2
# -*- coding: UTF-8 -*-
4
# This program is free software; you can redistribute it and/or modify
5
# it under the terms of the GNU General Public License as published by
6
# the Free Software Foundation; either version 2 of the License, or
7
# (at your option) any later version.
9
# This program is distributed in the hope that it will be useful,
10
# but WITHOUT ANY WARRANTY; without even the implied warranty of
11
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12
# GNU General Public License for more details.
14
# You should have received a copy of the GNU General Public License
15
# along with this program; if not, write to the Free Software
16
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
22
import traceback, socket, fnmatch, difflib, time
23
from binascii import hexlify
26
from inventory import Inventory
27
from trace import mutter, note
28
from tree import Tree, EmptyTree, RevisionTree, WorkingTree
29
from inventory import InventoryEntry, Inventory
30
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, chomp, \
31
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
32
joinpath, sha_string, file_kind, local_time_offset
33
from store import ImmutableStore
34
from revision import Revision
35
from errors import bailout
36
from textui import show_status
37
from diff import diff_trees
39
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
40
## TODO: Maybe include checks for common corruption of newlines, etc?
46
######################################################################
50
"""Branch holding a history of revisions.
52
:todo: Perhaps use different stores for different classes of object,
53
so that we can keep track of how much space each one uses,
54
or garbage-collect them.
56
:todo: Add a RemoteBranch subclass. For the basic case of read-only
57
HTTP access this should be very easy by,
58
just redirecting controlfile access into HTTP requests.
59
We would need a RemoteStore working similarly.
61
:todo: Keep the on-disk branch locked while the object exists.
63
:todo: mkdir() method.
65
def __init__(self, base, init=False):
66
"""Create new branch object at a particular location.
68
:param base: Base directory for the branch.
70
:param init: If True, create new control files in a previously
71
unversioned directory. If False, the branch must already
74
In the test suite, creation of new trees is tested using the
75
`ScratchBranch` class.
77
self.base = os.path.realpath(base)
81
if not isdir(self.controlfilename('.')):
82
bailout("not a bzr branch: %s" % quotefn(base),
83
['use "bzr init" to initialize a new working tree',
84
'current bzr can only operate from top-of-tree'])
87
self.text_store = ImmutableStore(self.controlfilename('text-store'))
88
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
89
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
93
return '%s(%r)' % (self.__class__.__name__, self.base)
100
"""Return filename relative to branch top"""
101
return os.path.join(self.base, name)
104
def controlfilename(self, file_or_path):
105
"""Return location relative to branch."""
106
if isinstance(file_or_path, types.StringTypes):
107
file_or_path = [file_or_path]
108
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
111
def controlfile(self, file_or_path, mode='r'):
112
"""Open a control file for this branch"""
113
return file(self.controlfilename(file_or_path), mode)
116
def _make_control(self):
117
os.mkdir(self.controlfilename([]))
118
self.controlfile('README', 'w').write(
119
"This is a Bazaar-NG control directory.\n"
120
"Do not change any files in this directory.")
121
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
122
for d in ('text-store', 'inventory-store', 'revision-store'):
123
os.mkdir(self.controlfilename(d))
124
for f in ('revision-history', 'merged-patches',
125
'pending-merged-patches', 'branch-name'):
126
self.controlfile(f, 'w').write('')
127
mutter('created control directory in ' + self.base)
128
Inventory().write_xml(self.controlfile('inventory','w'))
131
def _check_format(self):
132
"""Check this branch format is supported.
134
The current tool only supports the current unstable format.
136
In the future, we might need different in-memory Branch
137
classes to support downlevel branches. But not yet.
139
# read in binary mode to detect newline wierdness.
140
fmt = self.controlfile('branch-format', 'rb').read()
141
if fmt != BZR_BRANCH_FORMAT:
142
bailout('sorry, branch format %r not supported' % fmt,
143
['use a different bzr version',
144
'or remove the .bzr directory and "bzr init" again'])
147
def read_working_inventory(self):
148
"""Read the working inventory."""
150
inv = Inventory.read_xml(self.controlfile('inventory', 'r'))
151
mutter("loaded inventory of %d items in %f"
152
% (len(inv), time.time() - before))
156
def _write_inventory(self, inv):
157
"""Update the working inventory.
159
That is to say, the inventory describing changes underway, that
160
will be committed to the next revision.
162
inv.write_xml(self.controlfile('inventory', 'w'))
163
mutter('wrote inventory to %s' % quotefn(self.controlfilename('inventory')))
166
inventory = property(read_working_inventory, _write_inventory, None,
167
"""Inventory for the working copy.""")
