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
17
"""Experiment in converting existing bzr branches to weaves."""
19
# To make this properly useful
21
# 1. assign text version ids, and put those text versions into
22
# the inventory as they're converted.
24
# 2. keep track of the previous version of each file, rather than
25
# just using the last one imported
27
# 3. assign entry versions when files are added, renamed or moved.
29
# 4. when merged-in versions are observed, walk down through them
30
# to discover everything, then commit bottom-up
32
# 5. track ancestry as things are merged in, and commit that in each
35
# Perhaps it's best to first walk the whole graph and make a plan for
36
# what should be imported in what order? Need a kind of topological
37
# sort of all revisions. (Or do we, can we just before doing a revision
38
# see that all its parents have either been converted or abandoned?)
41
# Cannot import a revision until all its parents have been
42
# imported. in other words, we can only import revisions whose
43
# parents have all been imported. the first step must be to
44
# import a revision with no parents, of which there must be at
45
# least one. (So perhaps it's useful to store forward pointers
46
# from a list of parents to their children?)
48
# Another (equivalent?) approach is to build up the ordered
49
# ancestry list for the last revision, and walk through that. We
50
# are going to need that.
52
# We don't want to have to recurse all the way back down the list.
54
# Suppose we keep a queue of the revisions able to be processed at
55
# any point. This starts out with all the revisions having no
58
# This seems like a generally useful algorithm...
60
# The current algorithm is dumb (O(n**2)?) but will do the job, and
61
# takes less than a second on the bzr.dev branch.
63
# This currently does a kind of lazy conversion of file texts, where a
64
# new text is written in every version. That's unnecessary but for
65
# the moment saves us having to worry about when files need new
74
from bzrlib.branch import Branch, find_branch
75
from bzrlib.branch import BZR_BRANCH_FORMAT_5, BZR_BRANCH_FORMAT_6
76
import bzrlib.hashcache as hashcache
77
from bzrlib.weave import Weave
78
from bzrlib.weavefile import read_weave, write_weave
79
from bzrlib.ui import ui_factory
80
from bzrlib.atomicfile import AtomicFile
81
from bzrlib.xml4 import serializer_v4
82
from bzrlib.xml5 import serializer_v5
83
from bzrlib.trace import mutter, note, warning
84
from bzrlib.osutils import sha_strings, sha_string, pathjoin, abspath
87
# TODO: jam 20060108 Create a new branch format, and as part of upgrade
88
# make sure that ancestry.weave is deleted (it is never used, but
92
class Convert(object):
93
def __init__(self, base_dir):
95
self.converted_revs = set()
96
self.absent_revisions = set()
103
if not self._open_branch():
105
note('starting upgrade of %s', os.path.abspath(self.base))
106
self._backup_control_dir()
107
self.pb = ui_factory.progress_bar()
108
if self.old_format == 4:
109
note('starting upgrade from format 4 to 5')
110
self._convert_to_weaves()
112
if self.old_format == 5:
113
note('starting upgrade from format 5 to 6')
114
self._convert_to_prefixed()
116
cache = hashcache.HashCache(abspath(self.base))
122
def _convert_to_prefixed(self):
123
from bzrlib.store import hash_prefix
124
for store_name in ["weaves", "revision-store"]:
125
note("adding prefixes to %s" % store_name)
126
store_dir = pathjoin(self.base, ".bzr", store_name)
127
for filename in os.listdir(store_dir):
128
if filename.endswith(".weave") or filename.endswith(".gz"):
129
file_id = os.path.splitext(filename)[0]
132
prefix_dir = pathjoin(store_dir, hash_prefix(file_id))
133
if not os.path.isdir(prefix_dir):
135
os.rename(pathjoin(store_dir, filename),
136
pathjoin(prefix_dir, filename))
137
self._set_new_format(BZR_BRANCH_FORMAT_6)
140
def _convert_to_weaves(self):
141
note('note: upgrade may be faster if all store files are ungzipped first')
142
if not os.path.isdir(self.base + '/.bzr/weaves'):
143
os.mkdir(self.base + '/.bzr/weaves')
144
self.inv_weave = Weave('inventory')
145
# holds in-memory weaves for all files
146
self.text_weaves = {}
147
os.remove(self.branch.controlfilename('branch-format'))
148
self._convert_working_inv()
149
