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
81
import hotshot, hotshot.stats
86
from bzrlib.branch import Branch, find_branch, BZR_BRANCH_FORMAT_5
87
from bzrlib.revfile import Revfile
88
from bzrlib.weave import Weave
89
from bzrlib.weavefile import read_weave, write_weave
90
from bzrlib.progress import ProgressBar
91
from bzrlib.atomicfile import AtomicFile
92
from bzrlib.xml4 import serializer_v4
93
from bzrlib.xml5 import serializer_v5
94
from bzrlib.trace import mutter, note, warning, enable_default_logging
95
from bzrlib.osutils import sha_strings, sha_string
96
from bzrlib.commit import merge_ancestry_lines
99
class Convert(object):
100
def __init__(self, base_dir):
102
self.converted_revs = set()
103
self.absent_revisions = set()
106
self.inventories = {}
111
note('starting upgrade of %s', self.base)
112
self._backup_control_dir()
113
note('starting upgrade')
114
note('note: upgrade will be faster if all store files are ungzipped first')
115
self.pb = ProgressBar()
116
if not os.path.isdir(self.base + '/.bzr/weaves'):
117
os.mkdir(self.base + '/.bzr/weaves')
118
self.inv_weave = Weave('__inventory')
119
self.anc_weave = Weave('__ancestry')
121
# holds in-memory weaves for all files
122
self.text_weaves = {}
123
self.branch = Branch(self.base, relax_version_check=True)
124
if self.branch._branch_format == 5:
125
note('this branch is already in the most current format')
127
if self.branch._branch_format != 4:
128
raise BzrError("cannot upgrade from branch format %r" %
129
self.branch._branch_format)
130
os.remove(self.branch.controlfilename('branch-format'))
131
self._convert_working_inv()
132
rev_history = self.branch.revision_history()
133
# to_read is a stack holding the revisions we still need to process;
134
# appending to it adds new highest-priority revisions
135
self.known_revisions = set(rev_history)
136
self.to_read = [rev_history[-1]]
138
rev_id = self.to_read.pop()
139
if (rev_id not in self.revisions
140
and rev_id not in self.absent_revisions):
141
self._load_one_rev(rev_id)
143
to_import = self._make_order()
144
for i, rev_id in enumerate(to_import):
145
self.pb.update('converting revision', i, len(to_import))
146
self._convert_one_rev(rev_id)
148
note('upgraded to weaves:')
149
note(' %6d revisions and inventories' % len(self.revisions))
150
note(' %6d absent revisions removed' % len(self.absent_revisions))
151
note(' %6d texts' % self.text_count)
152
self._write_all_weaves()
153
self._write_all_revs()
154
self._set_new_format()
155
self._cleanup_spare_files()
158
def _set_new_format(self):
159
f = self.branch.controlfile('branch-format', 'wb')
161
f.write(BZR_BRANCH_FORMAT_5)
166
def _cleanup_spare_files(self):
167
for n in 'merged-patches', 'pending-merged-patches':
168
p = self.branch.controlfilename(n)
169
if not os.path.exists(p):
171
assert os.path.getsize(p) == 0
173
os.remove(self.base + '/.bzr/allow-upgrade')
174
shutil.rmtree(self.base + '/.bzr/inventory-store')
175
shutil.rmtree(self.base + '/.bzr/text-store')
178
def _backup_control_dir(self):
179
orig = self.base + '/.bzr'
180
backup = orig + '.backup'
181
shutil.copytree(orig, backup)
182
note('%s has been backed up to %s', orig, backup)
183
note('if conversion fails, you can move this directory back to .bzr')
184
note('if it succeeds, you can remove this directory if you wish')
187
def _convert_working_inv(self):
189
inv = serializer_v4.read_inventory(branch.controlfile('inventory', 'rb'))
190
serializer_v5.write_inventory(inv, branch.controlfile('inventory', 'wb'))
194
def _write_all_weaves(self):
195
write_a_weave(self.inv_weave, self.base + '/.bzr/inventory.weave')
196
write_a_weave(self.anc_weave, self.base + '/.bzr/ancestry.weave')
199
for file_id, file_weave in self.text_weaves.items():
200
self.pb.update('writing weave', i, len(self.text_weaves))
201
write_a_weave(file_weave, self.base + '/.bzr/weaves/%s.weave' % file_id)
207
def _write_all_revs(self):
208
"""Write all revisions out in new form."""
209
shutil.rmtree(self.base + '/.bzr/revision-store')
210
os.mkdir(self.base + '/.bzr/revision-store')
212
for i, rev_id in enumerate(self.converted_revs):
213
self.pb.update('write revision', i, len(self.converted_revs))
214
f = file(self.base + '/.bzr/revision-store/%s' % rev_id, 'wb')
216
serializer_v5.write_revision(self.revisions[rev_id], f)
223
def _load_one_rev(self, rev_id):
224
"""Load a revision object into memory.
