13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""bzr upgrade logic."""
20
from bzrlib.bzrdir import BzrDir, format_registry
21
import bzrlib.errors as errors
22
from bzrlib.remote import RemoteBzrDir
23
import bzrlib.ui as ui
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
74
from bzrlib.branch import Branch, find_branch
75
from bzrlib.branch import BZR_BRANCH_FORMAT_5, BZR_BRANCH_FORMAT_6
76
from bzrlib.branch import BzrBranchFormat, BzrBranchFormat4, BzrBranchFormat5, BzrBranchFormat6
77
from bzrlib.errors import NoSuchFile, UpgradeReadonly
78
import bzrlib.hashcache as hashcache
79
from bzrlib.osutils import sha_strings, sha_string, pathjoin, abspath
80
from bzrlib.ui import ui_factory
81
from bzrlib.store.text import TextStore
82
from bzrlib.store.weave import WeaveStore
83
from bzrlib.trace import mutter, note, warning
84
from bzrlib.transactions import PassThroughTransaction
85
from bzrlib.transport import get_transport
86
from bzrlib.transport.local import LocalTransport
87
from bzrlib.weave import Weave
88
from bzrlib.weavefile import read_weave, write_weave
89
from bzrlib.xml4 import serializer_v4
90
from bzrlib.xml5 import serializer_v5
26
93
class Convert(object):
28
def __init__(self, url, format=None):
30
self.bzrdir = BzrDir.open_unsupported(url)
31
# XXX: Change to cleanup
32
warning_id = 'cross_format_fetch'
33
saved_warning = warning_id in ui.ui_factory.suppressed_warnings
34
if isinstance(self.bzrdir, RemoteBzrDir):
35
self.bzrdir._ensure_real()
36
self.bzrdir = self.bzrdir._real_bzrdir
37
if self.bzrdir.root_transport.is_readonly():
38
raise errors.UpgradeReadonly
39
self.transport = self.bzrdir.root_transport
40
ui.ui_factory.suppressed_warnings.add(warning_id)
45
ui.ui_factory.suppressed_warnings.remove(warning_id)
95
def __init__(self, transport):
96
self.base = transport.base
97
self.converted_revs = set()
98
self.absent_revisions = set()
101
self.transport = transport
47
104
def convert(self):
105
if self.transport.is_readonly():
106
raise UpgradeReadonly
107
if not self._open_branch():
109
note('starting upgrade of %s', self.base)
110
self._backup_control_dir()
111
self.pb = ui_factory.progress_bar()
112
if isinstance(self.old_format, BzrBranchFormat4):
113
note('starting upgrade from format 4 to 5')
114
self._convert_to_weaves()
115
if isinstance(self.old_format, BzrBranchFormat5):
116
note('starting upgrade from format 5 to 6')
117
self._convert_to_prefixed()
118
if isinstance(self.transport, LocalTransport):
119
cache = hashcache.HashCache(abspath(self.base))
124
def _convert_to_prefixed(self):
125
from bzrlib.store import hash_prefix
126
bzr_transport = self.transport.clone('.bzr')
127
bzr_transport.delete('branch-format')
128
for store_name in ["weaves", "revision-store"]:
129
note("adding prefixes to %s" % store_name)
130
store_transport = bzr_transport.clone(store_name)
131
for filename in store_transport.list_dir('.'):
132
if filename.endswith(".weave") or filename.endswith(".gz"):
133
file_id = os.path.splitext(filename)[0]
136
prefix_dir = hash_prefix(file_id)
137
# FIXME keep track of the dirs made RBC 20060121
139
store_transport.move(filename, prefix_dir + '/' + filename)
140
except NoSuchFile: # catches missing dirs strangely enough
141
store_transport.mkdir(prefix_dir)
142
store_transport.move(filename, prefix_dir + '/' + filename)
143
self._set_new_format(BZR_BRANCH_FORMAT_6)
144
self.branch = BzrBranchFormat6().open(self.transport)
145
self.old_format = self.branch._branch_format
147
def _convert_to_weaves(self):
148
note('note: upgrade may be faster if all store files are ungzipped first')
149
bzr_transport = self.transport.clone('.bzr')
49
branch = self.bzrdir.open_branch()
50
if branch.user_url != self.bzrdir.user_url:
51
ui.ui_factory.note("This is a checkout. The branch (%s) needs to be "
52
"upgraded separately." %
55
except (errors.NotBranchError, errors.IncompatibleRepositories):
