1
# Copyright (C) 2005 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
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, 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"
39
## TODO: Maybe include checks for common corruption of newlines, etc?
43
def find_branch_root(f=None):
44
"""Find the branch root enclosing f, or pwd.
46
It is not necessary that f exists.
48
Basically we keep looking up until we find the control directory or
52
elif hasattr(os.path, 'realpath'):
53
f = os.path.realpath(f)
55
f = os.path.abspath(f)
61
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
63
head, tail = os.path.split(f)
65
# reached the root, whatever that may be
66
bailout('%r is not in a branch' % orig_f)
71
######################################################################
75
"""Branch holding a history of revisions.
77
:todo: Perhaps use different stores for different classes of object,
78
so that we can keep track of how much space each one uses,
79
or garbage-collect them.
81
:todo: Add a RemoteBranch subclass. For the basic case of read-only
82
HTTP access this should be very easy by,
83
just redirecting controlfile access into HTTP requests.
84
We would need a RemoteStore working similarly.
86
:todo: Keep the on-disk branch locked while the object exists.
88
:todo: mkdir() method.
90
def __init__(self, base, init=False, find_root=True):
91
"""Create new branch object at a particular location.
93
:param base: Base directory for the branch.
95
:param init: If True, create new control files in a previously
96
unversioned directory. If False, the branch must already
99
:param find_root: If true and init is false, find the root of the
100
existing branch containing base.
102
In the test suite, creation of new trees is tested using the
103
`ScratchBranch` class.
106
self.base = os.path.realpath(base)
109
self.base = find_branch_root(base)
111
self.base = os.path.realpath(base)
112
if not isdir(self.controlfilename('.')):
113
bailout("not a bzr branch: %s" % quotefn(base),
114
['use "bzr init" to initialize a new working tree',
115
'current bzr can only operate from top-of-tree'])
118
self.text_store = ImmutableStore(self.controlfilename('text-store'))
119
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
120
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
124
return '%s(%r)' % (self.__class__.__name__, self.base)
130
def abspath(self, name):
131
"""Return absolute filename for something in the branch"""
132
return os.path.join(self.base, name)
135
def relpath(self, path):
136
"""Return path relative to this branch of something inside it.
138
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)
148
def controlfilename(self, file_or_path):
149
"""Return location relative to branch."""
150
if isinstance(file_or_path, types.StringTypes):
151
file_or_path = [file_or_path]
152
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
155
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)
160
def _make_control(self):
161
os.mkdir(self.controlfilename([]))
162
self.controlfile('README', 'w').write(
163
"This is a Bazaar-NG control directory.\n"
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'):
167
os.mkdir(self.controlfilename(d))
168
for f in ('revision-history', 'merged-patches',
169
'pending-merged-patches', 'branch-name'):
170
self.controlfile(f, 'w').write('')
171
mutter('created control directory in ' + self.base)
172
Inventory().write_xml(self.controlfile('inventory','w'))
175
def _check_format(self):
176
"""Check this branch format is supported.
178
The current tool only supports the current unstable format.
180
In the future, we might need different in-memory Branch
181
classes to support downlevel branches. But not yet.
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'])
194
def read_working_inventory(self):
195
"""Read the working inventory."""
197
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
198
mutter("loaded inventory of %d items in %f"
199
% (len(inv), time.time() - before))
203
def _write_inventory(self, inv):
204
"""Update the working inventory.
206
That is to say, the inventory describing changes underway, that
207
will be committed to the next revision.
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)
219
mutter('wrote working inventory')
222
inventory = property(read_working_inventory, _write_inventory, None,
223
"""Inventory for the working copy.""")
226
def add(self, files, verbose=False):
227
"""Make files versioned.
229
This puts the files in the Added state, so that they will be
230
recorded by the next commit.
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', [])
267
# TODO: Re-adding a file that is removed in the working copy
268
# should probably put it back with the previous ID.
269
if isinstance(files, types.StringTypes):
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)
304
def print_file(self, file, revno):
305
"""Print `file` to stdout."""
