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, find_branch
75
from bzrlib.branch import BZR_BRANCH_FORMAT_5, BZR_BRANCH_FORMAT_6
76
from bzrlib.branch import BzrBranchFormat4, BzrBranchFormat5, BzrBranchFormat6
77
import bzrlib.hashcache as hashcache
78
from bzrlib.weave import Weave
79
from bzrlib.weavefile import read_weave, write_weave
80
from bzrlib.ui import ui_factory
81
from bzrlib.atomicfile import AtomicFile
82
from bzrlib.xml4 import serializer_v4
83
from bzrlib.xml5 import serializer_v5
84
from bzrlib.trace import mutter, note, warning
85
from bzrlib.osutils import sha_strings, sha_string, pathjoin, abspath
88
class Convert(object):
89
def __init__(self, base_dir):
91
self.converted_revs = set()
92
self.absent_revisions = set()
99
if not self._open_branch():
101
note('starting upgrade of %s', os.path.abspath(self.base))
102
self._backup_control_dir()
103
self.pb = ui_factory.progress_bar()
104
if isinstance(self.old_format, BzrBranchFormat4):
105
note('starting upgrade from format 4 to 5')
106
self._convert_to_weaves()
108
if isinstance(self.old_format, BzrBranchFormat5):
109
note('starting upgrade from format 5 to 6')
110
self._convert_to_prefixed()
112
cache = hashcache.HashCache(abspath(self.base))
117
def _convert_to_prefixed(self):
118
from bzrlib.store import hash_prefix
119
for store_name in ["weaves", "revision-store"]:
120
note("adding prefixes to %s" % store_name)
121
store_dir = pathjoin(self.base, ".bzr", store_name)
122
for filename in os.listdir(store_dir):
123
if filename.endswith(".weave") or filename.endswith(".gz"):
124
file_id = os.path.splitext(filename)[0]
127
prefix_dir = pathjoin(store_dir, hash_prefix(file_id))
128
if not os.path.isdir(prefix_dir):
130
os.rename(pathjoin(store_dir, filename),
131
pathjoin(prefix_dir, filename))
132
self._set_new_format(BZR_BRANCH_FORMAT_6)
135
def _convert_to_weaves(self):
136
note('note: upgrade may be faster if all store files are ungzipped first')
137
if not os.path.isdir(self.base + '/.bzr/weaves'):
138
os.mkdir(self.base + '/.bzr/weaves')
139
self.inv_weave = Weave('inventory')
140
# holds in-memory weaves for all files
141
self.text_weaves = {}
142
os.remove(self.branch.controlfilename('branch-format'))
143
self._convert_working_inv()
144
rev_history = self.branch.revision_history()
145
# to_read is a stack holding the revisions we still need to process;
146
# appending to it adds new highest-priority revisions
147
self.known_revisions = set(rev_history)
148
self.to_read = rev_history[-1:]
150
rev_id = self.to_read.pop()
151
if (rev_id not in self.revisions
152
and rev_id not in self.absent_revisions):
153
self._load_one_rev(rev_id)
155
to_import = self._make_order()
156
for i, rev_id in enumerate(to_import):
157
self.pb.update('converting revision', i, len(to_import))
158
self._convert_one_rev(rev_id)
160
note('upgraded to weaves:')
161
note(' %6d revisions and inventories' % len(self.revisions))
162
note(' %6d revisions not present' % len(self.absent_revisions))
163
note(' %6d texts' % self.text_count)
164
self._write_all_weaves()
165
self._write_all_revs()
166
self._cleanup_spare_files()
167
self._set_new_format(BZR_BRANCH_FORMAT_5)
170
def _open_branch(self):
171
self.branch = Branch.open_downlevel(self.base)
172
self.old_format = self.branch._branch_format
173
if isinstance(self.old_format, BzrBranchFormat6):
174
note('this branch is in the most current format (%s)', self.old_format)
176
if (not isinstance(self.old_format, BzrBranchFormat4) and
177
not isinstance(self.old_format, BzrBranchFormat5)):
178
raise BzrError("cannot upgrade from branch format %s" %
179
self.branch._branch_format)
183
def _set_new_format(self, format):
184
self.branch.put_controlfile('branch-format', format)
187
def _cleanup_spare_files(self):
188
for n in 'merged-patches', 'pending-merged-patches':
189
p = self.branch.controlfilename(n)
190
if not os.path.exists(p):
192
## assert os.path.getsize(p) == 0
194
shutil.rmtree(self.base + '/.bzr/inventory-store')
195
shutil.rmtree(self.base + '/.bzr/text-store')
198
def _backup_control_dir(self):
199
orig = self.base + '/.bzr'
200
backup = orig + '.backup'
201
note('making backup of tree history')
202
shutil.copytree(orig, backup)
203
note('%s has been backed up to %s', orig, backup)
204
note('if conversion fails, you can move this directory back to .bzr')
205
note('if it succeeds, you can remove this directory if you wish')
208
def _convert_working_inv(self):
210
inv = serializer_v4.read_inventory(branch.controlfile('inventory', 'rb'))
211
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
212
branch.put_controlfile('inventory', new_inv_xml)
216
def _write_all_weaves(self):
217
write_a_weave(self.inv_weave, self.base + '/.bzr/inventory.weave')
220
for file_id, file_weave in self.text_weaves.items():
221
self.pb.update('writing weave', i, len(self.text_weaves))
222
write_a_weave(file_weave, self.base + '/.bzr/weaves/%s.weave' % file_id)
228
def _write_all_revs(self):
229
"""Write all revisions out in new form."""