56
# might not be a format we can open without upgrading; see e.g.
57
# https://bugs.launchpad.net/bzr/+bug/253891
59
if self.format is None:
152
stat = bzr_transport.stat('weaves')
153
if not S_ISDIR(stat.st_mode):
154
bzr_transport.delete('weaves')
155
bzr_transport.mkdir('weaves')
157
bzr_transport.mkdir('weaves')
158
self.inv_weave = Weave('inventory')
159
# holds in-memory weaves for all files
160
self.text_weaves = {}
161
bzr_transport.delete('branch-format')
162
self._convert_working_inv()
163
rev_history = self.branch.revision_history()
164
# to_read is a stack holding the revisions we still need to process;
165
# appending to it adds new highest-priority revisions
166
self.known_revisions = set(rev_history)
167
self.to_read = rev_history[-1:]
169
rev_id = self.to_read.pop()
170
if (rev_id not in self.revisions
171
and rev_id not in self.absent_revisions):
172
self._load_one_rev(rev_id)
174
to_import = self._make_order()
175
for i, rev_id in enumerate(to_import):
176
self.pb.update('converting revision', i, len(to_import))
177
self._convert_one_rev(rev_id)
179
self._write_all_weaves()
180
self._write_all_revs()
181
note('upgraded to weaves:')
182
note(' %6d revisions and inventories' % len(self.revisions))
183
note(' %6d revisions not present' % len(self.absent_revisions))
184
note(' %6d texts' % self.text_count)
185
self._cleanup_spare_files_after_format4()
186
self._set_new_format(BZR_BRANCH_FORMAT_5)
187
self.branch = BzrBranchFormat5().open(self.transport)
188
self.old_format = self.branch._branch_format
190
def _open_branch(self):
191
self.old_format = BzrBranchFormat.find_format(self.transport)
192
self.branch = self.old_format.open(self.transport)
193
if isinstance(self.old_format, BzrBranchFormat6):
194
note('this branch is in the most current format (%s)', self.old_format)
196
if (not isinstance(self.old_format, BzrBranchFormat4) and
197
not isinstance(self.old_format, BzrBranchFormat5)):
198
raise BzrError("cannot upgrade from branch format %s" %
199
self.branch._branch_format)
202
def _set_new_format(self, format):
203
self.branch.put_controlfile('branch-format', format)
205
def _cleanup_spare_files_after_format4(self):
206
transport = self.transport.clone('.bzr')
207
for n in 'merged-patches', 'pending-merged-patches':
61
rich_root = self.bzrdir.find_repository()._format.rich_root_data
62
except errors.NoRepositoryPresent:
63
rich_root = False # assume no rich roots
65
format_name = "default-rich-root"
67
format_name = "default"
68
format = format_registry.make_bzrdir(format_name)
71
if not self.bzrdir.needs_format_conversion(format):
72
raise errors.UpToDateFormat(self.bzrdir._format)
73
if not self.bzrdir.can_convert_format():
74
raise errors.BzrError("cannot upgrade from bzrdir format %s" %
76
self.bzrdir.check_conversion_target(format)
77
ui.ui_factory.note('starting upgrade of %s' % self.transport.base)
79
self.bzrdir.backup_bzrdir()
80
while self.bzrdir.needs_format_conversion(format):
81
converter = self.bzrdir._format.get_converter(format)
82
self.bzrdir = converter.convert(self.bzrdir, None)
83
ui.ui_factory.note("finished")
86
def upgrade(url, format=None):
87
"""Upgrade to format, or the default bzrdir format if not supplied."""
209
## assert os.path.getsize(p) == 0
213
transport.delete_tree('inventory-store')
214
transport.delete_tree('text-store')
216
def _backup_control_dir(self):
217
note('making backup of tree history')
218
self.transport.copy_tree('.bzr', '.bzr.backup')
219
note('%s.bzr has been backed up to %s.bzr.backup',
222
note('if conversion fails, you can move this directory back to .bzr')
223
note('if it succeeds, you can remove this directory if you wish')
225
def _convert_working_inv(self):
227
inv = serializer_v4.read_inventory(branch.controlfile('inventory', 'rb'))
228
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
229
branch.put_controlfile('inventory', new_inv_xml)
231
def _write_all_weaves(self):
232
bzr_transport = self.transport.clone('.bzr')
233
controlweaves = WeaveStore(bzr_transport, prefixed=False)
234
weave_transport = bzr_transport.clone('weaves')
235
weaves = WeaveStore(weave_transport, prefixed=False)
236
transaction = PassThroughTransaction()
238
controlweaves.put_weave('inventory', self.inv_weave, transaction)
241
for file_id, file_weave in self.text_weaves.items():
242
self.pb.update('writing weave', i, len(self.text_weaves))
243
weaves.put_weave(file_id, file_weave, transaction)
248
def _write_all_revs(self):
249
"""Write all revisions out in new form."""
250
transport = self.transport.clone('.bzr')
251
transport.delete_tree('revision-store')
252
transport.mkdir('revision-store')
253
revision_transport = transport.clone('revision-store')
255
revision_store = TextStore(revision_transport,
259
for i, rev_id in enumerate(self.converted_revs):
260
self.pb.update('write revision', i, len(self.converted_revs))
262
serializer_v5.write_revision(self.revisions[rev_id], rev_tmp)
264
revision_store.add(rev_tmp, rev_id)
269
def _load_one_rev(self, rev_id):
270
"""Load a revision object into memory.
