1
# Copyright (C) 2005, 2006, 2008 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
18
"""Copying of history from one branch to another.
20
The basic plan is that every branch knows the history of everything
21
that has merged into it. As the first step of a merge, pull, or
22
branch operation we copy history from the source into the destination
25
The copying is done in a slightly complicated order. We don't want to
26
add a revision to the store until everything it refers to is also
27
stored, so that if a revision is present we can totally recreate it.
28
However, we can't know what files are included in a revision until we
29
read its inventory. So we query the inventory store of the source for
30
the ids we need, and then pull those ids and then return to the inventories.
36
import bzrlib.errors as errors
37
from bzrlib.errors import InstallFailed
38
from bzrlib.progress import ProgressPhase
39
from bzrlib.revision import NULL_REVISION
40
from bzrlib.tsort import topo_sort
41
from bzrlib.trace import mutter
43
from bzrlib.versionedfile import filter_absent, FulltextContentFactory
45
# TODO: Avoid repeatedly opening weaves so many times.
47
# XXX: This doesn't handle ghost (not present in branch) revisions at
48
# all yet. I'm not sure they really should be supported.
50
# NOTE: This doesn't copy revisions which may be present but not
51
# merged into the last revision. I'm not sure we want to do that.
53
# - get a list of revisions that need to be pulled in
54
# - for each one, pull in that revision file
55
# and get the inventory, and store the inventory with right
57
# - and get the ancestry, and store that with right parents too
58
# - and keep a note of all file ids and version seen
59
# - then go through all files; for each one get the weave,
60
# and add in all file versions
63
class RepoFetcher(object):
64
"""Pull revisions and texts from one repository to another.
67
if set, try to limit to the data this revision references.
70
count_copied -- number of revisions copied
72
This should not be used directly, it's essential a object to encapsulate
73
the logic in InterRepository.fetch().
76
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
78
"""Create a repo fetcher.
80
:param find_ghosts: If True search the entire history for ghosts.
81
:param _write_group_acquired_callable: Don't use; this parameter only
82
exists to facilitate a hack done in InterPackRepo.fetch. We would
83
like to remove this parameter.
86
self.failed_revisions = []
88
if to_repository.has_same_location(from_repository):
89
# repository.fetch should be taking care of this case.
90
raise errors.BzrError('RepoFetcher run '
91
'between two objects at the same location: '
92
'%r and %r' % (to_repository, from_repository))
93
self.to_repository = to_repository
94
self.from_repository = from_repository
95
self.sink = to_repository._get_sink()
96
# must not mutate self._last_revision as its potentially a shared instance
97
self._last_revision = last_revision
98
self.find_ghosts = find_ghosts
100
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
101
self.nested_pb = self.pb
104
self.nested_pb = None
105
self.from_repository.lock_read()
110
if self.nested_pb is not None:
111
self.nested_pb.finished()
113
self.from_repository.unlock()
116
"""Primary worker function.
118
This initialises all the needed variables, and then fetches the
119
requested revisions, finally clearing the progress bar.
121
# Roughly this is what we're aiming for fetch to become:
123
# missing = self.sink.insert_stream(self.source.get_stream(search))
125
# missing = self.sink.insert_stream(self.source.get_items(missing))
128
self.file_ids_names = {}
129
pp = ProgressPhase('Transferring', 4, self.pb)
132
search = self._revids_to_fetch()
135
self._fetch_everything_for_search(search, pp)
139
def _fetch_everything_for_search(self, search, pp):
140
"""Fetch all data for the given set of revisions."""