230
shutil.rmtree(self.base + '/.bzr/revision-store')
231
os.mkdir(self.base + '/.bzr/revision-store')
233
for i, rev_id in enumerate(self.converted_revs):
234
self.pb.update('write revision', i, len(self.converted_revs))
235
f = file(self.base + '/.bzr/revision-store/%s' % rev_id, 'wb')
237
serializer_v5.write_revision(self.revisions[rev_id], f)
244
def _load_one_rev(self, rev_id):
245
"""Load a revision object into memory.
247
Any parents not either loaded or abandoned get queued to be
249
self.pb.update('loading revision',
251
len(self.known_revisions))
252
if not self.branch.revision_store.has_id(rev_id):
254
note('revision {%s} not present in branch; '
255
'will be converted as a ghost',
257
self.absent_revisions.add(rev_id)
259
rev_xml = self.branch.revision_store.get(rev_id).read()
260
rev = serializer_v4.read_revision_from_string(rev_xml)
261
for parent_id in rev.parent_ids:
262
self.known_revisions.add(parent_id)
263
self.to_read.append(parent_id)
264
self.revisions[rev_id] = rev
267
def _load_old_inventory(self, rev_id):
268
assert rev_id not in self.converted_revs
269
old_inv_xml = self.branch.inventory_store.get(rev_id).read()
270
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
271
rev = self.revisions[rev_id]
272
if rev.inventory_sha1:
273
assert rev.inventory_sha1 == sha_string(old_inv_xml), \
274
'inventory sha mismatch for {%s}' % rev_id
278
def _load_updated_inventory(self, rev_id):
279
assert rev_id in self.converted_revs
280
inv_xml = self.inv_weave.get_text(rev_id)
281
inv = serializer_v5.read_inventory_from_string(inv_xml)
285
def _convert_one_rev(self, rev_id):
286
"""Convert revision and all referenced objects to new format."""
287
rev = self.revisions[rev_id]
288
inv = self._load_old_inventory(rev_id)
289
present_parents = [p for p in rev.parent_ids
290
if p not in self.absent_revisions]
291
self._convert_revision_contents(rev, inv, present_parents)
292
self._store_new_weave(rev, inv, present_parents)
293
self.converted_revs.add(rev_id)
296
def _store_new_weave(self, rev, inv, present_parents):
297
# the XML is now updated with text versions
301
if ie.kind == 'root_directory':
303
assert hasattr(ie, 'revision'), \
304
'no revision on {%s} in {%s}' % \
305
(file_id, rev.revision_id)
306
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
307
new_inv_sha1 = sha_string(new_inv_xml)
308
self.inv_weave.add(rev.revision_id,
310
new_inv_xml.splitlines(True),
312
rev.inventory_sha1 = new_inv_sha1
314
def _convert_revision_contents(self, rev, inv, present_parents):
315
"""Convert all the files within a revision.
317
Also upgrade the inventory to refer to the text revision ids."""
318
rev_id = rev.revision_id
319
mutter('converting texts of revision {%s}',
321
parent_invs = map(self._load_updated_inventory, present_parents)
324
self._convert_file_version(rev, ie, parent_invs)
326
def _convert_file_version(self, rev, ie, parent_invs):
327
"""Convert one version of one file.
329
The file needs to be added into the weave if it is a merge
330
of >=2 parents or if it's changed from its parent.
332
if ie.kind == 'root_directory':
335
rev_id = rev.revision_id
336
w = self.text_weaves.get(file_id)
339
self.text_weaves[file_id] = w
341
previous_entries = ie.find_previous_heads(parent_invs, w)
342
for old_revision in previous_entries:
343
# if this fails, its a ghost ?
344
assert old_revision in self.converted_revs
345
self.snapshot_ie(previous_entries, ie, w, rev_id)
347
assert getattr(ie, 'revision', None) is not None
349
def snapshot_ie(self, previous_revisions, ie, w, rev_id):
350
# TODO: convert this logic, which is ~= snapshot to
351
# a call to:. This needs the path figured out. rather than a work_tree
352
# a v4 revision_tree can be given, or something that looks enough like
353
# one to give the file content to the entry if it needs it.
354
# and we need something that looks like a weave store for snapshot to
356
#ie.snapshot(rev, PATH, previous_revisions, REVISION_TREE, InMemoryWeaveStore(self.text_weaves))
357
if len(previous_revisions) == 1:
358
previous_ie = previous_revisions.values()[0]
359
if ie._unchanged(previous_ie):
360
ie.revision = previous_ie.revision
362
parent_indexes = map(w.lookup, previous_revisions)
364
file_lines = self.branch.text_store.get(ie.text_id).readlines()
365
assert sha_strings(file_lines) == ie.text_sha1
366
assert sum(map(len, file_lines)) == ie.text_size
367
w.add(rev_id, parent_indexes, file_lines, ie.text_sha1)
370
w.add(rev_id, parent_indexes, [], None)
372
##mutter('import text {%s} of {%s}',
373
## ie.text_id, file_id)
375
def _make_order(self):
376
"""Return a suitable order for importing revisions.
378
The order must be such that an revision is imported after all
379
its (present) parents.
381
todo = set(self.revisions.keys())
382
done = self.absent_revisions.copy()
385
# scan through looking for a revision whose parents
387
for rev_id in sorted(list(todo)):
388
rev = self.revisions[rev_id]
389
parent_ids = set(rev.parent_ids)
390
if parent_ids.issubset(done):
391
# can take this one now
398
def write_a_weave(weave, filename):
399
inv_wf = file(filename, 'wb')
401
write_weave(weave, inv_wf)
406
def upgrade(base_dir):
added
removed
bzrlib/upgrade.py
