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
80
import hotshot, hotshot.stats
85
from bzrlib.branch import Branch, find_branch
86
from bzrlib.revfile import Revfile
87
from bzrlib.weave import Weave
88
from bzrlib.weavefile import read_weave, write_weave
89
from bzrlib.progress import ProgressBar
90
from bzrlib.atomicfile import AtomicFile
91
from bzrlib.xml4 import serializer_v4
92
from bzrlib.xml5 import serializer_v5
93
from bzrlib.trace import mutter, note, warning, enable_default_logging
94
from bzrlib.osutils import sha_strings
98
class Convert(object):
100
self.converted_revs = set()
101
self.absent_revisions = set()
104
self.inventories = {}
111
enable_default_logging()
112
self.pb = ProgressBar()
113
self.inv_weave = Weave('__inventory')
114
self.anc_weave = Weave('__ancestry')
118
# holds in-memory weaves for all files
119
self.text_weaves = {}
121
b = self.branch = Branch('.', relax_version_check=True)
124
rev_history = b.revision_history()
128
# to_read is a stack holding the revisions we still need to process;
129
# appending to it adds new highest-priority revisions
131
self.known_revisions = set(rev_history)
132
self.to_read = [rev_history[-1]]
134
rev_id = self.to_read.pop()
135
if (rev_id not in self.revisions
136
and rev_id not in self.absent_revisions):
137
self._load_one_rev(rev_id)
139
to_import = self._make_order()
140
for i, rev_id in enumerate(to_import):
141
self.pb.update('converting revision', i, len(to_import))
142
self._convert_one_rev(rev_id)
144
print '(not really) upgraded to weaves:'
145
print ' %6d revisions and inventories' % len(self.revisions)
146
print ' %6d absent revisions removed' % len(self.absent_revisions)
147
print ' %6d texts' % self.text_count
149
self._write_all_weaves()
152
def _write_all_weaves(self):
154
write_atomic_weave(self.inv_weave, 'weaves/inventory.weave')
156
for file_id, file_weave in self.text_weaves.items():
157
self.pb.update('writing weave', i, len(self.text_weaves))
158
write_atomic_weave(file_weave, 'weaves/%s.weave' % file_id)
162
## write_atomic_weave(self.anc_weave, 'weaves/ancestry.weave')
165
def _load_one_rev(self, rev_id):
166
"""Load a revision object into memory.
168
Any parents not either loaded or abandoned get queued to be
170
self.pb.update('loading revision',
172
len(self.known_revisions))
173
if rev_id not in self.branch.revision_store:
175
note('revision {%s} not present in branch; '
176
'will not be converted',
178
self.absent_revisions.add(rev_id)
180
rev_xml = self.branch.revision_store[rev_id].read()
181
rev = serializer_v4.read_revision_from_string(rev_xml)
182
for parent_id in rev.parent_ids:
183
self.known_revisions.add(parent_id)
184
self.to_read.append(parent_id)
185
self.revisions[rev_id] = rev
186
old_inv_xml = self.branch.inventory_store[rev_id].read()
187
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
188
self.inventories[rev_id] = inv
191
def _convert_one_rev(self, rev_id):
192
"""Convert revision and all referenced objects to new format."""
193
rev = self.revisions[rev_id]
194
inv = self.inventories[rev_id]
195
self._convert_revision_contents(rev, inv)
196
# the XML is now updated with text versions
197
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
198
inv_parents = [x for x in self.revisions[rev_id].parent_ids
199
if x not in self.absent_revisions]
200
self.inv_weave.add(rev_id, inv_parents,
201
new_inv_xml.splitlines(True))
202
# TODO: Upgrade revision XML and write that out
203
self.converted_revs.add(rev_id)
206
def _convert_revision_contents(self, rev, inv):
207
"""Convert all the files within a revision.
209
Also upgrade the inventory to refer to the text revision ids."""
210
rev_id = rev.revision_id
211
mutter('converting texts of revision {%s}',
213
for path, ie in inv.iter_entries():
214
if ie.kind != 'file':
216
self._convert_file_version(rev, ie)
217
# TODO: Check and convert name versions
220
def _convert_file_version(self, rev, ie):
221
"""Convert one version of one file.
223
The file needs to be added into the weave if it is a merge
224
of >=2 parents or if it's changed from its parent.
227
rev_id = rev.revision_id
228
w = self.text_weaves.get(file_id)
231
self.text_weaves[file_id] = w
232
file_lines = self.branch.text_store[ie.text_id].readlines()
233
assert sha_strings(file_lines) == ie.text_sha1
234
assert sum(map(len, file_lines)) == ie.text_size
237
for parent_id in rev.parent_ids:
238
if parent_id in self.absent_revisions:
240
assert parent_id in self.converted_revs
241
parent_inv = self.inventories[parent_id]
242
if parent_inv.has_id(file_id):
243
parent_ie = parent_inv[file_id]
244
old_text_version = parent_ie.text_version
245
assert old_text_version in self.converted_revs
246
if old_text_version not in file_parents:
247
file_parents.append(old_text_version)
248
if parent_ie.text_sha1 != ie.text_sha1:
250
if len(file_parents) != 1 or text_changed:
251
w.add(rev_id, file_parents, file_lines)
252
ie.name_version = ie.text_version = rev_id
253
mutter('import text {%s} of {%s}',
256
mutter('text of {%s} unchanged from parent', file_id)
257
ie.text_version = file_parents[0]
258
ie.name_version = file_parents[0]
263
def _make_order(self):
264
"""Return a suitable order for importing revisions.
266
The order must be such that an revision is imported after all
267
its (present) parents.
269
todo = set(self.revisions.keys())
270
done = self.absent_revisions.copy()
273
# scan through looking for a revision whose parents
275
for rev_id in sorted(list(todo)):
276
rev = self.revisions[rev_id]
277
parent_ids = set(rev.parent_ids)
278
if parent_ids.issubset(done):
279
# can take this one now
286
def write_atomic_weave(weave, filename):
287
inv_wf = AtomicFile(filename)
289
write_weave(weave, inv_wf)
297
def profile_convert():
298
prof_f = tempfile.NamedTemporaryFile()
300
prof = hotshot.Profile(prof_f.name)
302
prof.runcall(Convert)
305
stats = hotshot.stats.load(prof_f.name)
307
stats.sort_stats('time')
308
# XXX: Might like to write to stderr or the trace file instead but
309
# print_stats seems hardcoded to stdout
310
stats.print_stats(20)
313
enable_default_logging()
315
if '-p' in sys.argv[1:]: