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
69
# TODO: Don't create a progress bar here, have it passed by the caller.
70
# At least do it from the UI factory.
86
from bzrlib.branch import Branch, find_branch
87
from bzrlib.branch import BZR_BRANCH_FORMAT_5, BZR_BRANCH_FORMAT_6
88
import bzrlib.hashcache as hashcache
89
from bzrlib.revfile import Revfile
90
from bzrlib.weave import Weave
91
from bzrlib.weavefile import read_weave, write_weave
92
from bzrlib.ui import ui_factory
93
from bzrlib.atomicfile import AtomicFile
94
from bzrlib.xml4 import serializer_v4
95
from bzrlib.xml5 import serializer_v5
96
from bzrlib.trace import mutter, note, warning, enable_default_logging
97
from bzrlib.osutils import sha_strings, sha_string
26
100
class Convert(object):
101
def __init__(self, base_dir):
103
self.converted_revs = set()
104
self.absent_revisions = set()
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)
47
110
def convert(self):
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:
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."""
111
if not self._open_branch():
113
note('starting upgrade of %s', os.path.abspath(self.base))
114
self._backup_control_dir()
115
self.pb = ui_factory.progress_bar()
116
if self.old_format == 4:
117
note('starting upgrade from format 4 to 5')
118
self._convert_to_weaves()
120
if self.old_format == 5:
121
note('starting upgrade from format 5 to 6')
122
self._convert_to_prefixed()
124
cache = hashcache.HashCache(os.path.abspath(self.base))
130
def _convert_to_prefixed(self):
131
from bzrlib.store import hash_prefix
132
for store_name in ["weaves", "revision-store"]:
133
note("adding prefixes to %s" % store_name)
134
store_dir = os.path.join(self.base, ".bzr", store_name)
135
for filename in os.listdir(store_dir):
136
if filename.endswith(".weave") or filename.endswith(".gz"):
137
file_id = os.path.splitext(filename)[0]
140
prefix_dir = os.path.join(store_dir, hash_prefix(file_id))
141
if not os.path.isdir(prefix_dir):
143
os.rename(os.path.join(store_dir, filename),
144
os.path.join(prefix_dir, filename))
145
self._set_new_format(BZR_BRANCH_FORMAT_6)
148
def _convert_to_weaves(self):
149
note('note: upgrade may be faster if all store files are ungzipped first')
150
if not os.path.isdir(self.base + '/.bzr/weaves'):
151
os.mkdir(self.base + '/.bzr/weaves')
152
self.inv_weave = Weave('inventory')
153
# holds in-memory weaves for all files
154
self.text_weaves = {}
155
os.remove(self.branch.controlfilename('branch-format'))
156
self._convert_working_inv()
157
rev_history = self.branch.revision_history()
158
# to_read is a stack holding the revisions we still need to process;
159
# appending to it adds new highest-priority revisions
160
self.known_revisions = set(rev_history)
161
self.to_read = rev_history[-1:]
163
rev_id = self.to_read.pop()
164
if (rev_id not in self.revisions
165
and rev_id not in self.absent_revisions):
166
self._load_one_rev(rev_id)
168
to_import = self._make_order()
169
for i, rev_id in enumerate(to_import):
170
self.pb.update('converting revision', i, len(to_import))
171
self._convert_one_rev(rev_id)
173
note('upgraded to weaves:')
174
note(' %6d revisions and inventories' % len(self.revisions))
175
note(' %6d revisions not present' % len(self.absent_revisions))
176
note(' %6d texts' % self.text_count)
177
self._write_all_weaves()
178
self._write_all_revs()
179
self._cleanup_spare_files()
180
self._set_new_format(BZR_BRANCH_FORMAT_5)
183
def _open_branch(self):
184
self.branch = Branch.open_downlevel(self.base)
185
self.old_format = self.branch._branch_format
186
if self.old_format == 6:
187
note('this branch is in the most current format')
189
if self.old_format not in (4, 5):
190
raise BzrError("cannot upgrade from branch format %r" %
191
self.branch._branch_format)
195
def _set_new_format(self, format):
196
self.branch.put_controlfile('branch-format', format)
199
def _cleanup_spare_files(self):
200
for n in 'merged-patches', 'pending-merged-patches':
201
p = self.branch.controlfilename(n)
202
if not os.path.exists(p):
204
## assert os.path.getsize(p) == 0
206
shutil.rmtree(self.base + '/.bzr/inventory-store')
207
shutil.rmtree(self.base + '/.bzr/text-store')
210
def _backup_control_dir(self):
211
orig = self.base + '/.bzr'
212
backup = orig + '.backup'
213
note('making backup of tree history')
214
shutil.copytree(orig, backup)
215
note('%s has been backed up to %s', orig, backup)
216
note('if conversion fails, you can move this directory back to .bzr')
217
note('if it succeeds, you can remove this directory if you wish')
220
def _convert_working_inv(self):
222
inv = serializer_v4.read_inventory(branch.controlfile('inventory', 'rb'))
223
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
224
branch.put_controlfile('inventory', new_inv_xml)
228
def _write_all_weaves(self):
229
write_a_weave(self.inv_weave, self.base + '/.bzr/inventory.weave')
232
for file_id, file_weave in self.text_weaves.items():
233
self.pb.update('writing weave', i, len(self.text_weaves))
234
write_a_weave(file_weave, self.base + '/.bzr/weaves/%s.weave' % file_id)
240
def _write_all_revs(self):
241
"""Write all revisions out in new form."""
