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()
107
self.to_repository.lock_write()
109
self.to_repository.start_write_group()
113
self.to_repository.abort_write_group(suppress_errors=True)
116
self.to_repository.commit_write_group()
119
if self.nested_pb is not None:
120
self.nested_pb.finished()
122
self.to_repository.unlock()
124
self.from_repository.unlock()
127
"""Primary worker function.
129
This initialises all the needed variables, and then fetches the
130
requested revisions, finally clearing the progress bar.
132
# Roughly this is what we're aiming for fetch to become:
134
# missing = self.sink.insert_stream(self.source.get_stream(search))
136
# missing = self.sink.insert_stream(self.source.get_items(missing))
139
self.file_ids_names = {}
140
pp = ProgressPhase('Transferring', 4, self.pb)
143
search = self._revids_to_fetch()
146
self._fetch_everything_for_search(search, pp)
150
def _fetch_everything_for_search(self, search, pp):
151
"""Fetch all data for the given set of revisions."""
152
# The first phase is "file". We pass the progress bar for it directly
153
# into item_keys_introduced_by, which has more information about how
154
# that phase is progressing than we do. Progress updates for the other
155
# phases are taken care of in this function.
156
# XXX: there should be a clear owner of the progress reporting. Perhaps
157
# item_keys_introduced_by should have a richer API than it does at the
158
# moment, so that it can feed the progress information back to this
160
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
162
from_format = self.from_repository._format
163
stream = self.get_stream(search, pp)
164
missing_keys = self.sink.insert_stream(stream, from_format)
166
stream = self.get_stream_for_missing_keys(missing_keys)
167
missing_keys = self.sink.insert_stream(stream, from_format)
169
raise AssertionError(
170
"second push failed to complete a fetch %r." % (
174
if self.pb is not None:
177
def get_stream(self, search, pp):
179
revs = search.get_keys()
180
graph = self.from_repository.get_graph()
181
revs = list(graph.iter_topo_order(revs))
182
data_to_fetch = self.from_repository.item_keys_introduced_by(
185
for knit_kind, file_id, revisions in data_to_fetch:
186
if knit_kind != phase:
188
# Make a new progress bar for this phase
191
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
192
if knit_kind == "file":
193
# Accumulate file texts
194
text_keys.extend([(file_id, revision) for revision in
196
elif knit_kind == "inventory":
197
# Now copy the file texts.
198
to_texts = self.to_repository.texts
199
from_texts = self.from_repository.texts
200
yield ('texts', from_texts.get_record_stream(
201
text_keys, self.to_repository._fetch_order,
202
not self.to_repository._fetch_uses_deltas))
203
# Cause an error if a text occurs after we have done the
206
# Before we process the inventory we generate the root
207
# texts (if necessary) so that the inventories references
209
for _ in self._generate_root_texts(revs):
211
# NB: This currently reopens the inventory weave in source;
212
# using a single stream interface instead would avoid this.
213
self.pb.update("fetch inventory", 0, 1)
214
from_weave = self.from_repository.inventories
215
# we fetch only the referenced inventories because we do not
216
# know for unselected inventories whether all their required
217
# texts are present in the other repository - it could be
219
yield ('inventories', from_weave.get_record_stream(
220
[(rev_id,) for rev_id in revs],
221
self.inventory_fetch_order(),
222
not self.delta_on_metadata()))
223
elif knit_kind == "signatures":
224
# Nothing to do here; this will be taken care of when
225
# _fetch_revision_texts happens.
227
elif knit_kind == "revisions":
228
for _ in self._fetch_revision_texts(revs, self.pb):
231
raise AssertionError("Unknown knit kind %r" % knit_kind)
232
self.count_copied += len(revs)
234
def get_stream_for_missing_keys(self, missing_keys):
235
# missing keys can only occur when we are byte copying and not
236
# translating (because translation means we don't send
237
# unreconstructable deltas ever).
239
keys['texts'] = set()
240
keys['revisions'] = set()
241
keys['inventories'] = set()
242
keys['signatures'] = set()
243
for key in missing_keys:
244
keys[key[0]].add(key[1:])
245
if len(keys['revisions']):
246
# If we allowed copying revisions at this point, we could end up
247
# copying a revision without copying its required texts: a
248
# violation of the requirements for repository integrity.
249
raise AssertionError(
250
'cannot copy revisions to fill in missing deltas %s' % (
252
for substream_kind, keys in keys.iteritems():
253
vf = getattr(self.from_repository, substream_kind)
254
# Ask for full texts always so that we don't need more round trips
256
stream = vf.get_record_stream(keys,
257
self.to_repository._fetch_order, True)
258
yield substream_kind, stream
260
def _revids_to_fetch(self):
261
"""Determines the exact revisions needed from self.from_repository to
262
install self._last_revision in self.to_repository.
264
If no revisions need to be fetched, then this just returns None.