170
def add(self, files, verbose=False):
171
"""Make files versioned.
173
This puts the files in the Added state, so that they will be
174
recorded by the next commit.
176
:todo: Perhaps have an option to add the ids even if the files do
179
:todo: Perhaps return the ids of the files? But then again it
180
is easy to retrieve them if they're needed.
182
:todo: Option to specify file id.
184
:todo: Adding a directory should optionally recurse down and
185
add all non-ignored children. Perhaps do that in a
188
>>> b = ScratchBranch(files=['foo'])
189
>>> 'foo' in b.unknowns()
194
>>> 'foo' in b.unknowns()
196
>>> bool(b.inventory.path2id('foo'))
202
Traceback (most recent call last):
204
BzrError: ('foo is already versioned', [])
206
>>> b.add(['nothere'])
207
Traceback (most recent call last):
208
BzrError: ('cannot add: not a regular file or directory: nothere', [])
211
# TODO: Re-adding a file that is removed in the working copy
212
# should probably put it back with the previous ID.
213
if isinstance(files, types.StringTypes):
216
inv = self.read_working_inventory()
218
if is_control_file(f):
219
bailout("cannot add control file %s" % quotefn(f))
224
bailout("cannot add top-level %r" % f)
226
fullpath = os.path.normpath(self._rel(f))
230
elif isdir(fullpath):
233
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
236
parent_name = joinpath(fp[:-1])
237
mutter("lookup parent %r" % parent_name)
238
parent_id = inv.path2id(parent_name)
239
if parent_id == None:
240
bailout("cannot add: parent %r is not versioned"
245
file_id = _gen_file_id(fp[-1])
246
inv.add(InventoryEntry(file_id, fp[-1], kind=kind, parent_id=parent_id))
248
show_status('A', kind, quotefn(f))
250
mutter("add file %s file_id:{%s} kind=%r parent_id={%s}"
251
% (f, file_id, kind, parent_id))
252
self._write_inventory(inv)
256
def remove(self, files, verbose=False):
257
"""Mark nominated files for removal from the inventory.
259
This does not remove their text. This does not run on
261
:todo: Refuse to remove modified files unless --force is given?
263
>>> b = ScratchBranch(files=['foo'])
265
>>> b.inventory.has_filename('foo')
268
>>> b.working_tree().has_filename('foo')
270
>>> b.inventory.has_filename('foo')
273
>>> b = ScratchBranch(files=['foo'])
278
>>> b.inventory.has_filename('foo')
280
>>> b.basis_tree().has_filename('foo')
282
>>> b.working_tree().has_filename('foo')
285
:todo: Do something useful with directories.
287
:todo: Should this remove the text or not? Tough call; not
288
removing may be useful and the user can just use use rm, and
289
is the opposite of add. Removing it is consistent with most
290
other tools. Maybe an option.
292
## TODO: Normalize names
293
## TODO: Remove nested loops; better scalability
295
if isinstance(files, types.StringTypes):
298
inv = self.read_working_inventory()
300
# do this before any modifications
304
bailout("cannot remove unversioned file %s" % quotefn(f))
305
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
307
show_status('D', inv[fid].kind, quotefn(f))
310
self._write_inventory(inv)
314
"""Return all unknown files.
316
These are files in the working directory that are not versioned or
317
control files or ignored.
319
>>> b = ScratchBranch(files=['foo', 'foo~'])
320
>>> list(b.unknowns())
323
>>> list(b.unknowns())
326
>>> list(b.unknowns())
329
return self.working_tree().unknowns()
332
def commit(self, message, timestamp=None, timezone=None,
335
"""Commit working copy as a new revision.
337
The basic approach is to add all the file texts into the
338
store, then the inventory, then make a new revision pointing
339
to that inventory and store that.
341
This is not quite safe if the working copy changes during the
342
commit; for the moment that is simply not allowed. A better
343
approach is to make a temporary copy of the files before
344
computing their hashes, and then add those hashes in turn to
345
the inventory. This should mean at least that there are no
346
broken hash pointers. There is no way we can get a snapshot
347
of the whole directory at an instant. This would also have to
348
be robust against files disappearing, moving, etc. So the
349
whole thing is a bit hard.
351
:param timestamp: if not None, seconds-since-epoch for a
352
postdated/predated commit.