306
tree = self.revision_tree(self.lookup_revision(revno))
307
# use inventory as it was in that revision
308
file_id = tree.inventory.path2id(file)
310
bailout("%r is not present in revision %d" % (file, revno))
311
tree.print_file(file_id)
314
def remove(self, files, verbose=False):
315
"""Mark nominated files for removal from the inventory.
317
This does not remove their text. This does not run on
319
:todo: Refuse to remove modified files unless --force is given?
321
>>> b = ScratchBranch(files=['foo'])
323
>>> b.inventory.has_filename('foo')
326
>>> b.working_tree().has_filename('foo')
328
>>> b.inventory.has_filename('foo')
331
>>> b = ScratchBranch(files=['foo'])
336
>>> b.inventory.has_filename('foo')
338
>>> b.basis_tree().has_filename('foo')
340
>>> b.working_tree().has_filename('foo')
343
:todo: Do something useful with directories.
345
:todo: Should this remove the text or not? Tough call; not
346
removing may be useful and the user can just use use rm, and
347
is the opposite of add. Removing it is consistent with most
348
other tools. Maybe an option.
350
## TODO: Normalize names
351
## TODO: Remove nested loops; better scalability
353
if isinstance(files, types.StringTypes):
356
tree = self.working_tree()
359
# do this before any modifications
363
bailout("cannot remove unversioned file %s" % quotefn(f))
364
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
366
# having remove it, it must be either ignored or unknown
367
if tree.is_ignored(f):
371
show_status(new_status, inv[fid].kind, quotefn(f))
374
self._write_inventory(inv)
378
"""Return all unknown files.
380
These are files in the working directory that are not versioned or
381
control files or ignored.
383
>>> b = ScratchBranch(files=['foo', 'foo~'])
384
>>> list(b.unknowns())
387
>>> list(b.unknowns())
390
>>> list(b.unknowns())
393
return self.working_tree().unknowns()
396
def commit(self, message, timestamp=None, timezone=None,
399
"""Commit working copy as a new revision.
401
The basic approach is to add all the file texts into the
402
store, then the inventory, then make a new revision pointing
403
to that inventory and store that.
405
This is not quite safe if the working copy changes during the
406
commit; for the moment that is simply not allowed. A better
407
approach is to make a temporary copy of the files before
408
computing their hashes, and then add those hashes in turn to
409
the inventory. This should mean at least that there are no
410
broken hash pointers. There is no way we can get a snapshot
411
of the whole directory at an instant. This would also have to
412
be robust against files disappearing, moving, etc. So the
413
whole thing is a bit hard.
415
:param timestamp: if not None, seconds-since-epoch for a
416
postdated/predated commit.
419
## TODO: Show branch names
421
# TODO: Don't commit if there are no changes, unless forced?
423
# First walk over the working inventory; and both update that
424
# and also build a new revision inventory. The revision
425
# inventory needs to hold the text-id, sha1 and size of the
426
# actual file versions committed in the revision. (These are
427
# not present in the working inventory.) We also need to
428
# detect missing/deleted files, and remove them from the
431
work_inv = self.read_working_inventory()
433
basis = self.basis_tree()
434
basis_inv = basis.inventory
436
for path, entry in work_inv.iter_entries():
437
## TODO: Cope with files that have gone missing.
439
## TODO: Check that the file kind has not changed from the previous
440
## revision of this file (if any).
444
p = self.abspath(path)
445
file_id = entry.file_id
446
mutter('commit prep file %s, id %r ' % (p, file_id))
448
if not os.path.exists(p):
449
mutter(" file is missing, removing from inventory")
451
show_status('D', entry.kind, quotefn(path))
452
missing_ids.append(file_id)
