3
# Copyright (C) 2005 Canonical Ltd
5
# This program is free software; you can redistribute it and/or modify
6
# it under the terms of the GNU General Public License as published by
7
# the Free Software Foundation; either version 2 of the License, or
8
# (at your option) any later version.
10
# This program is distributed in the hope that it will be useful,
11
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
# GNU General Public License for more details.
15
# You should have received a copy of the GNU General Public License
16
# along with this program; if not, write to the Free Software
17
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
"""Experiment in converting existing bzr branches to weaves."""
21
# To make this properly useful
23
# 1. assign text version ids, and put those text versions into
24
# the inventory as they're converted.
26
# 2. keep track of the previous version of each file, rather than
27
# just using the last one imported
29
# 3. assign entry versions when files are added, renamed or moved.
31
# 4. when merged-in versions are observed, walk down through them
32
# to discover everything, then commit bottom-up
34
# 5. track ancestry as things are merged in, and commit that in each
37
# Perhaps it's best to first walk the whole graph and make a plan for
38
# what should be imported in what order? Need a kind of topological
39
# sort of all revisions. (Or do we, can we just before doing a revision
40
# see that all its parents have either been converted or abandoned?)
43
# Cannot import a revision until all its parents have been
44
# imported. in other words, we can only import revisions whose
45
# parents have all been imported. the first step must be to
46
# import a revision with no parents, of which there must be at
47
# least one. (So perhaps it's useful to store forward pointers
48
# from a list of parents to their children?)
50
# Another (equivalent?) approach is to build up the ordered
51
# ancestry list for the last revision, and walk through that. We
52
# are going to need that.
54
# We don't want to have to recurse all the way back down the list.
56
# Suppose we keep a queue of the revisions able to be processed at
57
# any point. This starts out with all the revisions having no
60
# This seems like a generally useful algorithm...
62
# The current algorithm is dumb (O(n**2)?) but will do the job, and
63
# takes less than a second on the bzr.dev branch.
65
# This currently does a kind of lazy conversion of file texts, where a
66
# new text is written in every version. That's unnecessary but for
67
# the moment saves us having to worry about when files need new
76
from bzrlib.branch import Branch, find_branch
77
from bzrlib.branch import BZR_BRANCH_FORMAT_5, BZR_BRANCH_FORMAT_6
78
import bzrlib.hashcache as hashcache
79
from bzrlib.revfile import Revfile
80
from bzrlib.weave import Weave
81
from bzrlib.weavefile import read_weave, write_weave
82
from bzrlib.ui import ui_factory
83
from bzrlib.atomicfile import AtomicFile
84
from bzrlib.xml4 import serializer_v4
85
from bzrlib.xml5 import serializer_v5
86
from bzrlib.trace import mutter, note, warning
87
from bzrlib.osutils import sha_strings, sha_string
90
class Convert(object):
91
def __init__(self, base_dir):
93
self.converted_revs = set()
94
self.absent_revisions = set()
101
if not self._open_branch():
103
note('starting upgrade of %s', os.path.abspath(self.base))
104
self._backup_control_dir()
105
self.pb = ui_factory.progress_bar()
106
if self.old_format == 4:
107
note('starting upgrade from format 4 to 5')
108
self._convert_to_weaves()
110
if self.old_format == 5:
111
note('starting upgrade from format 5 to 6')
112
self._convert_to_prefixed()
114
cache = hashcache.HashCache(os.path.abspath(self.base))
120
def _convert_to_prefixed(self):
121
from bzrlib.store import hash_prefix
122
for store_name in ["weaves", "revision-store"]:
123
note("adding prefixes to %s" % store_name)
124
store_dir = os.path.join(self.base, ".bzr", store_name)
125
for filename in os.listdir(store_dir):
126
if filename.endswith(".weave") or filename.endswith(".gz"):
127
file_id = os.path.splitext(filename)[0]
130
prefix_dir = os.path.join(store_dir, hash_prefix(file_id))
131
if not os.path.isdir(prefix_dir):
133
os.rename(os.path.join(store_dir, filename),
134
os.path.join(prefix_dir, filename))
135
self._set_new_format(BZR_BRANCH_FORMAT_6)
138
def _convert_to_weaves(self):
139
note('note: upgrade may be faster if all store files are ungzipped first')
140
if not os.path.isdir(self.base + '/.bzr/weaves'):
141