355
## TODO: Show branch names
357
# TODO: Don't commit if there are no changes, unless forced?
359
# First walk over the working inventory; and both update that
360
# and also build a new revision inventory. The revision
361
# inventory needs to hold the text-id, sha1 and size of the
362
# actual file versions committed in the revision. (These are
363
# not present in the working inventory.) We also need to
364
# detect missing/deleted files, and remove them from the
367
work_inv = self.read_working_inventory()
369
basis = self.basis_tree()
370
basis_inv = basis.inventory
372
for path, entry in work_inv.iter_entries():
373
## TODO: Cope with files that have gone missing.
375
## TODO: Check that the file kind has not changed from the previous
376
## revision of this file (if any).
381
file_id = entry.file_id
382
mutter('commit prep file %s, id %r ' % (p, file_id))
384
if not os.path.exists(p):
385
mutter(" file is missing, removing from inventory")
387
show_status('D', entry.kind, quotefn(path))
388
missing_ids.append(file_id)
391
# TODO: Handle files that have been deleted
393
# TODO: Maybe a special case for empty files? Seems a
394
# waste to store them many times.
398
if basis_inv.has_id(file_id):
399
old_kind = basis_inv[file_id].kind
400
if old_kind != entry.kind:
401
bailout("entry %r changed kind from %r to %r"
402
% (file_id, old_kind, entry.kind))
404
if entry.kind == 'directory':
406
bailout("%s is entered as directory but not a directory" % quotefn(p))
407
elif entry.kind == 'file':
409
bailout("%s is entered as file but is not a file" % quotefn(p))
411
content = file(p, 'rb').read()
413
entry.text_sha1 = sha_string(content)
414
entry.text_size = len(content)
416
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
418
and (old_ie.text_size == entry.text_size)
419
and (old_ie.text_sha1 == entry.text_sha1)):
420
## assert content == basis.get_file(file_id).read()
421
entry.text_id = basis_inv[file_id].text_id
422
mutter(' unchanged from previous text_id {%s}' %
426
entry.text_id = _gen_file_id(entry.name)
427
self.text_store.add(content, entry.text_id)
428
mutter(' stored with text_id {%s}' % entry.text_id)
432
elif (old_ie.name == entry.name
433
and old_ie.parent_id == entry.parent_id):
438
show_status(state, entry.kind, quotefn(path))
440
for file_id in missing_ids:
441
# have to do this later so we don't mess up the iterator.
442
# since parents may be removed before their children we
445
# FIXME: There's probably a better way to do this; perhaps
446
# the workingtree should know how to filter itself.
447
if work_inv.has_id(file_id):
448
del work_inv[file_id]
451
inv_id = rev_id = _gen_revision_id(time.time())
453
inv_tmp = tempfile.TemporaryFile()
454
inv.write_xml(inv_tmp)
456
self.inventory_store.add(inv_tmp, inv_id)
457
mutter('new inventory_id is {%s}' % inv_id)
459
self._write_inventory(work_inv)
461
if timestamp == None:
462
timestamp = time.time()
464
if committer == None:
465
committer = username()
468
timezone = local_time_offset()
470
mutter("building commit log message")
471
rev = Revision(timestamp=timestamp,
474
precursor = self.last_patch(),
479
rev_tmp = tempfile.TemporaryFile()
480
rev.write_xml(rev_tmp)
482
self.revision_store.add(rev_tmp, rev_id)
483
mutter("new revision_id is {%s}" % rev_id)
485
## XXX: Everything up to here can simply be orphaned if we abort
486
## the commit; it will leave junk files behind but that doesn't
489
## TODO: Read back the just-generated changeset, and make sure it
490
## applies and recreates the right state.
492
## TODO: Also calculate and store the inventory SHA1
493
mutter("committing patch r%d" % (self.revno() + 1))
495
mutter("append to revision-history")
496
self.controlfile('revision-history', 'at').write(rev_id + '\n')
501
def get_revision(self, revision_id):
502
"""Return the Revision object for a named revision"""
503
r = Revision.read_xml(self.revision_store[revision_id])
504
assert r.revision_id == revision_id
508
def get_inventory(self, inventory_id):
509
"""Get Inventory object by hash.
511
:todo: Perhaps for this and similar methods, take a revision
512
parameter which can be either an integer revno or a
514
i = Inventory.read_xml(self.inventory_store[inventory_id])
518
def get_revision_inventory(self, revision_id):
519
"""Return inventory of a past revision."""
520
if revision_id == None:
523
return self.get_inventory(self.get_revision(revision_id).inventory_id)
526
def revision_history(self):
527
"""Return sequence of revision hashes on to this branch.
529
>>> ScratchBranch().revision_history()
532
return [chomp(l) for l in self.controlfile('revision-history').readlines()]
536
"""Return current revision number for this branch.
538
That is equivalent to the number of revisions committed to
541
>>> b = ScratchBranch()
544
>>> b.commit('no foo')
548
return len(self.revision_history())
551
def last_patch(self):
552
"""Return last patch hash, or None if no history.
