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
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.branch.lock_write()
112
self._convert_to_weaves()
116
if self.old_format == 5:
117
note('starting upgrade from format 5 to 6')
118
self.branch.lock_write()
120
self._convert_to_prefixed()
124
cache = hashcache.HashCache(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 = pathjoin(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 = pathjoin(store_dir, hash_prefix(file_id))
141
if not os.path.isdir(prefix_dir):
143
os.rename(pathjoin(store_dir, filename),
144
pathjoin(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.control_files.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)
194
def _set_new_format(self, format):
195
self.branch.control_files.put_utf8('branch-format', format)
197
def _cleanup_spare_files(self):
198
for n in 'merged-patches', 'pending-merged-patches':
199
p = self.branch.control_files.controlfilename(n)
200
if not os.path.exists(p):
202
## assert os.path.getsize(p) == 0
204
shutil.rmtree(self.base + '/.bzr/inventory-store')
205
shutil.rmtree(self.base + '/.bzr/text-store')
207
def _backup_control_dir(self):
208
orig = self.base + '/.bzr'
209
backup = orig + '.backup'
210
note('making backup of tree history')
211
shutil.copytree(orig, backup)
212
note('%s has been backed up to %s', orig, backup)
213
note('if conversion fails, you can move this directory back to .bzr')
214
note('if it succeeds, you can remove this directory if you wish')
216
def _convert_working_inv(self):
218
inv = serializer_v4.read_inventory(branch.control_files.get('inventory'))
219
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
220
branch.control_files.put('inventory', new_inv_xml)
222
def _write_all_weaves(self):
223
write_a_weave(self.inv_weave, self.base + '/.bzr/inventory.weave')
226
for file_id, file_weave in self.text_weaves.items():
227
self.pb.update('writing weave', i, len(self.text_weaves))
228
write_a_weave(file_weave, self.base + '/.bzr/weaves/%s.weave' % file_id)
234
def _write_all_revs(self):
235
"""Write all revisions out in new form."""
236
shutil.rmtree(self.base + '/.bzr/revision-store')
237
os.mkdir(self.base + '/.bzr/revision-store')
239
for i, rev_id in enumerate(self.converted_revs):
240
self.pb.update('write revision', i, len(self.converted_revs))
241
f = file(self.base + '/.bzr/revision-store/%s' % rev_id, 'wb')
243
serializer_v5.write_revision(self.revisions[rev_id], f)
250
def _load_one_rev(self, rev_id):
251
"""Load a revision object into memory.
253
Any parents not either loaded or abandoned get queued to be
255
self.pb.update('loading revision',
257
len(self.known_revisions))
258
if not self.branch.repository.revision_store.has_id(rev_id):
260
note('revision {%s} not present in branch; '
261
'will be converted as a ghost',
263
self.absent_revisions.add(rev_id)
265
rev_xml = self.branch.repository.revision_store.get(rev_id).read()
266
rev = serializer_v4.read_revision_from_string(rev_xml)
267
for parent_id in rev.parent_ids:
268
self.known_revisions.add(parent_id)
269
self.to_read.append(parent_id)
270
self.revisions[rev_id] = rev
273
def _load_old_inventory(self, rev_id):
274
assert rev_id not in self.converted_revs
275
old_inv_xml = self.branch.repository.inventory_store.get(rev_id).read()
276
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
277
rev = self.revisions[rev_id]
278
if rev.inventory_sha1:
279
assert rev.inventory_sha1 == sha_string(old_inv_xml), \
280
'inventory sha mismatch for {%s}' % rev_id
284
def _load_updated_inventory(self, rev_id):
285
assert rev_id in self.converted_revs
286
inv_xml = self.inv_weave.get_text(rev_id)
287
inv = serializer_v5.read_inventory_from_string(inv_xml)
291
def _convert_one_rev(self, rev_id):
292
"""Convert revision and all referenced objects to new format."""