455
# TODO: Handle files that have been deleted
457
# TODO: Maybe a special case for empty files? Seems a
458
# waste to store them many times.
462
if basis_inv.has_id(file_id):
463
old_kind = basis_inv[file_id].kind
464
if old_kind != entry.kind:
465
bailout("entry %r changed kind from %r to %r"
466
% (file_id, old_kind, entry.kind))
468
if entry.kind == 'directory':
470
bailout("%s is entered as directory but not a directory" % quotefn(p))
471
elif entry.kind == 'file':
473
bailout("%s is entered as file but is not a file" % quotefn(p))
475
content = file(p, 'rb').read()
477
entry.text_sha1 = sha_string(content)
478
entry.text_size = len(content)
480
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
482
and (old_ie.text_size == entry.text_size)
483
and (old_ie.text_sha1 == entry.text_sha1)):
484
## assert content == basis.get_file(file_id).read()
485
entry.text_id = basis_inv[file_id].text_id
486
mutter(' unchanged from previous text_id {%s}' %
490
entry.text_id = gen_file_id(entry.name)
491
self.text_store.add(content, entry.text_id)
492
mutter(' stored with text_id {%s}' % entry.text_id)
496
elif (old_ie.name == entry.name
497
and old_ie.parent_id == entry.parent_id):
502
show_status(state, entry.kind, quotefn(path))
504
for file_id in missing_ids:
505
# have to do this later so we don't mess up the iterator.
506
# since parents may be removed before their children we
509
# FIXME: There's probably a better way to do this; perhaps
510
# the workingtree should know how to filter itself.
511
if work_inv.has_id(file_id):
512
del work_inv[file_id]
515
inv_id = rev_id = _gen_revision_id(time.time())
517
inv_tmp = tempfile.TemporaryFile()
518
inv.write_xml(inv_tmp)
520
self.inventory_store.add(inv_tmp, inv_id)
521
mutter('new inventory_id is {%s}' % inv_id)
523
self._write_inventory(work_inv)
525
if timestamp == None:
526
timestamp = time.time()
528
if committer == None:
529
committer = username()
532
timezone = local_time_offset()
534
mutter("building commit log message")
535
rev = Revision(timestamp=timestamp,
538
precursor = self.last_patch(),
543
rev_tmp = tempfile.TemporaryFile()
544
rev.write_xml(rev_tmp)
546
self.revision_store.add(rev_tmp, rev_id)
547
mutter("new revision_id is {%s}" % rev_id)
549
## XXX: Everything up to here can simply be orphaned if we abort
550
## the commit; it will leave junk files behind but that doesn't
553
## TODO: Read back the just-generated changeset, and make sure it
554
## applies and recreates the right state.
556
## TODO: Also calculate and store the inventory SHA1
557
mutter("committing patch r%d" % (self.revno() + 1))
559
mutter("append to revision-history")
560
f = self.controlfile('revision-history', 'at')
561
f.write(rev_id + '\n')
565
note("commited r%d" % self.revno())
568
def get_revision(self, revision_id):
569
"""Return the Revision object for a named revision"""
570
r = Revision.read_xml(self.revision_store[revision_id])
571
assert r.revision_id == revision_id
575
def get_inventory(self, inventory_id):
576
"""Get Inventory object by hash.
578
:todo: Perhaps for this and similar methods, take a revision
579
parameter which can be either an integer revno or a
581
i = Inventory.read_xml(self.inventory_store[inventory_id])
585
def get_revision_inventory(self, revision_id):
586
"""Return inventory of a past revision."""
587
if revision_id == None:
590
return self.get_inventory(self.get_revision(revision_id).inventory_id)
593
def revision_history(self):
594
"""Return sequence of revision hashes on to this branch.
596
>>> ScratchBranch().revision_history()
599
return [chomp(l) for l in self.controlfile('revision-history').readlines()]
603
"""Return current revision number for this branch.
605
That is equivalent to the number of revisions committed to
608
>>> b = ScratchBranch()
611
>>> b.commit('no foo')
615
return len(self.revision_history())
618
def last_patch(self):
619
"""Return last patch hash, or None if no history.
621
>>> ScratchBranch().last_patch() == None
624
ph = self.revision_history()
629
def lookup_revision(self, revno):
630
"""Return revision hash for revision number."""
635
# list is 0-based; revisions are 1-based
636
return self.revision_history()[revno-1]
638
bailout("no such revision %s" % revno)
641
def revision_tree(self, revision_id):
642
"""Return Tree for a revision on this branch.
