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 is_null, NULL_REVISION
40
from bzrlib.symbol_versioning import (deprecated_function,
43
from bzrlib.tsort import topo_sort
44
from bzrlib.trace import mutter
46
from bzrlib.versionedfile import filter_absent, FulltextContentFactory
48
# TODO: Avoid repeatedly opening weaves so many times.
50
# XXX: This doesn't handle ghost (not present in branch) revisions at
51
# all yet. I'm not sure they really should be supported.
53
# NOTE: This doesn't copy revisions which may be present but not
54
# merged into the last revision. I'm not sure we want to do that.
56
# - get a list of revisions that need to be pulled in
57
# - for each one, pull in that revision file
58
# and get the inventory, and store the inventory with right
60
# - and get the ancestry, and store that with right parents too
61
# - and keep a note of all file ids and version seen
62
# - then go through all files; for each one get the weave,
63
# and add in all file versions
66
class RepoFetcher(object):
67
"""Pull revisions and texts from one repository to another.
70
if set, try to limit to the data this revision references.
73
count_copied -- number of revisions copied
75
This should not be used directly, it's essential a object to encapsulate
76
the logic in InterRepository.fetch().
79
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
81
"""Create a repo fetcher.
83
:param find_ghosts: If True search the entire history for ghosts.
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
# must not mutate self._last_revision as its potentially a shared instance
96
self._last_revision = last_revision
97
self.find_ghosts = find_ghosts
99
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
100
self.nested_pb = self.pb
103
self.nested_pb = None
104
self.from_repository.lock_read()
106
self.to_repository.lock_write()
108
self.to_repository.start_write_group()
112
self.to_repository.abort_write_group()
115
self.to_repository.commit_write_group()
118
if self.nested_pb is not None:
119
self.nested_pb.finished()
121
self.to_repository.unlock()
123
self.from_repository.unlock()
126
"""Primary worker function.
128
This initialises all the needed variables, and then fetches the
129
requested revisions, finally clearing the progress bar.
132
self.file_ids_names = {}
133
pp = ProgressPhase('Transferring', 4, self.pb)
136
search = self._revids_to_fetch()
139
if getattr(self, '_fetch_everything_for_search', None) is not None:
140
self._fetch_everything_for_search(search, pp)
142
# backward compatibility
143
self._fetch_everything_for_revisions(search.get_keys, pp)
147
def _fetch_everything_for_search(self, search, pp):
148
"""Fetch all data for the given set of revisions."""
149
# The first phase is "file". We pass the progress bar for it directly
150
# into item_keys_introduced_by, which has more information about how
151
# that phase is progressing than we do. Progress updates for the other
152
# phases are taken care of in this function.
153
# XXX: there should be a clear owner of the progress reporting. Perhaps
154
# item_keys_introduced_by should have a richer API than it does at the
155
# moment, so that it can feed the progress information back to this
158
pb = bzrlib.ui.ui_factory.nested_progress_bar()
160
revs = search.get_keys()
161
graph = self.from_repository.get_graph()
162
revs = list(graph.iter_topo_order(revs))
163
data_to_fetch = self.from_repository.item_keys_introduced_by(revs,
166
for knit_kind, file_id, revisions in data_to_fetch:
167
if knit_kind != phase:
169
# Make a new progress bar for this phase
172
pb = bzrlib.ui.ui_factory.nested_progress_bar()
173
if knit_kind == "file":
174
# Accumulate file texts
175
text_keys.extend([(file_id, revision) for revision in
177
elif knit_kind == "inventory":
178
# Now copy the file texts.
179
to_texts = self.to_repository.texts
180
from_texts = self.from_repository.texts
181
to_texts.insert_record_stream(from_texts.get_record_stream(
182
text_keys, 'topological', False))
183
# Cause an error if a text occurs after we have done the
186
# Before we process the inventory we generate the root
187
# texts (if necessary) so that the inventories references
189
self._generate_root_texts(revs)
190
# NB: This currently reopens the inventory weave in source;
191
# using a single stream interface instead would avoid this.
192
self._fetch_inventory_weave(revs, pb)
193
elif knit_kind == "signatures":
194
# Nothing to do here; this will be taken care of when
195
# _fetch_revision_texts happens.
197
elif knit_kind == "revisions":
198
self._fetch_revision_texts(revs, pb)
200
raise AssertionError("Unknown knit kind %r" % knit_kind)
204
self.count_copied += len(revs)
206
def _revids_to_fetch(self):
207
"""Determines the exact revisions needed from self.from_repository to
208
install self._last_revision in self.to_repository.
