_revision(p)
686
print 'committer:', rev.committer
687
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
690
## opportunistic consistency check, same as check_patch_chaining
691
if rev.precursor != precursor:
692
bailout("mismatched precursor!")
696
print ' (no message)'
698
for l in rev.message.split('\n'):
965
705
def rename_one(self, from_rel, to_rel):
968
This can change the directory or the filename or both.
972
tree = self.working_tree()
974
if not tree.has_filename(from_rel):
975
raise BzrError("can't rename: old working file %r does not exist" % from_rel)
976
if tree.has_filename(to_rel):
977
raise BzrError("can't rename: new working file %r already exists" % to_rel)
979
file_id = inv.path2id(from_rel)
981
raise BzrError("can't rename: old name %r is not versioned" % from_rel)
983
if inv.path2id(to_rel):
984
raise BzrError("can't rename: new name %r is already versioned" % to_rel)
986
to_dir, to_tail = os.path.split(to_rel)
987
to_dir_id = inv.path2id(to_dir)
988
if to_dir_id == None and to_dir != '':
989
raise BzrError("can't determine destination directory id for %r" % to_dir)
991
mutter("rename_one:")
992
mutter(" file_id {%s}" % file_id)
993
mutter(" from_rel %r" % from_rel)
994
mutter(" to_rel %r" % to_rel)
995
mutter(" to_dir %r" % to_dir)
996
mutter(" to_dir_id {%s}" % to_dir_id)
998
inv.rename(file_id, to_dir_id, to_tail)
1000
from_abs = self.abspath(from_rel)
1001
to_abs = self.abspath(to_rel)
1003
rename(from_abs, to_abs)
1005
raise BzrError("failed to rename %r to %r: %s"
1006
% (from_abs, to_abs, e[1]),
1007
["rename rolled back"])
1009
self._write_inventory(inv)
1014
def move(self, from_paths, to_name):
706
tree = self.working_tree()
708
if not tree.has_filename(from_rel):
709
bailout("can't rename: old working file %r does not exist" % from_rel)
710
if tree.has_filename(to_rel):
711
bailout("can't rename: new working file %r already exists" % to_rel)
713
file_id = inv.path2id(from_rel)
715
bailout("can't rename: old name %r is not versioned" % from_rel)
717
if inv.path2id(to_rel):
718
bailout("can't rename: new name %r is already versioned" % to_rel)
720
to_dir, to_tail = os.path.split(to_rel)
721
to_dir_id = inv.path2id(to_dir)
722
if to_dir_id == None and to_dir != '':
723
bailout("can't determine destination directory id for %r" % to_dir)
725
mutter("rename_one:")
726
mutter(" file_id {%s}" % file_id)
727
mutter(" from_rel %r" % from_rel)
728
mutter(" to_rel %r" % to_rel)
729
mutter(" to_dir %r" % to_dir)
730
mutter(" to_dir_id {%s}" % to_dir_id)
732
inv.rename(file_id, to_dir_id, to_tail)
733
os.rename(self.abspath(from_rel), self.abspath(to_rel))
735
self._write_inventory(inv)
739
def rename(self, from_paths, to_name):
1015
740
"""Rename files.
1017
to_name must exist as a versioned directory.
1019
742
If to_name exists and is a directory, the files are moved into
1020
743
it, keeping their old names. If it is a directory,
1022
745
Note that to_name is only the last component of the new name;
1023
746
this doesn't change the directory.
1025
This returns a list of (from_path, to_path) pairs for each
1026
entry that is moved.
1031
## TODO: Option to move IDs only
1032
assert not isinstance(from_paths, basestring)
1033
tree = self.working_tree()
1034
inv = tree.inventory
1035
to_abs = self.abspath(to_name)
1036
if not isdir(to_abs):
1037
raise BzrError("destination %r is not a directory" % to_abs)
1038
if not tree.has_filename(to_name):
1039
raise BzrError("destination %r not in working directory" % to_abs)
1040
to_dir_id = inv.path2id(to_name)
1041
if to_dir_id == None and to_name != '':
1042
raise BzrError("destination %r is not a versioned directory" % to_name)
1043
to_dir_ie = inv[to_dir_id]
1044
if to_dir_ie.kind not in ('directory', 'root_directory'):
1045
raise BzrError("destination %r is not a directory" % to_abs)
1047
to_idpath = inv.get_idpath(to_dir_id)
1049
for f in from_paths:
1050
if not tree.has_filename(f):
1051
raise BzrError("%r does not exist in working tree" % f)
1052
f_id = inv.path2id(f)
1054
raise BzrError("%r is not versioned" % f)
1055
name_tail = splitpath(f)[-1]
1056
dest_path = appendpath(to_name, name_tail)
1057
if tree.has_filename(dest_path):
1058
raise BzrError("destination %r already exists" % dest_path)
1059
if f_id in to_idpath:
1060
raise BzrError("can't move %r to a subdirectory of itself" % f)
1062
# OK, so there's a race here, it's possible that someone will
1063
# create a file in this interval and then the rename might be
1064
# left half-done. But we should have caught most problems.
1066
for f in from_paths:
1067
name_tail = splitpath(f)[-1]
1068
dest_path = appendpath(to_name, name_tail)
1069
result.append((f, dest_path))
1070
inv.rename(inv.path2id(f), to_dir_id, name_tail)
1072
rename(self.abspath(f), self.abspath(dest_path))
1074
raise BzrError("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
1075
["rename rolled back"])
748
## TODO: Option to move IDs only
749
assert not isinstance(from_paths, basestring)
750
tree = self.working_tree()
752
dest_dir = isdir(self.abspath(to_name))
754
# TODO: Wind back properly if some can't be moved?
755
dest_dir_id = inv.path2id(to_name)
756
if not dest_dir_id and to_name != '':
757
bailout("destination %r is not a versioned directory" % to_name)
759
name_tail = splitpath(f)[-1]
760
dest_path = appendpath(to_name, name_tail)
761
print "%s => %s" % (f, dest_path)
762
inv.rename(inv.path2id(f), dest_dir_id, name_tail)
763
os.rename(self.abspath(f), self.abspath(dest_path))
1077
764
self._write_inventory(inv)
1084
def revert(self, filenames, old_tree=None, backups=True):
1085
"""Restore selected files to the versions from a previous tree.