141
# The first phase is "file". We pass the progress bar for it directly
142
# into item_keys_introduced_by, which has more information about how
143
# that phase is progressing than we do. Progress updates for the other
144
# phases are taken care of in this function.
145
# XXX: there should be a clear owner of the progress reporting. Perhaps
146
# item_keys_introduced_by should have a richer API than it does at the
147
# moment, so that it can feed the progress information back to this
149
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
151
from_format = self.from_repository._format
152
stream = self.get_stream(search, pp)
153
resume_tokens, missing_keys = self.sink.insert_stream(
154
stream, from_format, [])
156
stream = self.get_stream_for_missing_keys(missing_keys)
157
resume_tokens, missing_keys = self.sink.insert_stream(
158
stream, from_format, resume_tokens)
160
raise AssertionError(
161
"second push failed to complete a fetch %r." % (
164
raise AssertionError(
165
"second push failed to commit the fetch %r." % (
169
if self.pb is not None:
172
def get_stream(self, search, pp):
174
revs = search.get_keys()
175
graph = self.from_repository.get_graph()
176
revs = list(graph.iter_topo_order(revs))
177
data_to_fetch = self.from_repository.item_keys_introduced_by(
180
for knit_kind, file_id, revisions in data_to_fetch:
181
if knit_kind != phase:
183
# Make a new progress bar for this phase
186
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
187
if knit_kind == "file":
188
# Accumulate file texts
189
text_keys.extend([(file_id, revision) for revision in
191
elif knit_kind == "inventory":
192
# Now copy the file texts.
193
from_texts = self.from_repository.texts
194
yield ('texts', from_texts.get_record_stream(
195
text_keys, self.to_repository._fetch_order,
196
not self.to_repository._fetch_uses_deltas))
197
# Cause an error if a text occurs after we have done the
200
# Before we process the inventory we generate the root
201
# texts (if necessary) so that the inventories references
203
for _ in self._generate_root_texts(revs):
205
# NB: This currently reopens the inventory weave in source;
206
# using a single stream interface instead would avoid this.
207
self.pb.update("fetch inventory", 0, 1)
208
from_weave = self.from_repository.inventories
209
# we fetch only the referenced inventories because we do not
210
# know for unselected inventories whether all their required
211
# texts are present in the other repository - it could be
213
yield ('inventories', from_weave.get_record_stream(
214
[(rev_id,) for rev_id in revs],
215
self.inventory_fetch_order(),
216
not self.delta_on_metadata()))
217
elif knit_kind == "signatures":
218
# Nothing to do here; this will be taken care of when
219
# _fetch_revision_texts happens.
221
elif knit_kind == "revisions":
222
for _ in self._fetch_revision_texts(revs, self.pb):
225
raise AssertionError("Unknown knit kind %r" % knit_kind)
226
self.count_copied += len(revs)
228
def get_stream_for_missing_keys(self, missing_keys):
229
# missing keys can only occur when we are byte copying and not
230
# translating (because translation means we don't send
231
# unreconstructable deltas ever).
233
keys['texts'] = set()
234
keys['revisions'] = set()
235
keys['inventories'] = set()
236
keys['signatures'] = set()
237
for key in missing_keys:
238
keys[key[0]].add(key[1:])
239
if len(keys['revisions']):
240
# If we allowed copying revisions at this point, we could end up
241
# copying a revision without copying its required texts: a
242
# violation of the requirements for repository integrity.
243
raise AssertionError(
244
'cannot copy revisions to fill in missing deltas %s' % (
246
for substream_kind, keys in keys.iteritems():
247
vf = getattr(self.from_repository, substream_kind)
248
# Ask for full texts always so that we don't need more round trips
250
stream = vf.get_record_stream(keys,
251
self.to_repository._fetch_order, True)
252
yield substream_kind, stream
254
def _revids_to_fetch(self):
255
"""Determines the exact revisions needed from self.from_repository to
256
install self._last_revision in self.to_repository.
258
If no revisions need to be fetched, then this just returns None.
260
mutter('fetch up to rev {%s}', self._last_revision)
261
if self._last_revision is NULL_REVISION:
262
# explicit limit of no revisions needed
264
if (self._last_revision is not None and
265
self.to_repository.has_revision(self._last_revision)):
268
return self.to_repository.search_missing_revision_ids(
269
self.from_repository, self._last_revision,
270
find_ghosts=self.find_ghosts)
271
except errors.NoSuchRevision, e:
272
raise InstallFailed([self._last_revision])
274
def _fetch_revision_texts(self, revs, pb):
275
# fetch signatures first and then the revision texts
276
# may need to be a InterRevisionStore call here.
277
from_sf = self.from_repository.signatures
278
# A missing signature is just skipped.
279
keys = [(rev_id,) for rev_id in revs]
280
signatures = filter_absent(from_sf.get_record_stream(
282
self.to_repository._fetch_order,
283
not self.to_repository._fetch_uses_deltas))
284
# If a revision has a delta, this is actually expanded inside the
285
# insert_record_stream code now, which is an alternate fix for
287
from_rf = self.from_repository.revisions
288
revisions = from_rf.get_record_stream(
290
self.to_repository._fetch_order,
291
not self.delta_on_metadata())
292
return [('signatures', signatures), ('revisions', revisions)]
294
def _generate_root_texts(self, revs):
295
"""This will be called by __fetch between fetching weave texts and
296
fetching the inventory weave.