242
shutil.rmtree(self.base + '/.bzr/revision-store')
243
os.mkdir(self.base + '/.bzr/revision-store')
245
for i, rev_id in enumerate(self.converted_revs):
246
self.pb.update('write revision', i, len(self.converted_revs))
247
f = file(self.base + '/.bzr/revision-store/%s' % rev_id, 'wb')
249
serializer_v5.write_revision(self.revisions[rev_id], f)
256
def _load_one_rev(self, rev_id):
257
"""Load a revision object into memory.
259
Any parents not either loaded or abandoned get queued to be
261
self.pb.update('loading revision',
263
len(self.known_revisions))
264
if not self.branch.revision_store.has_id(rev_id):
266
note('revision {%s} not present in branch; '
267
'will be converted as a ghost',
269
self.absent_revisions.add(rev_id)
271
rev_xml = self.branch.revision_store.get(rev_id).read()
272
rev = serializer_v4.read_revision_from_string(rev_xml)
273
for parent_id in rev.parent_ids:
274
self.known_revisions.add(parent_id)
275
self.to_read.append(parent_id)
276
self.revisions[rev_id] = rev
279
def _load_old_inventory(self, rev_id):
280
assert rev_id not in self.converted_revs
281
old_inv_xml = self.branch.inventory_store.get(rev_id).read()
282
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
283
rev = self.revisions[rev_id]
284
if rev.inventory_sha1:
285
assert rev.inventory_sha1 == sha_string(old_inv_xml), \
286
'inventory sha mismatch for {%s}' % rev_id
290
def _load_updated_inventory(self, rev_id):
291
assert rev_id in self.converted_revs
292
inv_xml = self.inv_weave.get_text(rev_id)
293
inv = serializer_v5.read_inventory_from_string(inv_xml)
297
def _convert_one_rev(self, rev_id):
298
"""Convert revision and all referenced objects to new format."""
299
rev = self.revisions[rev_id]
300
inv = self._load_old_inventory(rev_id)
301
present_parents = [p for p in rev.parent_ids
302
if p not in self.absent_revisions]
303
self._convert_revision_contents(rev, inv, present_parents)
304
self._store_new_weave(rev, inv, present_parents)
305
self.converted_revs.add(rev_id)
308
def _store_new_weave(self, rev, inv, present_parents):
309
# the XML is now updated with text versions
313
if ie.kind == 'root_directory':
315
assert hasattr(ie, 'revision'), \
316
'no revision on {%s} in {%s}' % \
317
(file_id, rev.revision_id)
318
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
319
new_inv_sha1 = sha_string(new_inv_xml)
320
self.inv_weave.add(rev.revision_id,
322
new_inv_xml.splitlines(True),
324
rev.inventory_sha1 = new_inv_sha1
326
def _convert_revision_contents(self, rev, inv, present_parents):
327
"""Convert all the files within a revision.
329
Also upgrade the inventory to refer to the text revision ids."""
330
rev_id = rev.revision_id
331
mutter('converting texts of revision {%s}',
333
parent_invs = map(self._load_updated_inventory, present_parents)
336
self._convert_file_version(rev, ie, parent_invs)
338
def _convert_file_version(self, rev, ie, parent_invs):
339
"""Convert one version of one file.
341
The file needs to be added into the weave if it is a merge
342
of >=2 parents or if it's changed from its parent.
344
if ie.kind == 'root_directory':
347
rev_id = rev.revision_id
348
w = self.text_weaves.get(file_id)
351
self.text_weaves[file_id] = w
353
previous_entries = ie.find_previous_heads(parent_invs, w)
354
for old_revision in previous_entries:
355
# if this fails, its a ghost ?
356
assert old_revision in self.converted_revs
357
self.snapshot_ie(previous_entries, ie, w, rev_id)
359
assert getattr(ie, 'revision', None) is not None
361
def snapshot_ie(self, previous_revisions, ie, w, rev_id):
362
# TODO: convert this logic, which is ~= snapshot to
363
# a call to:. This needs the path figured out. rather than a work_tree
364
# a v4 revision_tree can be given, or something that looks enough like
365
# one to give the file content to the entry if it needs it.
366
# and we need something that looks like a weave store for snapshot to
368
#ie.snapshot(rev, PATH, previous_revisions, REVISION_TREE, InMemoryWeaveStore(self.text_weaves))
369
if len(previous_revisions) == 1:
370
previous_ie = previous_revisions.values()[0]
371
if ie._unchanged(previous_ie):
372
ie.revision = previous_ie.revision
374
parent_indexes = map(w.lookup, previous_revisions)
376
file_lines = self.branch.text_store.get(ie.text_id).readlines()
377
assert sha_strings(file_lines) == ie.text_sha1
378
assert sum(map(len, file_lines)) == ie.text_size
379
w.add(rev_id, parent_indexes, file_lines, ie.text_sha1)
382
w.add(rev_id, parent_indexes, [], None)
384
##mutter('import text {%s} of {%s}',
385
## ie.text_id, file_id)
387
def _make_order(self):
388
"""Return a suitable order for importing revisions.
390
The order must be such that an revision is imported after all
391
its (present) parents.
393
todo = set(self.revisions.keys())
394
done = self.absent_revisions.copy()
397
# scan through looking for a revision whose parents
399
for rev_id in sorted(list(todo)):
400
rev = self.revisions[rev_id]
401
parent_ids = set(rev.parent_ids)
402
if parent_ids.issubset(done):
403
# can take this one now
410
def write_a_weave(weave, filename):
411
inv_wf = file(filename, 'wb')
413
write_weave(weave, inv_wf)
418
def upgrade(base_dir):
added
removed
bzrlib/upgrade.py