rev_history = self.branch.revision_history()
150
# to_read is a stack holding the revisions we still need to process;
151
# appending to it adds new highest-priority revisions
152
self.known_revisions = set(rev_history)
153
self.to_read = rev_history[-1:]
155
rev_id = self.to_read.pop()
156
if (rev_id not in self.revisions
157
and rev_id not in self.absent_revisions):
158
self._load_one_rev(rev_id)
160
to_import = self._make_order()
161
for i, rev_id in enumerate(to_import):
162
self.pb.update('converting revision', i, len(to_import))
163
self._convert_one_rev(rev_id)
165
note('upgraded to weaves:')
166
note(' %6d revisions and inventories' % len(self.revisions))
167
note(' %6d revisions not present' % len(self.absent_revisions))
168
note(' %6d texts' % self.text_count)
169
self._write_all_weaves()
170
self._write_all_revs()
171
self._cleanup_spare_files()
172
self._set_new_format(BZR_BRANCH_FORMAT_5)
175
def _open_branch(self):
176
self.branch = Branch.open_downlevel(self.base)
177
self.old_format = self.branch._branch_format
178
if self.old_format == 6:
179
note('this branch is in the most current format')
181
if self.old_format not in (4, 5):
182
raise BzrError("cannot upgrade from branch format %r" %
183
self.branch._branch_format)
187
def _set_new_format(self, format):
188
self.branch.put_controlfile('branch-format', format)
191
def _cleanup_spare_files(self):
192
for n in 'merged-patches', 'pending-merged-patches':
193
p = self.branch.controlfilename(n)
194
if not os.path.exists(p):
196
## assert os.path.getsize(p) == 0
198
shutil.rmtree(self.base + '/.bzr/inventory-store')
199
shutil.rmtree(self.base + '/.bzr/text-store')
202
def _backup_control_dir(self):
203
orig = self.base + '/.bzr'
204
backup = orig + '.backup'
205
note('making backup of tree history')
206
shutil.copytree(orig, backup)
207
note('%s has been backed up to %s', orig, backup)
208
note('if conversion fails, you can move this directory back to .bzr')
209
note('if it succeeds, you can remove this directory if you wish')
212
def _convert_working_inv(self):
214
inv = serializer_v4.read_inventory(branch.controlfile('inventory', 'rb'))
215
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
216
branch.put_controlfile('inventory', new_inv_xml)
220
def _write_all_weaves(self):
221
write_a_weave(self.inv_weave, self.base + '/.bzr/inventory.weave')
224
for file_id, file_weave in self.text_weaves.items():
225
self.pb.update('writing weave', i, len(self.text_weaves))
226
write_a_weave(file_weave, self.base + '/.bzr/weaves/%s.weave' % file_id)
232
def _write_all_revs(self):
233
"""Write all revisions out in new form."""
234
shutil.rmtree(self.base + '/.bzr/revision-store')
235
os.mkdir(self.base + '/.bzr/revision-store')
237
for i, rev_id in enumerate(self.converted_revs):
238
self.pb.update('write revision', i, len(self.converted_revs))
239
f = file(self.base + '/.bzr/revision-store/%s' % rev_id, 'wb')
241
serializer_v5.write_revision(self.revisions[rev_id], f)
248
def _load_one_rev(self, rev_id):
249
"""Load a revision object into memory.
251
Any parents not either loaded or abandoned get queued to be
253
self.pb.update('loading revision',
255
len(self.known_revisions))
256
if not self.branch.revision_store.has_id(rev_id):
258
note('revision {%s} not present in branch; '
259
'will be converted as a ghost',
261
self.absent_revisions.add(rev_id)
263
rev_xml = self.branch.revision_store.get(rev_id).read()
264
rev = serializer_v4.read_revision_from_string(rev_xml)
265
for parent_id in rev.parent_ids:
266
self.known_revisions.add(parent_id)
267
self.to_read.append(parent_id)
268
self.revisions[rev_id] = rev
271
def _load_old_inventory(self, rev_id):
272
assert rev_id not in self.converted_revs
273
old_inv_xml = self.branch.inventory_store.get(rev_id).read()
274
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
275
rev = self.revisions[rev_id]
276
if rev.inventory_sha1:
277
assert rev.inventory_sha1 == sha_string(old_inv_xml), \
278
'inventory sha mismatch for {%s}' % rev_id
282
def _load_updated_inventory(self, rev_id):
283
assert rev_id in self.converted_revs
284
inv_xml = self.inv_weave.get_text(rev_id)
285
inv = serializer_v5.read_inventory_from_string(inv_xml)
289
def _convert_one_rev(self, rev_id):
290
"""Convert revision and all referenced objects to new format."""
