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
68
from cStringIO import StringIO
75
from bzrlib.branch import Branch
76
from bzrlib.branch import BZR_BRANCH_FORMAT_5, BZR_BRANCH_FORMAT_6
77
from bzrlib.branch import BzrBranchFormat, BzrBranchFormat4, BzrBranchFormat5, BzrBranchFormat6
78
from bzrlib.errors import NoSuchFile, UpgradeReadonly
79
import bzrlib.hashcache as hashcache
80
from bzrlib.lockable_files import LockableFiles
81
from bzrlib.osutils import sha_strings, sha_string, pathjoin, abspath
82
from bzrlib.ui import ui_factory
83
from bzrlib.store.text import TextStore
84
from bzrlib.store.weave import WeaveStore
85
from bzrlib.trace import mutter, note, warning
86
from bzrlib.transactions import PassThroughTransaction
87
from bzrlib.transport import get_transport
88
from bzrlib.transport.local import LocalTransport
89
from bzrlib.weave import Weave
90
from bzrlib.weavefile import read_weave, write_weave
91
from bzrlib.xml4 import serializer_v4
92
from bzrlib.xml5 import serializer_v5
95
class Convert(object):
97
def __init__(self, transport):
98
self.base = transport.base
99
self.converted_revs = set()
100
self.absent_revisions = set()
103
self.transport = transport
104
if self.transport.is_readonly():
105
raise UpgradeReadonly
106
self.control_files = LockableFiles(transport.clone(bzrlib.BZRDIR), 'branch-lock')
107
# Lock the branch (soon to be meta dir) to prevent anyone racing with us
108
# This is currently windows incompatible, it will deadlock. When the upgrade
109
# logic becomes format specific, then we can have the format know how to pass this
110
# on. Also note that we probably have an 'upgrade meta' which upgrades the constituent
112
print "FIXME: control files reuse"
113
self.control_files.lock_write()
117
self.control_files.unlock()
120
if not self._open_branch():
122
note('starting upgrade of %s', self.base)
123
self._backup_control_dir()
124
self.pb = ui_factory.progress_bar()
125
if isinstance(self.old_format, BzrBranchFormat4):
126
note('starting upgrade from format 4 to 5')
127
self._convert_to_weaves()
128
if isinstance(self.old_format, BzrBranchFormat5):
129
note('starting upgrade from format 5 to 6')
130
self._convert_to_prefixed()
131
if isinstance(self.transport, LocalTransport):
132
cache = hashcache.HashCache(abspath(self.base))
137
def _convert_to_prefixed(self):
138
from bzrlib.store import hash_prefix
139
bzr_transport = self.transport.clone('.bzr')
140
bzr_transport.delete('branch-format')
141
for store_name in ["weaves", "revision-store"]:
142
note("adding prefixes to %s" % store_name)
143
store_transport = bzr_transport.clone(store_name)
144
for filename in store_transport.list_dir('.'):
145
if (filename.endswith(".weave") or
146
filename.endswith(".gz") or
147
filename.endswith(".sig")):
148
file_id = os.path.splitext(filename)[0]
151
prefix_dir = hash_prefix(file_id)
152
# FIXME keep track of the dirs made RBC 20060121
154
store_transport.move(filename, prefix_dir + '/' + filename)
155
except NoSuchFile: # catches missing dirs strangely enough
156
store_transport.mkdir(prefix_dir)
157
store_transport.move(filename, prefix_dir + '/' + filename)
158
self._set_new_format(BZR_BRANCH_FORMAT_6)
159
self.branch = BzrBranchFormat6().open(self.transport)
160
self.old_format = self.branch._branch_format
162
def _convert_to_weaves(self):
163
note('note: upgrade may be faster if all store files are ungzipped first')
164
bzr_transport = self.transport.clone('.bzr')
167
stat = bzr_transport.stat('weaves')
168
if not S_ISDIR(stat.st_mode):
169
bzr_transport.delete('weaves')
170
bzr_transport.mkdir('weaves')
172
bzr_transport.mkdir('weaves')
173
self.inv_weave = Weave('inventory')
174
# holds in-memory weaves for all files
175
self.text_weaves = {}
176