210
If no revisions need to be fetched, then this just returns None.
212
mutter('fetch up to rev {%s}', self._last_revision)
213
if self._last_revision is NULL_REVISION:
214
# explicit limit of no revisions needed
216
if (self._last_revision is not None and
217
self.to_repository.has_revision(self._last_revision)):
220
return self.to_repository.search_missing_revision_ids(
221
self.from_repository, self._last_revision,
222
find_ghosts=self.find_ghosts)
223
except errors.NoSuchRevision, e:
224
raise InstallFailed([self._last_revision])
226
def _fetch_inventory_weave(self, revs, pb):
227
pb.update("fetch inventory", 0, 2)
228
to_weave = self.to_repository.inventories
229
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
231
# just merge, this is optimisable and its means we don't
232
# copy unreferenced data such as not-needed inventories.
233
pb.update("fetch inventory", 1, 3)
234
from_weave = self.from_repository.inventories
235
pb.update("fetch inventory", 2, 3)
236
# we fetch only the referenced inventories because we do not
237
# know for unselected inventories whether all their required
238
# texts are present in the other repository - it could be
240
to_weave.insert_record_stream(from_weave.get_record_stream(
241
[(rev_id,) for rev_id in revs],
242
'topological', False))
246
def _generate_root_texts(self, revs):
247
"""This will be called by __fetch between fetching weave texts and
248
fetching the inventory weave.
250
Subclasses should override this if they need to generate root texts
251
after fetching weave texts.
256
class GenericRepoFetcher(RepoFetcher):
257
"""This is a generic repo to repo fetcher.
259
This makes minimal assumptions about repo layout and contents.
260
It triggers a reconciliation after fetching to ensure integrity.
263
def _fetch_revision_texts(self, revs, pb):
264
"""Fetch revision object texts"""
268
pb.update('copying revisions', count, total)
270
sig_text = self.from_repository.get_signature_text(rev)
271
self.to_repository.add_signature_text(rev, sig_text)
272
except errors.NoSuchRevision:
275
self._copy_revision(rev)
277
# fixup inventory if needed:
278
# this is expensive because we have no inverse index to current ghosts.
279
# but on local disk its a few seconds and sftp push is already insane.
281
# FIXME: repository should inform if this is needed.
282
self.to_repository.reconcile()
284
def _copy_revision(self, rev_id):
285
rev = self.from_repository.get_revision(rev_id)
286
self.to_repository.add_revision(rev_id, rev)
289
class KnitRepoFetcher(RepoFetcher):
290
"""This is a knit format repository specific fetcher.
292
This differs from the GenericRepoFetcher by not doing a
293
reconciliation after copying, and using knit joining to
297
def _fetch_revision_texts(self, revs, pb):
298
# may need to be a InterRevisionStore call here.
299
to_sf = self.to_repository.signatures
300
from_sf = self.from_repository.signatures
301
# A missing signature is just skipped.
302
to_sf.insert_record_stream(filter_absent(from_sf.get_record_stream(
303
[(rev_id,) for rev_id in revs],
304
'unordered', False)))
305
self._fetch_just_revision_texts(revs)
307
def _fetch_just_revision_texts(self, version_ids):
308
to_rf = self.to_repository.revisions
309
from_rf = self.from_repository.revisions
310
to_rf.insert_record_stream(from_rf.get_record_stream(
311
[(rev_id,) for rev_id in version_ids],
312
'topological', False))
315
class Inter1and2Helper(object):
316
"""Helper for operations that convert data from model 1 and 2
318
This is for use by fetchers and converters.
321
def __init__(self, source, target):
324
:param source: The repository data comes from
325
:param target: The repository data goes to
330
def iter_rev_trees(self, revs):
331
"""Iterate through RevisionTrees efficiently.
333
Additionally, the inventory's revision_id is set if unset.