os.mkdir(self.base + '/.bzr/weaves')
142
self.inv_weave = Weave('inventory')
143
# holds in-memory weaves for all files
144
self.text_weaves = {}
145
os.remove(self.branch.controlfilename('branch-format'))
146
self._convert_working_inv()
147
rev_history = self.branch.revision_history()
148
# to_read is a stack holding the revisions we still need to process;
149
# appending to it adds new highest-priority revisions
150
self.known_revisions = set(rev_history)
151
self.to_read = rev_history[-1:]
153
rev_id = self.to_read.pop()
154
if (rev_id not in self.revisions
155
and rev_id not in self.absent_revisions):
156
self._load_one_rev(rev_id)
158
to_import = self._make_order()
159
for i, rev_id in enumerate(to_import):
160
self.pb.update('converting revision', i, len(to_import))
161
self._convert_one_rev(rev_id)
163
note('upgraded to weaves:')
164
note(' %6d revisions and inventories' % len(self.revisions))
165
note(' %6d revisions not present' % len(self.absent_revisions))
166
note(' %6d texts' % self.text_count)
167
self._write_all_weaves()
168
self._write_all_revs()
169
self._cleanup_spare_files()
170
self._set_new_format(BZR_BRANCH_FORMAT_5)
173
def _open_branch(self):
174
self.branch = Branch.open_downlevel(self.base)
175
self.old_format = self.branch._branch_format
176
if self.old_format == 6:
177
note('this branch is in the most current format')
179
if self.old_format not in (4, 5):
180
raise BzrError("cannot upgrade from branch format %r" %
181
self.branch._branch_format)
185
def _set_new_format(self, format):
186
self.branch.put_controlfile('branch-format', format)
189
def _cleanup_spare_files(self):
190
for n in 'merged-patches', 'pending-merged-patches':
191
p = self.branch.controlfilename(n)
192
if not os.path.exists(p):
194
## assert os.path.getsize(p) == 0
196
shutil.rmtree(self.base + '/.bzr/inventory-store')
197
shutil.rmtree(self.base + '/.bzr/text-store')
200
def _backup_control_dir(self):
201
orig = self.base + '/.bzr'
202
backup = orig + '.backup'
203
note('making backup of tree history')
204
shutil.copytree(orig, backup)
205
note('%s has been backed up to %s', orig, backup)
206
note('if conversion fails, you can move this directory back to .bzr')
207
note('if it succeeds, you can remove this directory if you wish')
210
def _convert_working_inv(self):
212
inv = serializer_v4.read_inventory(branch.controlfile('inventory', 'rb'))
213
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
214
branch.put_controlfile('inventory', new_inv_xml)
218
def _write_all_weaves(self):
219
write_a_weave(self.inv_weave, self.base + '/.bzr/inventory.weave')
222
for file_id, file_weave in self.text_weaves.items():
223
self.pb.update('writing weave', i, len(self.text_weaves))
224
write_a_weave(file_weave, self.base + '/.bzr/weaves/%s.weave' % file_id)
230
def _write_all_revs(self):
231
"""Write all revisions out in new form."""
232
shutil.rmtree(self.base + '/.bzr/revision-store')
233
os.mkdir(self.base + '/.bzr/revision-store')
235
for i, rev_id in enumerate(self.converted_revs):
236
self.pb.update('write revision', i, len(self.converted_revs))
237
f = file(self.base + '/.bzr/revision-store/%s' % rev_id, 'wb')
239
serializer_v5.write_revision(self.revisions[rev_id], f)
246
def _load_one_rev(self, rev_id):
247
"""Load a revision object into memory.
249
Any parents not either loaded or abandoned get queued to be
251
self.pb.update('loading revision',
253
len(self.known_revisions))
254
if not self.branch.revision_store.has_id(rev_id):
256
note('revision {%s} not present in branch; '
257
'will be converted as a ghost',
259
self.absent_revisions.add(rev_id)
261
rev_xml = self.branch.revision_store.get(rev_id).read()
262
rev = serializer_v4.read_revision_from_string(rev_xml)
263
for parent_id in rev.parent_ids:
264
self.known_revisions.add(parent_id)
265
self.to_read.append(parent_id)
266
self.revisions[rev_id] = rev
269
def _load_old_inventory(self, rev_id):
270
assert rev_id not in self.converted_revs
271
old_inv_xml = self.branch.inventory_store.get(rev_id).read()
272
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
273
rev = self.revisions[rev_id]
274
if rev.inventory_sha1:
275
assert rev.inventory_sha1 == sha_string(old_inv_xml), \
276
'inventory sha mismatch for {%s}' % rev_id
280
def _load_updated_inventory(self, rev_id):
281
assert rev_id in self.converted_revs
282
inv_xml = self.inv_weave.get_text(rev_id)
283
inv = serializer_v5.read_inventory_from_string(inv_xml)
287
def _convert_one_rev(self, rev_id):
288
"""Convert revision and all referenced objects to new format."""