644
`revision_id` may be None for the null revision, in which case
645
an `EmptyTree` is returned."""
647
if revision_id == None:
650
inv = self.get_revision_inventory(revision_id)
651
return RevisionTree(self.text_store, inv)
654
def working_tree(self):
655
"""Return a `Tree` for the working copy."""
656
return WorkingTree(self.base, self.read_working_inventory())
659
def basis_tree(self):
660
"""Return `Tree` object for last revision.
662
If there are no revisions yet, return an `EmptyTree`.
664
>>> b = ScratchBranch(files=['foo'])
665
>>> b.basis_tree().has_filename('foo')
667
>>> b.working_tree().has_filename('foo')
670
>>> b.commit('add foo')
671
>>> b.basis_tree().has_filename('foo')
674
r = self.last_patch()
678
return RevisionTree(self.text_store, self.get_revision_inventory(r))
682
def write_log(self, show_timezone='original'):
683
"""Write out human-readable log of commits to this branch
685
:param utc: If true, show dates in universal time, not local time."""
686
## TODO: Option to choose either original, utc or local timezone
689
for p in self.revision_history():
691
print 'revno:', revno
692
## TODO: Show hash if --id is given.
693
##print 'revision-hash:', p
694
rev = self.get_revision(p)
695
print 'committer:', rev.committer
696
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0,
699
## opportunistic consistency check, same as check_patch_chaining
700
if rev.precursor != precursor:
701
bailout("mismatched precursor!")
705
print ' (no message)'
707
for l in rev.message.split('\n'):
714
def rename_one(self, from_rel, to_rel):
715
tree = self.working_tree()
717
if not tree.has_filename(from_rel):
718
bailout("can't rename: old working file %r does not exist" % from_rel)
719
if tree.has_filename(to_rel):
720
bailout("can't rename: new working file %r already exists" % to_rel)
722
file_id = inv.path2id(from_rel)
724
bailout("can't rename: old name %r is not versioned" % from_rel)
726
if inv.path2id(to_rel):
727
bailout("can't rename: new name %r is already versioned" % to_rel)
729
to_dir, to_tail = os.path.split(to_rel)
730
to_dir_id = inv.path2id(to_dir)
731
if to_dir_id == None and to_dir != '':
732
bailout("can't determine destination directory id for %r" % to_dir)
734
mutter("rename_one:")
735
mutter(" file_id {%s}" % file_id)
736
mutter(" from_rel %r" % from_rel)
737
mutter(" to_rel %r" % to_rel)
738
mutter(" to_dir %r" % to_dir)
739
mutter(" to_dir_id {%s}" % to_dir_id)
741
inv.rename(file_id, to_dir_id, to_tail)
743
print "%s => %s" % (from_rel, to_rel)
745
from_abs = self.abspath(from_rel)
746
to_abs = self.abspath(to_rel)
748
os.rename(from_abs, to_abs)
750
bailout("failed to rename %r to %r: %s"
751
% (from_abs, to_abs, e[1]),
752
["rename rolled back"])
754
self._write_inventory(inv)
758
def move(self, from_paths, to_name):
761
to_name must exist as a versioned directory.
763
If to_name exists and is a directory, the files are moved into
764
it, keeping their old names. If it is a directory,
766
Note that to_name is only the last component of the new name;
767
this doesn't change the directory.