bzr_transport.delete('branch-format')
177
self._convert_working_inv()
178
rev_history = self.branch.revision_history()
179
# to_read is a stack holding the revisions we still need to process;
180
# appending to it adds new highest-priority revisions
181
self.known_revisions = set(rev_history)
182
self.to_read = rev_history[-1:]
184
rev_id = self.to_read.pop()
185
if (rev_id not in self.revisions
186
and rev_id not in self.absent_revisions):
187
self._load_one_rev(rev_id)
189
to_import = self._make_order()
190
for i, rev_id in enumerate(to_import):
191
self.pb.update('converting revision', i, len(to_import))
192
self._convert_one_rev(rev_id)
194
self._write_all_weaves()
195
self._write_all_revs()
196
note('upgraded to weaves:')
197
note(' %6d revisions and inventories' % len(self.revisions))
198
note(' %6d revisions not present' % len(self.absent_revisions))
199
note(' %6d texts' % self.text_count)
200
self._cleanup_spare_files_after_format4()
201
self._set_new_format(BZR_BRANCH_FORMAT_5)
202
self.branch = BzrBranchFormat5().open(self.transport)
203
self.old_format = self.branch._branch_format
205
def _open_branch(self):
206
self.old_format = BzrBranchFormat.find_format(self.transport)
207
self.branch = self.old_format.open(self.transport)
208
if isinstance(self.old_format, BzrBranchFormat6):
209
note('this branch is in the most current format (%s)', self.old_format)
211
if (not isinstance(self.old_format, BzrBranchFormat4) and
212
not isinstance(self.old_format, BzrBranchFormat5)):
213
raise BzrError("cannot upgrade from branch format %s" %
214
self.branch._branch_format)
217
def _set_new_format(self, format):
218
self.branch.control_files.put_utf8('branch-format', format)
220
def _cleanup_spare_files_after_format4(self):
221
transport = self.transport.clone('.bzr')
222
print "FIXME working tree upgrade foo."
223
for n in 'merged-patches', 'pending-merged-patches':
225
## assert os.path.getsize(p) == 0
229
transport.delete_tree('inventory-store')
230
transport.delete_tree('text-store')
232
def _backup_control_dir(self):
233
note('making backup of tree history')
234
self.transport.copy_tree('.bzr', '.bzr.backup')
235
note('%s.bzr has been backed up to %s.bzr.backup',
238
note('if conversion fails, you can move this directory back to .bzr')
239
note('if it succeeds, you can remove this directory if you wish')
241
def _convert_working_inv(self):
243
inv = serializer_v4.read_inventory(branch.control_files.get('inventory'))
244
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
245
print "fixme inventory is a working tree change."
246
branch.control_files.put('inventory', new_inv_xml)
248
def _write_all_weaves(self):
249
bzr_transport = self.transport.clone('.bzr')
250
controlweaves = WeaveStore(bzr_transport, prefixed=False)
251
weave_transport = bzr_transport.clone('weaves')
252
weaves = WeaveStore(weave_transport, prefixed=False)
253
transaction = PassThroughTransaction()
255
controlweaves.put_weave('inventory', self.inv_weave, transaction)
258
for file_id, file_weave in self.text_weaves.items():
259
self.pb.update('writing weave', i, len(self.text_weaves))
260
weaves.put_weave(file_id, file_weave, transaction)
265
def _write_all_revs(self):
266
"""Write all revisions out in new form."""
267
transport = self.transport.clone('.bzr')
268
transport.delete_tree('revision-store')
269
transport.mkdir('revision-store')
270
revision_transport = transport.clone('revision-store')
272
revision_store = TextStore(revision_transport,
276
for i, rev_id in enumerate(self.converted_revs):
277
self.pb.update('write revision', i, len(self.converted_revs))
279
serializer_v5.write_revision(self.revisions[rev_id], rev_tmp)
281
revision_store.add(rev_tmp, rev_id)
286
def _load_one_rev(self, rev_id):
287
"""Load a revision object into memory.