291
rev = self.revisions[rev_id]
292
inv = self._load_old_inventory(rev_id)
293
present_parents = [p for p in rev.parent_ids
294
if p not in self.absent_revisions]
295
self._convert_revision_contents(rev, inv, present_parents)
296
self._store_new_weave(rev, inv, present_parents)
297
self.converted_revs.add(rev_id)
300
def _store_new_weave(self, rev, inv, present_parents):
301
# the XML is now updated with text versions
305
if ie.kind == 'root_directory':
307
assert hasattr(ie, 'revision'), \
308
'no revision on {%s} in {%s}' % \
309
(file_id, rev.revision_id)
310
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
311
new_inv_sha1 = sha_string(new_inv_xml)
312
self.inv_weave.add(rev.revision_id,
314
new_inv_xml.splitlines(True),
316
rev.inventory_sha1 = new_inv_sha1
318
def _convert_revision_contents(self, rev, inv, present_parents):
319
"""Convert all the files within a revision.
321
Also upgrade the inventory to refer to the text revision ids."""
322
rev_id = rev.revision_id
323
mutter('converting texts of revision {%s}',
325
parent_invs = map(self._load_updated_inventory, present_parents)
328
self._convert_file_version(rev, ie, parent_invs)
330
def _convert_file_version(self, rev, ie, parent_invs):
331
"""Convert one version of one file.
333
The file needs to be added into the weave if it is a merge
334
of >=2 parents or if it's changed from its parent.
336
if ie.kind == 'root_directory':
339
rev_id = rev.revision_id
340
w = self.text_weaves.get(file_id)
343
self.text_weaves[file_id] = w
345
previous_entries = ie.find_previous_heads(parent_invs, w)
346
for old_revision in previous_entries:
347
# if this fails, its a ghost ?
348
assert old_revision in self.converted_revs
349
self.snapshot_ie(previous_entries, ie, w, rev_id)
351
assert getattr(ie, 'revision', None) is not None
353
def snapshot_ie(self, previous_revisions, ie, w, rev_id):
354
# TODO: convert this logic, which is ~= snapshot to
355
# a call to:. This needs the path figured out. rather than a work_tree
356
# a v4 revision_tree can be given, or something that looks enough like
357
# one to give the file content to the entry if it needs it.
358
# and we need something that looks like a weave store for snapshot to
360
#ie.snapshot(rev, PATH, previous_revisions, REVISION_TREE, InMemoryWeaveStore(self.text_weaves))
361
if len(previous_revisions) == 1:
362
previous_ie = previous_revisions.values()[0]
363
if ie._unchanged(previous_ie):
364
ie.revision = previous_ie.revision
366
parent_indexes = map(w.lookup, previous_revisions)
368
file_lines = self.branch.text_store.get(ie.text_id).readlines()
369
assert sha_strings(file_lines) == ie.text_sha1
370
assert sum(map(len, file_lines)) == ie.text_size
371
w.add(rev_id, parent_indexes, file_lines, ie.text_sha1)
374
w.add(rev_id, parent_indexes, [], None)
376
##mutter('import text {%s} of {%s}',
377
## ie.text_id, file_id)
379
def _make_order(self):
380
"""Return a suitable order for importing revisions.
382
The order must be such that an revision is imported after all
383
its (present) parents.
385
todo = set(self.revisions.keys())
386
done = self.absent_revisions.copy()
389
# scan through looking for a revision whose parents
391
for rev_id in sorted(list(todo)):
392
rev = self.revisions[rev_id]
393
parent_ids = set(rev.parent_ids)
394
if parent_ids.issubset(done):
395
# can take this one now
402
def write_a_weave(weave, filename):
403
inv_wf = file(filename, 'wb')
405
write_weave(weave, inv_wf)
410
def upgrade(base_dir):