1088
If true (default) backups are made of files before
1091
from bzrlib.errors import NotVersionedError, BzrError
1092
from bzrlib.atomicfile import AtomicFile
1093
from bzrlib.osutils import backup_file
1095
inv = self.read_working_inventory()
1096
if old_tree is None:
1097
old_tree = self.basis_tree()
1098
old_inv = old_tree.inventory
1101
for fn in filenames:
1102
file_id = inv.path2id(fn)
1104
raise NotVersionedError("not a versioned file", fn)
1105
if not old_inv.has_id(file_id):
1106
raise BzrError("file not present in old tree", fn, file_id)
1107
nids.append((fn, file_id))
1109
# TODO: Rename back if it was previously at a different location
1111
# TODO: If given a directory, restore the entire contents from
1112
# the previous version.
1114
# TODO: Make a backup to a temporary file.
1116
# TODO: If the file previously didn't exist, delete it?
1117
for fn, file_id in nids:
1120
f = AtomicFile(fn, 'wb')
1122
f.write(old_tree.get_file(file_id).read())
1128
def pending_merges(self):
1129
"""Return a list of pending merges.
1131
These are revisions that have been merged into the working
1132
directory but not yet committed.
1134
cfn = self.controlfilename('pending-merges')
1135
if not os.path.exists(cfn):
1138
for l in self.controlfile('pending-merges', 'r').readlines():
1139
p.append(l.rstrip('\n'))
1143
def add_pending_merge(self, revision_id):
1144
validate_revision_id(revision_id)
1145
# TODO: Perhaps should check at this point that the
1146
# history of the revision is actually present?
1147
p = self.pending_merges()
1148
if revision_id in p:
1150
p.append(revision_id)
1151
self.set_pending_merges(p)
1154
def set_pending_merges(self, rev_list):
1155
from bzrlib.atomicfile import AtomicFile
1158
f = AtomicFile(self.controlfilename('pending-merges'))
1169
def get_parent(self):
1170
"""Return the parent location of the branch.
1172
This is the default location for push/pull/missing. The usual
1173
pattern is that the user can override it by specifying a
1177
_locs = ['parent', 'pull', 'x-pull']
1180
return self.controlfile(l, 'r').read().strip('\n')
1182
if e.errno != errno.ENOENT:
1187
def set_parent(self, url):
1188
# TODO: Maybe delete old location files?
1189
from bzrlib.atomicfile import AtomicFile
1192
f = AtomicFile(self.controlfilename('parent'))
1201
def check_revno(self, revno):
1203
Check whether a revno corresponds to any revision.
1204
Zero (the NULL revision) is considered valid.
1207
self.check_real_revno(revno)
1209
def check_real_revno(self, revno):
1211
Check whether a revno corresponds to a real revision.
1212
Zero (the NULL revision) is considered invalid
1214
if revno < 1 or revno > self.revno():
1215
raise InvalidRevisionNumber(revno)
1221
class ScratchBranch(LocalBranch):
766
if len(from_paths) != 1:
767
bailout("when moving multiple files, destination must be a directory")
768
bailout("rename to non-directory %r not implemented sorry" % to_name)
772
def show_status(branch, show_all=False):
773
"""Display single-line status for non-ignored working files.
775
The list is show sorted in order by file name.
777
>>> b = ScratchBranch(files=['foo', 'foo~'])
783
>>> b.commit("add foo")
785
>>> os.unlink(b.abspath('foo'))
790
:todo: Get state for single files.
792
:todo: Perhaps show a slash at the end of directory names.
796
# We have to build everything into a list first so that it can
797
# sorted by name, incorporating all the different sources.
799
# FIXME: Rather than getting things in random order and then sorting,
800
# just step through in order.
802
# Interesting case: the old ID for a file has been removed,
803
# but a new file has been created under that name.
805
old = branch.basis_tree()
806
old_inv = old.inventory
807
new = branch.working_tree()
808
new_inv = new.inventory
810
for fs, fid, oldname, newname, kind in diff_trees(old, new):
812
show_status(fs, kind,
813
oldname + ' => ' + newname)
814
elif fs == 'A' or fs == 'M':
815
show_status(fs, kind, newname)
817
show_status(fs, kind, oldname)
820
show_status(fs, kind, newname)
823
show_status(fs, kind, newname)
825
show_status(fs, kind, newname)
827
bailout("wierd file state %r" % ((fs, fid),))
831
class ScratchBranch(Branch):
1222
832
"""Special test class: a branch that cleans up after itself.
1224
834
>>> b = ScratchBranch()
1225
835
>>> isdir(b.base)
1232
def __init__(self, files=[], dirs=[], base=None):
842
def __init__(self, files=[], dirs=[]):
1233
843
"""Make a test branch.
1235
845
This creates a temporary directory and runs init-tree in it.
1237
847
If any files are listed, they are created in the working copy.
1239
from tempfile import mkdtemp
1244
LocalBranch.__init__(self, base, init=init)
849
Branch.__init__(self, tempfile.mkdtemp(), init=True)
1246
851
os.mkdir(self.abspath(d))