298
Subclasses should override this if they need to generate root texts
299
after fetching weave texts.
303
def inventory_fetch_order(self):
304
return self.to_repository._fetch_order
306
def delta_on_metadata(self):
307
src_serializer = self.from_repository._format._serializer
308
target_serializer = self.to_repository._format._serializer
309
return (self.to_repository._fetch_uses_deltas and
310
src_serializer == target_serializer)
313
class Inter1and2Helper(object):
314
"""Helper for operations that convert data from model 1 and 2
316
This is for use by fetchers and converters.
319
def __init__(self, source):
322
:param source: The repository data comes from
326
def iter_rev_trees(self, revs):
327
"""Iterate through RevisionTrees efficiently.
329
Additionally, the inventory's revision_id is set if unset.
331
Trees are retrieved in batches of 100, and then yielded in the order
334
:param revs: A list of revision ids
336
# In case that revs is not a list.
339
for tree in self.source.revision_trees(revs[:100]):
340
if tree.inventory.revision_id is None:
341
tree.inventory.revision_id = tree.get_revision_id()
345
def _find_root_ids(self, revs, parent_map, graph):
347
planned_versions = {}
348
for tree in self.iter_rev_trees(revs):
349
revision_id = tree.inventory.root.revision
350
root_id = tree.get_root_id()
351
planned_versions.setdefault(root_id, []).append(revision_id)
352
revision_root[revision_id] = root_id
353
# Find out which parents we don't already know root ids for
355
for revision_parents in parent_map.itervalues():
356
parents.update(revision_parents)
357
parents.difference_update(revision_root.keys() + [NULL_REVISION])
358
# Limit to revisions present in the versionedfile
359
parents = graph.get_parent_map(parents).keys()
360
for tree in self.iter_rev_trees(parents):
361
root_id = tree.get_root_id()
362
revision_root[tree.get_revision_id()] = root_id
363
return revision_root, planned_versions
365
def generate_root_texts(self, revs):
366
"""Generate VersionedFiles for all root ids.
368
:param revs: the revisions to include
370
graph = self.source.get_graph()
371
parent_map = graph.get_parent_map(revs)
372
rev_order = topo_sort(parent_map)
373
rev_id_to_root_id, root_id_to_rev_ids = self._find_root_ids(
374
revs, parent_map, graph)
375
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
377
# Guaranteed stable, this groups all the file id operations together
378
# retaining topological order within the revisions of a file id.
379
# File id splits and joins would invalidate this, but they don't exist
380
# yet, and are unlikely to in non-rich-root environments anyway.
381
root_id_order.sort(key=operator.itemgetter(0))
382
# Create a record stream containing the roots to create.
384
for key in root_id_order:
385
root_id, rev_id = key
386
rev_parents = parent_map[rev_id]
387
# We drop revision parents with different file-ids, because
388
# that represents a rename of the root to a different location
389
# - its not actually a parent for us. (We could look for that
390
# file id in the revision tree at considerably more expense,
391
# but for now this is sufficient (and reconcile will catch and
392
# correct this anyway).
393
# When a parent revision is a ghost, we guess that its root id
394
# was unchanged (rather than trimming it from the parent list).
395
parent_keys = tuple((root_id, parent) for parent in rev_parents
396
if parent != NULL_REVISION and
397
rev_id_to_root_id.get(parent, root_id) == root_id)
398
yield FulltextContentFactory(key, parent_keys, None, '')
399
return [('texts', yield_roots())]
402
class Model1toKnit2Fetcher(RepoFetcher):
403
"""Fetch from a Model1 repository into a Knit2 repository
405
def __init__(self, to_repository, from_repository, last_revision=None,
406
pb=None, find_ghosts=True):
407
self.helper = Inter1and2Helper(from_repository)
408
RepoFetcher.__init__(self, to_repository, from_repository,
409
last_revision, pb, find_ghosts)
411
def _generate_root_texts(self, revs):
412
return self.helper.generate_root_texts(revs)
414
def inventory_fetch_order(self):
417
Knit1to2Fetcher = Model1toKnit2Fetcher
added
removed
bzrlib/fetch.py