293
rev = self.revisions[rev_id]
294
inv = self._load_old_inventory(rev_id)
295
present_parents = [p for p in rev.parent_ids
296
if p not in self.absent_revisions]
297
self._convert_revision_contents(rev, inv, present_parents)
298
self._store_new_weave(rev, inv, present_parents)
299
self.converted_revs.add(rev_id)
302
def _store_new_weave(self, rev, inv, present_parents):
303
# the XML is now updated with text versions
307
if ie.kind == 'root_directory':
309
assert hasattr(ie, 'revision'), \
310
'no revision on {%s} in {%s}' % \
311
(file_id, rev.revision_id)
312
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
313
new_inv_sha1 = sha_string(new_inv_xml)
314
self.inv_weave.add(rev.revision_id,
316
new_inv_xml.splitlines(True),
318
rev.inventory_sha1 = new_inv_sha1
320
def _convert_revision_contents(self, rev, inv, present_parents):
321
"""Convert all the files within a revision.
323
Also upgrade the inventory to refer to the text revision ids."""
324
rev_id = rev.revision_id
325
mutter('converting texts of revision {%s}',
327
parent_invs = map(self._load_updated_inventory, present_parents)
330
self._convert_file_version(rev, ie, parent_invs)
332
def _convert_file_version(self, rev, ie, parent_invs):
333
"""Convert one version of one file.
335
The file needs to be added into the weave if it is a merge
336
of >=2 parents or if it's changed from its parent.
338
if ie.kind == 'root_directory':
341
rev_id = rev.revision_id
342
w = self.text_weaves.get(file_id)
345
self.text_weaves[file_id] = w
347
previous_entries = ie.find_previous_heads(parent_invs, w)
348
for old_revision in previous_entries:
349
# if this fails, its a ghost ?
350
assert old_revision in self.converted_revs
351
self.snapshot_ie(previous_entries, ie, w, rev_id)
353
assert getattr(ie, 'revision', None) is not None
355
def snapshot_ie(self, previous_revisions, ie, w, rev_id):
356
# TODO: convert this logic, which is ~= snapshot to
357
# a call to:. This needs the path figured out. rather than a work_tree
358
# a v4 revision_tree can be given, or something that looks enough like
359
# one to give the file content to the entry if it needs it.
360
# and we need something that looks like a weave store for snapshot to
362
#ie.snapshot(rev, PATH, previous_revisions, REVISION_TREE, InMemoryWeaveStore(self.text_weaves))
363
if len(previous_revisions) == 1:
364
previous_ie = previous_revisions.values()[0]
365
if ie._unchanged(previous_ie):
366
ie.revision = previous_ie.revision
368
parent_indexes = map(w.lookup, previous_revisions)
370
text = self.branch.repository.text_store.get(ie.text_id)
371
file_lines = text.readlines()
372
assert sha_strings(file_lines) == ie.text_sha1
373
assert sum(map(len, file_lines)) == ie.text_size
374
w.add(rev_id, parent_indexes, file_lines, ie.text_sha1)
377
w.add(rev_id, parent_indexes, [], None)
379
##mutter('import text {%s} of {%s}',
380
## ie.text_id, file_id)
382
def _make_order(self):
383
"""Return a suitable order for importing revisions.
385
The order must be such that an revision is imported after all
386
its (present) parents.
388
todo = set(self.revisions.keys())
389
done = self.absent_revisions.copy()
392
# scan through looking for a revision whose parents
394
for rev_id in sorted(list(todo)):
395
rev = self.revisions[rev_id]
396
parent_ids = set(rev.parent_ids)
397
if parent_ids.issubset(done):
398
# can take this one now
405
def write_a_weave(weave, filename):
406
inv_wf = file(filename, 'wb')
408
write_weave(weave, inv_wf)
413
def upgrade(base_dir):
added
removed
bzrlib/upgrade.py