554
>>> ScratchBranch().last_patch() == None
557
ph = self.revision_history()
562
def lookup_revision(self, revno):
563
"""Return revision hash for revision number."""
568
# list is 0-based; revisions are 1-based
569
return self.revision_history()[revno-1]
571
bailout("no such revision %s" % revno)
574
def revision_tree(self, revision_id):
575
"""Return Tree for a revision on this branch.
577
`revision_id` may be None for the null revision, in which case
578
an `EmptyTree` is returned."""
580
if revision_id == None:
583
inv = self.get_revision_inventory(revision_id)
584
return RevisionTree(self.text_store, inv)
587
def working_tree(self):
588
"""Return a `Tree` for the working copy."""
589
return WorkingTree(self.base, self.read_working_inventory())
592
def basis_tree(self):
593
"""Return `Tree` object for last revision.
595
If there are no revisions yet, return an `EmptyTree`.
597
>>> b = ScratchBranch(files=['foo'])
598
>>> b.basis_tree().has_filename('foo')
600
>>> b.working_tree().has_filename('foo')
603
>>> b.commit('add foo')
604
>>> b.basis_tree().has_filename('foo')
607
r = self.last_patch()
611
return RevisionTree(self.text_store, self.get_revision_inventory(r))
615
def write_log(self, utc=False):
616
"""Write out human-readable log of commits to this branch
618
:param utc: If true, show dates in universal time, not local time."""
619
## TODO: Option to choose either original, utc or local timezone
622
for p in self.revision_history():
624
print 'revno:', revno
625
## TODO: Show hash if --id is given.
626
##print 'revision-hash:', p
627
rev = self.get_revision(p)
628
print 'committer:', rev.committer
629
print 'timestamp: %s' % (format_date(rev.timestamp, rev.timezone or 0))
631
## opportunistic consistency check, same as check_patch_chaining
632
if rev.precursor != precursor:
633
bailout("mismatched precursor!")
637
print ' (no message)'
639
for l in rev.message.split('\n'):
647
def show_status(branch, show_all=False):
648
"""Display single-line status for non-ignored working files.
650
The list is show sorted in order by file name.
652
>>> b = ScratchBranch(files=['foo', 'foo~'])
658
>>> b.commit("add foo")
661
:todo: Get state for single files.
663
:todo: Perhaps show a slash at the end of directory names.
667
# We have to build everything into a list first so that it can
668
# sorted by name, incorporating all the different sources.
670
# FIXME: Rather than getting things in random order and then sorting,
671
# just step through in order.
673
# Interesting case: the old ID for a file has been removed,
674
# but a new file has been created under that name.
676
old = branch.basis_tree()
677
old_inv = old.inventory
678
new = branch.working_tree()
679
new_inv = new.inventory
681
for fs, fid, oldname, newname, kind in diff_trees(old, new):
683
show_status(fs, kind,
684
oldname + ' => ' + newname)
685
elif fs == 'A' or fs == 'M':
686
show_status(fs, kind, newname)
688
show_status(fs, kind, oldname)
691
show_status(fs, kind, newname)
694
show_status(fs, kind, newname)
696
show_status(fs, kind, newname)
698
bailout("wierd file state %r" % ((fs, fid),))
702
class ScratchBranch(Branch):
703
"""Special test class: a branch that cleans up after itself.
705
>>> b = ScratchBranch()
713
def __init__(self, files = []):
714
"""Make a test branch.
716
This creates a temporary directory and runs init-tree in it.
718
If any files are listed, they are created in the working copy.
720
Branch.__init__(self, tempfile.mkdtemp(), init=True)
722
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
726
"""Destroy the test branch, removing the scratch directory."""
727
shutil.rmtree(self.base)
731
######################################################################
735
def is_control_file(filename):
736
## FIXME: better check
737
filename = os.path.normpath(filename)
738
while filename != '':
739
head, tail = os.path.split(filename)
740
## mutter('check %r for control file' % ((head, tail), ))
741
if tail == bzrlib.BZRDIR:
748
def _gen_revision_id(when):
749
"""Return new revision-id."""
750
s = '%s-%s-' % (user_email(), compact_date(when))
751
s += hexlify(rand_bytes(8))
755
def _gen_file_id(name):
756
"""Return new file id.
758
This should probably generate proper UUIDs, but for the moment we
759
cope with just randomness because running uuidgen every time is
761
assert '/' not in name
762
while name[0] == '.':
764
s = hexlify(rand_bytes(8))
765
return '-'.join((name, compact_date(time.time()), s))