289
rev = self.revisions[rev_id]
290
inv = self._load_old_inventory(rev_id)
291
present_parents = [p for p in rev.parent_ids
292
if p not in self.absent_revisions]
293
self._convert_revision_contents(rev, inv, present_parents)
294
self._store_new_weave(rev, inv, present_parents)
295
self.converted_revs.add(rev_id)
298
def _store_new_weave(self, rev, inv, present_parents):
299
# the XML is now updated with text versions
303
if ie.kind == 'root_directory':
305
assert hasattr(ie, 'revision'), \
306
'no revision on {%s} in {%s}' % \
307
(file_id, rev.revision_id)
308
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
309
new_inv_sha1 = sha_string(new_inv_xml)
310
self.inv_weave.add(rev.revision_id,
312
new_inv_xml.splitlines(True),
314
rev.inventory_sha1 = new_inv_sha1
316
def _convert_revision_contents(self, rev, inv, present_parents):
317
"""Convert all the files within a revision.
319
Also upgrade the inventory to refer to the text revision ids."""
320
rev_id = rev.revision_id
321
mutter('converting texts of revision {%s}',
323
parent_invs = map(self._load_updated_inventory, present_parents)
326
self._convert_file_version(rev, ie, parent_invs)
328
def _convert_file_version(self, rev, ie, parent_invs):
329
"""Convert one version of one file.
331
The file needs to be added into the weave if it is a merge
332
of >=2 parents or if it's changed from its parent.
334
if ie.kind == 'root_directory':
337
rev_id = rev.revision_id
338
w = self.text_weaves.get(file_id)
341
self.text_weaves[file_id] = w
343
previous_entries = ie.find_previous_heads(parent_invs, w)
344
for old_revision in previous_entries:
345
# if this fails, its a ghost ?
346
assert old_revision in self.converted_revs
347
self.snapshot_ie(previous_entries, ie, w, rev_id)
349
assert getattr(ie, 'revision', None) is not None
351
def snapshot_ie(self, previous_revisions, ie, w, rev_id):
352
# TODO: convert this logic, which is ~= snapshot to
353
# a call to:. This needs the path figured out. rather than a work_tree
354
# a v4 revision_tree can be given, or something that looks enough like
355
# one to give the file content to the entry if it needs it.
356
# and we need something that looks like a weave store for snapshot to
358
#ie.snapshot(rev, PATH, previous_revisions, REVISION_TREE, InMemoryWeaveStore(self.text_weaves))
359
if len(previous_revisions) == 1:
360
previous_ie = previous_revisions.values()[0]
361
if ie._unchanged(previous_ie):
362
ie.revision = previous_ie.revision
364
parent_indexes = map(w.lookup, previous_revisions)
366
file_lines = self.branch.text_store.get(ie.text_id).readlines()
367
assert sha_strings(file_lines) == ie.text_sha1
368
assert sum(map(len, file_lines)) == ie.text_size
369
w.add(rev_id, parent_indexes, file_lines, ie.text_sha1)
372
w.add(rev_id, parent_indexes, [], None)
374
##mutter('import text {%s} of {%s}',
375
## ie.text_id, file_id)
377
def _make_order(self):
378
"""Return a suitable order for importing revisions.
380
The order must be such that an revision is imported after all
381
its (present) parents.
383
todo = set(self.revisions.keys())
384
done = self.absent_revisions.copy()
387
# scan through looking for a revision whose parents
389
for rev_id in sorted(list(todo)):
390
rev = self.revisions[rev_id]
391
parent_ids = set(rev.parent_ids)
392
if parent_ids.issubset(done):
393
# can take this one now
400
def write_a_weave(weave, filename):
401
inv_wf = file(filename, 'wb')
403
write_weave(weave, inv_wf)
408
def upgrade(base_dir):
added
removed
bzrlib/upgrade.py