226
Any parents not either loaded or abandoned get queued to be
228
self.pb.update('loading revision',
230
len(self.known_revisions))
231
if rev_id not in self.branch.revision_store:
233
note('revision {%s} not present in branch; '
234
'will not be converted',
236
self.absent_revisions.add(rev_id)
238
rev_xml = self.branch.revision_store[rev_id].read()
239
rev = serializer_v4.read_revision_from_string(rev_xml)
240
for parent_id in rev.parent_ids:
241
self.known_revisions.add(parent_id)
242
self.to_read.append(parent_id)
243
self.revisions[rev_id] = rev
244
old_inv_xml = self.branch.inventory_store[rev_id].read()
245
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
246
assert rev.inventory_sha1 == sha_string(old_inv_xml)
247
self.inventories[rev_id] = inv
250
def _convert_one_rev(self, rev_id):
251
"""Convert revision and all referenced objects to new format."""
252
rev = self.revisions[rev_id]
253
inv = self.inventories[rev_id]
254
for parent_id in rev.parent_ids[:]:
255
if parent_id in self.absent_revisions:
256
rev.parent_ids.remove(parent_id)
258
note('remove {%s} as parent of {%s}', parent_id, rev_id)
259
self._convert_revision_contents(rev, inv)
260
# the XML is now updated with text versions
261
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
262
new_inv_sha1 = sha_string(new_inv_xml)
263
self.inv_weave.add(rev_id, rev.parent_ids,
264
new_inv_xml.splitlines(True),
266
# TODO: Upgrade revision XML and write that out
267
rev.inventory_sha1 = new_inv_sha1
268
self._make_rev_ancestry(rev)
269
self.converted_revs.add(rev_id)
272
def _make_rev_ancestry(self, rev):
273
rev_id = rev.revision_id
274
for parent_id in rev.parent_ids:
275
assert parent_id in self.converted_revs
277
lines = list(self.anc_weave.mash_iter(rev.parent_ids))
280
lines.append(rev_id + '\n')
282
parent_ancestries = [self.ancestries[p] for p in rev.parent_ids]
283
new_lines = merge_ancestry_lines(rev_id, parent_ancestries)
284
assert set(lines) == set(new_lines)
285
self.ancestries[rev_id] = new_lines
286
self.anc_weave.add(rev_id, rev.parent_ids, lines)
289
def _convert_revision_contents(self, rev, inv):
290
"""Convert all the files within a revision.
292
Also upgrade the inventory to refer to the text revision ids."""
293
rev_id = rev.revision_id
294
mutter('converting texts of revision {%s}',
298
self._set_name_version(rev, ie)
299
if ie.kind != 'file':
301
self._convert_file_version(rev, ie)
304
def _set_name_version(self, rev, ie):
305
"""Set name version for a file.
307
Done in a slightly lazy way: if the file is renamed or in a merge revision
308
it gets a new version, otherwise the same as before.
311
if len(rev.parent_ids) != 1:
312
ie.name_version = rev.revision_id
314
old_inv = self.inventories[rev.parent_ids[0]]
315
if not old_inv.has_id(file_id):
316
ie.name_version = rev.revision_id
318
old_ie = old_inv[file_id]
319
if (old_ie.parent_id != ie.parent_id
320
or old_ie.name != ie.name):
321
ie.name_version = rev.revision_id
323
ie.name_version = old_ie.name_version
327
def _convert_file_version(self, rev, ie):
328
"""Convert one version of one file.
330
The file needs to be added into the weave if it is a merge
331
of >=2 parents or if it's changed from its parent.
334
rev_id = rev.revision_id
335
w = self.text_weaves.get(file_id)
338
self.text_weaves[file_id] = w
339
file_lines = self.branch.text_store[ie.text_id].readlines()
340
assert sha_strings(file_lines) == ie.text_sha1
341
assert sum(map(len, file_lines)) == ie.text_size
344
for parent_id in rev.parent_ids:
345
##if parent_id in self.absent_revisions:
347
assert parent_id in self.converted_revs, \
348
'parent {%s} not converted' % parent_id
349
parent_inv = self.inventories[parent_id]
350
if parent_inv.has_id(file_id):
351
parent_ie = parent_inv[file_id]
352
old_text_version = parent_ie.text_version
353
assert old_text_version in self.converted_revs
354
if old_text_version not in file_parents:
355
file_parents.append(old_text_version)
356
if parent_ie.text_sha1 != ie.text_sha1:
358
if len(file_parents) != 1 or text_changed:
359
w.add(rev_id, file_parents, file_lines, ie.text_sha1)
360
ie.text_version = rev_id
362
##mutter('import text {%s} of {%s}',
363
## ie.text_id, file_id)
365
##mutter('text of {%s} unchanged from parent', file_id)
366
ie.text_version = file_parents[0]
371
def _make_order(self):
372
"""Return a suitable order for importing revisions.
374
The order must be such that an revision is imported after all
375
its (present) parents.
377
todo = set(self.revisions.keys())
378
done = self.absent_revisions.copy()
381
# scan through looking for a revision whose parents
383
for rev_id in sorted(list(todo)):
384
rev = self.revisions[rev_id]
385
parent_ids = set(rev.parent_ids)
386
if parent_ids.issubset(done):
387
# can take this one now
394
def write_a_weave(weave, filename):
395
inv_wf = file(filename, 'wb')
397
write_weave(weave, inv_wf)
402
def upgrade(base_dir):
added
removed
bzrlib/upgrade.py