289
Any parents not either loaded or abandoned get queued to be
291
self.pb.update('loading revision',
293
len(self.known_revisions))
294
if not self.branch.repository.revision_store.has_id(rev_id):
296
note('revision {%s} not present in branch; '
297
'will be converted as a ghost',
299
self.absent_revisions.add(rev_id)
301
rev_xml = self.branch.repository.revision_store.get(rev_id).read()
302
rev = serializer_v4.read_revision_from_string(rev_xml)
303
for parent_id in rev.parent_ids:
304
self.known_revisions.add(parent_id)
305
self.to_read.append(parent_id)
306
self.revisions[rev_id] = rev
309
def _load_old_inventory(self, rev_id):
310
assert rev_id not in self.converted_revs
311
old_inv_xml = self.branch.repository.inventory_store.get(rev_id).read()
312
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
313
rev = self.revisions[rev_id]
314
if rev.inventory_sha1:
315
assert rev.inventory_sha1 == sha_string(old_inv_xml), \
316
'inventory sha mismatch for {%s}' % rev_id
320
def _load_updated_inventory(self, rev_id):
321
assert rev_id in self.converted_revs
322
inv_xml = self.inv_weave.get_text(rev_id)
323
inv = serializer_v5.read_inventory_from_string(inv_xml)
327
def _convert_one_rev(self, rev_id):
328
"""Convert revision and all referenced objects to new format."""
329
rev = self.revisions[rev_id]
330
inv = self._load_old_inventory(rev_id)
331
present_parents = [p for p in rev.parent_ids
332
if p not in self.absent_revisions]
333
self._convert_revision_contents(rev, inv, present_parents)
334
self._store_new_weave(rev, inv, present_parents)
335
self.converted_revs.add(rev_id)
338
def _store_new_weave(self, rev, inv, present_parents):
339
# the XML is now updated with text versions
343
if ie.kind == 'root_directory':
345
assert hasattr(ie, 'revision'), \
346
'no revision on {%s} in {%s}' % \
347
(file_id, rev.revision_id)
348
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
349
new_inv_sha1 = sha_string(new_inv_xml)
350
self.inv_weave.add(rev.revision_id,
352
new_inv_xml.splitlines(True),
354
rev.inventory_sha1 = new_inv_sha1
356
def _convert_revision_contents(self, rev, inv, present_parents):
357
"""Convert all the files within a revision.
359
Also upgrade the inventory to refer to the text revision ids."""
360
rev_id = rev.revision_id
361
mutter('converting texts of revision {%s}',
363
parent_invs = map(self._load_updated_inventory, present_parents)
366
self._convert_file_version(rev, ie, parent_invs)
368
def _convert_file_version(self, rev, ie, parent_invs):
369
"""Convert one version of one file.
371
The file needs to be added into the weave if it is a merge
372
of >=2 parents or if it's changed from its parent.
374
if ie.kind == 'root_directory':
377
rev_id = rev.revision_id
378
w = self.text_weaves.get(file_id)
381
self.text_weaves[file_id] = w
383
previous_entries = ie.find_previous_heads(parent_invs, w)
384
for old_revision in previous_entries:
385
# if this fails, its a ghost ?
386
assert old_revision in self.converted_revs
387
self.snapshot_ie(previous_entries, ie, w, rev_id)
389
assert getattr(ie, 'revision', None) is not None
391
def snapshot_ie(self, previous_revisions, ie, w, rev_id):
392
# TODO: convert this logic, which is ~= snapshot to
393
# a call to:. This needs the path figured out. rather than a work_tree
394
# a v4 revision_tree can be given, or something that looks enough like
395
# one to give the file content to the entry if it needs it.
396
# and we need something that looks like a weave store for snapshot to
398
#ie.snapshot(rev, PATH, previous_revisions, REVISION_TREE, InMemoryWeaveStore(self.text_weaves))
399
if len(previous_revisions) == 1:
400
previous_ie = previous_revisions.values()[0]
401
if ie._unchanged(previous_ie):
402
ie.revision = previous_ie.revision
404
parent_indexes = map(w.lookup, previous_revisions)
406
text = self.branch.repository.text_store.get(ie.text_id)
407
file_lines = text.readlines()
408
assert sha_strings(file_lines) == ie.text_sha1
409
assert sum(map(len, file_lines)) == ie.text_size
410
w.add(rev_id, parent_indexes, file_lines, ie.text_sha1)
413
w.add(rev_id, parent_indexes, [], None)
415
##mutter('import text {%s} of {%s}',
416
## ie.text_id, file_id)
418
def _make_order(self):
419
"""Return a suitable order for importing revisions.
421
The order must be such that an revision is imported after all
422
its (present) parents.
424
todo = set(self.revisions.keys())
425
done = self.absent_revisions.copy()
428
# scan through looking for a revision whose parents
430
for rev_id in sorted(list(todo)):
431
rev = self.revisions[rev_id]
432
parent_ids = set(rev.parent_ids)
433
if parent_ids.issubset(done):
434
# can take this one now
442
t = get_transport(url)
added
removed
bzrlib/upgrade.py