266
mutter('fetch up to rev {%s}', self._last_revision)
267
if self._last_revision is NULL_REVISION:
268
# explicit limit of no revisions needed
270
if (self._last_revision is not None and
271
self.to_repository.has_revision(self._last_revision)):
274
return self.to_repository.search_missing_revision_ids(
275
self.from_repository, self._last_revision,
276
find_ghosts=self.find_ghosts)
277
except errors.NoSuchRevision, e:
278
raise InstallFailed([self._last_revision])
280
def _fetch_revision_texts(self, revs, pb):
281
# fetch signatures first and then the revision texts
282
# may need to be a InterRevisionStore call here.
283
from_sf = self.from_repository.signatures
284
# A missing signature is just skipped.
285
keys = [(rev_id,) for rev_id in revs]
286
signatures = filter_absent(from_sf.get_record_stream(
288
self.to_repository._fetch_order,
289
not self.to_repository._fetch_uses_deltas))
290
# If a revision has a delta, this is actually expanded inside the
291
# insert_record_stream code now, which is an alternate fix for
293
from_rf = self.from_repository.revisions
294
revisions = from_rf.get_record_stream(
296
self.to_repository._fetch_order,
297
not self.delta_on_metadata())
298
return [('signatures', signatures), ('revisions', revisions)]
300
def _generate_root_texts(self, revs):
301
"""This will be called by __fetch between fetching weave texts and
302
fetching the inventory weave.
304
Subclasses should override this if they need to generate root texts
305
after fetching weave texts.
309
def inventory_fetch_order(self):
310
return self.to_repository._fetch_order
312
def delta_on_metadata(self):
313
src_serializer = self.from_repository._format._serializer
314
target_serializer = self.to_repository._format._serializer
315
return (self.to_repository._fetch_uses_deltas and
316
src_serializer == target_serializer)
319
class Inter1and2Helper(object):
320
"""Helper for operations that convert data from model 1 and 2
322
This is for use by fetchers and converters.
325
def __init__(self, source):
328
:param source: The repository data comes from
332
def iter_rev_trees(self, revs):
333
"""Iterate through RevisionTrees efficiently.
335
Additionally, the inventory's revision_id is set if unset.
337
Trees are retrieved in batches of 100, and then yielded in the order
340
:param revs: A list of revision ids
342
# In case that revs is not a list.
345
for tree in self.source.revision_trees(revs[:100]):
346
if tree.inventory.revision_id is None:
347
tree.inventory.revision_id = tree.get_revision_id()
351
def _find_root_ids(self, revs, parent_map, graph):
353
planned_versions = {}
354
for tree in self.iter_rev_trees(revs):
355
revision_id = tree.inventory.root.revision
356
root_id = tree.get_root_id()
357
planned_versions.setdefault(root_id, []).append(revision_id)
358
revision_root[revision_id] = root_id
359
# Find out which parents we don't already know root ids for
361
for revision_parents in parent_map.itervalues():
362
parents.update(revision_parents)
363
parents.difference_update(revision_root.keys() + [NULL_REVISION])
364
# Limit to revisions present in the versionedfile
365
parents = graph.get_parent_map(parents).keys()
366
for tree in self.iter_rev_trees(parents):
367
root_id = tree.get_root_id()
368
revision_root[tree.get_revision_id()] = root_id
369
return revision_root, planned_versions
371
def generate_root_texts(self, revs):
372
"""Generate VersionedFiles for all root ids.
374
:param revs: the revisions to include
376
graph = self.source.get_graph()
377
parent_map = graph.get_parent_map(revs)
378
rev_order = topo_sort(parent_map)
379
rev_id_to_root_id, root_id_to_rev_ids = self._find_root_ids(
380
revs, parent_map, graph)
381
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
383
# Guaranteed stable, this groups all the file id operations together
384
# retaining topological order within the revisions of a file id.
385
# File id splits and joins would invalidate this, but they don't exist
386
# yet, and are unlikely to in non-rich-root environments anyway.
387
root_id_order.sort(key=operator.itemgetter(0))
388
# Create a record stream containing the roots to create.
390
for key in root_id_order:
391
root_id, rev_id = key
392
rev_parents = parent_map[rev_id]
393
# We drop revision parents with different file-ids, because
394
# that represents a rename of the root to a different location
395
# - its not actually a parent for us. (We could look for that
396
# file id in the revision tree at considerably more expense,
397
# but for now this is sufficient (and reconcile will catch and
398
# correct this anyway).
399
# When a parent revision is a ghost, we guess that its root id
400
# was unchanged (rather than trimming it from the parent list).
401
parent_keys = tuple((root_id, parent) for parent in rev_parents
402
if parent != NULL_REVISION and
403
rev_id_to_root_id.get(parent, root_id) == root_id)
404
yield FulltextContentFactory(key, parent_keys, None, '')
405
return [('texts', yield_roots())]
408
class Model1toKnit2Fetcher(RepoFetcher):
409
"""Fetch from a Model1 repository into a Knit2 repository
411
def __init__(self, to_repository, from_repository, last_revision=None,
412
pb=None, find_ghosts=True):
413
self.helper = Inter1and2Helper(from_repository)
414
RepoFetcher.__init__(self, to_repository, from_repository,
415
last_revision, pb, find_ghosts)
417
def _generate_root_texts(self, revs):
418
return self.helper.generate_root_texts(revs)
420
def inventory_fetch_order(self):
423
Knit1to2Fetcher = Model1toKnit2Fetcher
added
removed
bzrlib/fetch.py