335
Trees are retrieved in batches of 100, and then yielded in the order
338
:param revs: A list of revision ids
340
# In case that revs is not a list.
343
for tree in self.source.revision_trees(revs[:100]):
344
if tree.inventory.revision_id is None:
345
tree.inventory.revision_id = tree.get_revision_id()
349
def _find_root_ids(self, revs, parent_map, graph):
351
planned_versions = {}
352
for tree in self.iter_rev_trees(revs):
353
revision_id = tree.inventory.root.revision
354
root_id = tree.get_root_id()
355
planned_versions.setdefault(root_id, []).append(revision_id)
356
revision_root[revision_id] = root_id
357
# Find out which parents we don't already know root ids for
359
for revision_parents in parent_map.itervalues():
360
parents.update(revision_parents)
361
parents.difference_update(revision_root.keys() + [NULL_REVISION])
362
# Limit to revisions present in the versionedfile
363
parents = graph.get_parent_map(parents).keys()
364
for tree in self.iter_rev_trees(parents):
365
root_id = tree.get_root_id()
366
revision_root[tree.get_revision_id()] = root_id
367
return revision_root, planned_versions
369
def generate_root_texts(self, revs):
370
"""Generate VersionedFiles for all root ids.
372
:param revs: the revisions to include
374
to_texts = self.target.texts
375
graph = self.source.get_graph()
376
parent_map = graph.get_parent_map(revs)
377
rev_order = topo_sort(parent_map)
378
rev_id_to_root_id, root_id_to_rev_ids = self._find_root_ids(
379
revs, parent_map, graph)
380
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
382
# Guaranteed stable, this groups all the file id operations together
383
# retaining topological order within the revisions of a file id.
384
# File id splits and joins would invalidate this, but they don't exist
385
# yet, and are unlikely to in non-rich-root environments anyway.
386
root_id_order.sort(key=operator.itemgetter(0))
387
# Create a record stream containing the roots to create.
389
for key in root_id_order:
390
root_id, rev_id = key
391
rev_parents = parent_map[rev_id]
392
# We drop revision parents with different file-ids, because
393
# that represents a rename of the root to a different location
394
# - its not actually a parent for us. (We could look for that
395
# file id in the revision tree at considerably more expense,
396
# but for now this is sufficient (and reconcile will catch and
397
# correct this anyway).
398
# When a parent revision is a ghost, we guess that its root id
399
# was unchanged (rather than trimming it from the parent list).
400
parent_keys = tuple((root_id, parent) for parent in rev_parents
401
if parent != NULL_REVISION and
402
rev_id_to_root_id.get(parent, root_id) == root_id)
403
yield FulltextContentFactory(key, parent_keys, None, '')
404
to_texts.insert_record_stream(yield_roots())
406
def regenerate_inventory(self, revs):
407
"""Generate a new inventory versionedfile in target, convertin data.
409
The inventory is retrieved from the source, (deserializing it), and
410
stored in the target (reserializing it in a different format).
411
:param revs: The revisions to include
413
for tree in self.iter_rev_trees(revs):
414
parents = tree.get_parent_ids()
415
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
418
def fetch_revisions(self, revision_ids):
419
for revision in self.source.get_revisions(revision_ids):
420
self.target.add_revision(revision.revision_id, revision)
423
class Model1toKnit2Fetcher(GenericRepoFetcher):
424
"""Fetch from a Model1 repository into a Knit2 repository
426
def __init__(self, to_repository, from_repository, last_revision=None,
427
pb=None, find_ghosts=True):
428
self.helper = Inter1and2Helper(from_repository, to_repository)
429
GenericRepoFetcher.__init__(self, to_repository, from_repository,
430
last_revision, pb, find_ghosts)
432
def _generate_root_texts(self, revs):
433
self.helper.generate_root_texts(revs)
435
def _fetch_inventory_weave(self, revs, pb):
436
self.helper.regenerate_inventory(revs)
438
def _copy_revision(self, rev):
439
self.helper.fetch_revisions([rev])
442
class Knit1to2Fetcher(KnitRepoFetcher):
443
"""Fetch from a Knit1 repository into a Knit2 repository"""
445
def __init__(self, to_repository, from_repository, last_revision=None,
446
pb=None, find_ghosts=True):
447
self.helper = Inter1and2Helper(from_repository, to_repository)
448
KnitRepoFetcher.__init__(self, to_repository, from_repository,
449
last_revision, pb, find_ghosts)
451
def _generate_root_texts(self, revs):
452
self.helper.generate_root_texts(revs)
454
def _fetch_inventory_weave(self, revs, pb):
455
self.helper.regenerate_inventory(revs)
457
def _fetch_just_revision_texts(self, version_ids):
458
self.helper.fetch_revisions(version_ids)
added
removed
bzrlib/fetch.py