769
## TODO: Option to move IDs only
770
assert not isinstance(from_paths, basestring)
771
tree = self.working_tree()
773
to_abs = self.abspath(to_name)
774
if not isdir(to_abs):
775
bailout("destination %r is not a directory" % to_abs)
776
if not tree.has_filename(to_name):
777
bailout("destination %r not in working directory" % to_abs)
778
to_dir_id = inv.path2id(to_name)
779
if to_dir_id == None and to_name != '':
780
bailout("destination %r is not a versioned directory" % to_name)
781
to_dir_ie = inv[to_dir_id]
782
if to_dir_ie.kind not in ('directory', 'root_directory'):
783
bailout("destination %r is not a directory" % to_abs)
785
to_idpath = Set(inv.get_idpath(to_dir_id))
788
if not tree.has_filename(f):
789
bailout("%r does not exist in working tree" % f)
790
f_id = inv.path2id(f)
792
bailout("%r is not versioned" % f)
793
name_tail = splitpath(f)[-1]
794
dest_path = appendpath(to_name, name_tail)
795
if tree.has_filename(dest_path):
796
bailout("destination %r already exists" % dest_path)
797
if f_id in to_idpath:
798
bailout("can't move %r to a subdirectory of itself" % f)
800
# OK, so there's a race here, it's possible that someone will
801
# create a file in this interval and then the rename might be
802
# left half-done. But we should have caught most problems.
805
name_tail = splitpath(f)[-1]
806
dest_path = appendpath(to_name, name_tail)
807
print "%s => %s" % (f, dest_path)
808
inv.rename(inv.path2id(f), to_dir_id, name_tail)
810
os.rename(self.abspath(f), self.abspath(dest_path))
812
bailout("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
813
["rename rolled back"])
815
self._write_inventory(inv)
819
def show_status(branch, show_all=False):
820
"""Display single-line status for non-ignored working files.
822
The list is show sorted in order by file name.
824
>>> b = ScratchBranch(files=['foo', 'foo~'])
830
>>> b.commit("add foo")
832
>>> os.unlink(b.abspath('foo'))
837
:todo: Get state for single files.
839
:todo: Perhaps show a slash at the end of directory names.
843
# We have to build everything into a list first so that it can
844
# sorted by name, incorporating all the different sources.
846
# FIXME: Rather than getting things in random order and then sorting,
847
# just step through in order.
849
# Interesting case: the old ID for a file has been removed,
850
# but a new file has been created under that name.
852
old = branch.basis_tree()
853
old_inv = old.inventory
854
new = branch.working_tree()
855
new_inv = new.inventory
857
for fs, fid, oldname, newname, kind in diff_trees(old, new):
859
show_status(fs, kind,
860
oldname + ' => ' + newname)
861
elif fs == 'A' or fs == 'M':
862
show_status(fs, kind, newname)
864
show_status(fs, kind, oldname)
867
show_status(fs, kind, newname)
870
show_status(fs, kind, newname)
872
show_status(fs, kind, newname)
874
bailout("wierd file state %r" % ((fs, fid),))
878
class ScratchBranch(Branch):
879
"""Special test class: a branch that cleans up after itself.
881
>>> b = ScratchBranch()
889
def __init__(self, files=[], dirs=[]):
890
"""Make a test branch.
892
This creates a temporary directory and runs init-tree in it.
894
If any files are listed, they are created in the working copy.
896
Branch.__init__(self, tempfile.mkdtemp(), init=True)
898
os.mkdir(self.abspath(d))
901
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
905
"""Destroy the test branch, removing the scratch directory."""
907
shutil.rmtree(self.base)
909
# Work around for shutil.rmtree failing on Windows when
910
# readonly files are encountered
911
for root, dirs, files in os.walk(self.base, topdown=False):
913
os.chmod(os.path.join(root, name), 0700)
914
shutil.rmtree(self.base)
918
######################################################################
922
def is_control_file(filename):
923
## FIXME: better check
924
filename = os.path.normpath(filename)
925
while filename != '':
926
head, tail = os.path.split(filename)
927
## mutter('check %r for control file' % ((head, tail), ))
928
if tail == bzrlib.BZRDIR:
937
def _gen_revision_id(when):
938
"""Return new revision-id."""
939
s = '%s-%s-' % (user_email(), compact_date(when))
940
s += hexlify(rand_bytes(12))
944
def gen_file_id(name):
945
"""Return new file id.
947
This should probably generate proper UUIDs, but for the moment we
948
cope with just randomness because running uuidgen every time is
950
idx = name.rfind('/')
952
name = name[idx+1 : ]
954
name = name.lstrip('.')
956
s = hexlify(rand_bytes(12))
957
return '-'.join((name, compact_date(time.time()), s))