272
Any parents not either loaded or abandoned get queued to be
274
self.pb.update('loading revision',
276
len(self.known_revisions))
277
if not self.branch.revision_store.has_id(rev_id):
279
note('revision {%s} not present in branch; '
280
'will be converted as a ghost',
282
self.absent_revisions.add(rev_id)
284
rev_xml = self.branch.revision_store.get(rev_id).read()
285
rev = serializer_v4.read_revision_from_string(rev_xml)
286
for parent_id in rev.parent_ids:
287
self.known_revisions.add(parent_id)
288
self.to_read.append(parent_id)
289
self.revisions[rev_id] = rev
292
def _load_old_inventory(self, rev_id):
293
assert rev_id not in self.converted_revs
294
old_inv_xml = self.branch.inventory_store.get(rev_id).read()
295
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
296
rev = self.revisions[rev_id]
297
if rev.inventory_sha1:
298
assert rev.inventory_sha1 == sha_string(old_inv_xml), \
299
'inventory sha mismatch for {%s}' % rev_id
303
def _load_updated_inventory(self, rev_id):
304
assert rev_id in self.converted_revs
305
inv_xml = self.inv_weave.get_text(rev_id)
306
inv = serializer_v5.read_inventory_from_string(inv_xml)
310
def _convert_one_rev(self, rev_id):
311
"""Convert revision and all referenced objects to new format."""
312
rev = self.revisions[rev_id]
313
inv = self._load_old_inventory(rev_id)
314
present_parents = [p for p in rev.parent_ids
315
if p not in self.absent_revisions]
316
self._convert_revision_contents(rev, inv, present_parents)
317
self._store_new_weave(rev, inv, present_parents)
318
self.converted_revs.add(rev_id)
321
def _store_new_weave(self, rev, inv, present_parents):
322
# the XML is now updated with text versions
326
if ie.kind == 'root_directory':
328
assert hasattr(ie, 'revision'), \
329
'no revision on {%s} in {%s}' % \
330
(file_id, rev.revision_id)
331
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
332
new_inv_sha1 = sha_string(new_inv_xml)
333
self.inv_weave.add(rev.revision_id,
335
new_inv_xml.splitlines(True),
337
rev.inventory_sha1 = new_inv_sha1
339
def _convert_revision_contents(self, rev, inv, present_parents):
340
"""Convert all the files within a revision.
342
Also upgrade the inventory to refer to the text revision ids."""
343
rev_id = rev.revision_id
344
mutter('converting texts of revision {%s}',
346
parent_invs = map(self._load_updated_inventory, present_parents)
349
self._convert_file_version(rev, ie, parent_invs)
351
def _convert_file_version(self, rev, ie, parent_invs):
352
"""Convert one version of one file.
354
The file needs to be added into the weave if it is a merge
355
of >=2 parents or if it's changed from its parent.
357
if ie.kind == 'root_directory':
360
rev_id = rev.revision_id
361
w = self.text_weaves.get(file_id)
364
self.text_weaves[file_id] = w
366
previous_entries = ie.find_previous_heads(parent_invs, w)
367
for old_revision in previous_entries:
368
# if this fails, its a ghost ?
369
assert old_revision in self.converted_revs
370
self.snapshot_ie(previous_entries, ie, w, rev_id)
372
assert getattr(ie, 'revision', None) is not None
374
def snapshot_ie(self, previous_revisions, ie, w, rev_id):
375
# TODO: convert this logic, which is ~= snapshot to
376
# a call to:. This needs the path figured out. rather than a work_tree
377
# a v4 revision_tree can be given, or something that looks enough like
378
# one to give the file content to the entry if it needs it.
379
# and we need something that looks like a weave store for snapshot to
381
#ie.snapshot(rev, PATH, previous_revisions, REVISION_TREE, InMemoryWeaveStore(self.text_weaves))
382
if len(previous_revisions) == 1:
383
previous_ie = previous_revisions.values()[0]
384
if ie._unchanged(previous_ie):
385
ie.revision = previous_ie.revision
387
parent_indexes = map(w.lookup, previous_revisions)
389
file_lines = self.branch.text_store.get(ie.text_id).readlines()
390
assert sha_strings(file_lines) == ie.text_sha1
391
assert sum(map(len, file_lines)) == ie.text_size
392
w.add(rev_id, parent_indexes, file_lines, ie.text_sha1)
395
w.add(rev_id, parent_indexes, [], None)
397
##mutter('import text {%s} of {%s}',
398
## ie.text_id, file_id)
400
def _make_order(self):
401
"""Return a suitable order for importing revisions.
403
The order must be such that an revision is imported after all
404
its (present) parents.
406
todo = set(self.revisions.keys())
407
done = self.absent_revisions.copy()
410
# scan through looking for a revision whose parents
412
for rev_id in sorted(list(todo)):
413
rev = self.revisions[rev_id]
414
parent_ids = set(rev.parent_ids)
415
if parent_ids.issubset(done):
416
# can take this one now
424
t = get_transport(url)
added
removed
bzrlib/upgrade.py
