1
# Copyright (C) 2005, 2006 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 finally actually join
35
import bzrlib.errors as errors
36
from bzrlib.errors import InstallFailed
37
from bzrlib.progress import ProgressPhase
38
from bzrlib.revision import is_null, NULL_REVISION
39
from bzrlib.symbol_versioning import (deprecated_function,
42
from bzrlib.trace import mutter
45
from bzrlib.lazy_import import lazy_import
47
# TODO: Avoid repeatedly opening weaves so many times.
49
# XXX: This doesn't handle ghost (not present in branch) revisions at
50
# all yet. I'm not sure they really should be supported.
52
# NOTE: This doesn't copy revisions which may be present but not
53
# merged into the last revision. I'm not sure we want to do that.
55
# - get a list of revisions that need to be pulled in
56
# - for each one, pull in that revision file
57
# and get the inventory, and store the inventory with right
59
# - and get the ancestry, and store that with right parents too
60
# - and keep a note of all file ids and version seen
61
# - then go through all files; for each one get the weave,
62
# and add in all file versions
65
class RepoFetcher(object):
66
"""Pull revisions and texts from one repository to another.
69
if set, try to limit to the data this revision references.
72
count_copied -- number of revisions copied
74
This should not be used directly, it's essential a object to encapsulate
75
the logic in InterRepository.fetch().
78
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
80
"""Create a repo fetcher.
82
:param find_ghosts: If True search the entire history for ghosts.
85
self.failed_revisions = []
87
if to_repository.has_same_location(from_repository):
88
# repository.fetch should be taking care of this case.
89
raise errors.BzrError('RepoFetcher run '
90
'between two objects at the same location: '
91
'%r and %r' % (to_repository, from_repository))
92
self.to_repository = to_repository
93
self.from_repository = from_repository
94
# must not mutate self._last_revision as its potentially a shared instance
95
self._last_revision = last_revision
96
self.find_ghosts = find_ghosts
98
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
99
self.nested_pb = self.pb
102
self.nested_pb = None
103
self.from_repository.lock_read()
105
self.to_repository.lock_write()
107
self.to_repository.start_write_group()
111
self.to_repository.abort_write_group()
114
self.to_repository.commit_write_group()
116
if self.nested_pb is not None:
117
self.nested_pb.finished()
118
self.to_repository.unlock()
120
self.from_repository.unlock()
123
"""Primary worker function.
125
This initialises all the needed variables, and then fetches the
126
requested revisions, finally clearing the progress bar.
128
self.to_weaves = self.to_repository.weave_store
129
self.from_weaves = self.from_repository.weave_store
131
self.file_ids_names = {}
132
pp = ProgressPhase('Transferring', 4, self.pb)
135
search = self._revids_to_fetch()
138
if getattr(self, '_fetch_everything_for_search', None) is not None:
139
self._fetch_everything_for_search(search, pp)
141
# backward compatibility
142
self._fetch_everything_for_revisions(search.get_keys, pp)
146
def _fetch_everything_for_search(self, search, pp):
147
"""Fetch all data for the given set of revisions."""
148
# The first phase is "file". We pass the progress bar for it directly
149
# into item_keys_introduced_by, which has more information about how
150
# that phase is progressing than we do. Progress updates for the other
151
# phases are taken care of in this function.
152
# XXX: there should be a clear owner of the progress reporting. Perhaps
153
# item_keys_introduced_by should have a richer API than it does at the
154
# moment, so that it can feed the progress information back to this
157
pb = bzrlib.ui.ui_factory.nested_progress_bar()
159
revs = search.get_keys()
160
data_to_fetch = self.from_repository.item_keys_introduced_by(revs, pb)
161
for knit_kind, file_id, revisions in data_to_fetch:
162
if knit_kind != phase:
164
# Make a new progress bar for this phase
167
pb = bzrlib.ui.ui_factory.nested_progress_bar()
168
if knit_kind == "file":
169
self._fetch_weave_text(file_id, revisions)
170
elif knit_kind == "inventory":
171
# Before we process the inventory we generate the root
172
# texts (if necessary) so that the inventories references
174
self._generate_root_texts(revs)
175
# NB: This currently reopens the inventory weave in source;
176
# using a full get_data_stream instead would avoid this.
177
self._fetch_inventory_weave(revs, pb)
178
elif knit_kind == "signatures":
179
# Nothing to do here; this will be taken care of when
180
# _fetch_revision_texts happens.
182
elif knit_kind == "revisions":
183
self._fetch_revision_texts(revs, pb)
185
raise AssertionError("Unknown knit kind %r" % knit_kind)
189
self.count_copied += len(revs)
191
def _revids_to_fetch(self):
192
"""Determines the exact revisions needed from self.from_repository to
193
install self._last_revision in self.to_repository.
195
If no revisions need to be fetched, then this just returns None.
197
mutter('fetch up to rev {%s}', self._last_revision)
198
if self._last_revision is NULL_REVISION:
199
# explicit limit of no revisions needed
201
if (self._last_revision is not None and
202
self.to_repository.has_revision(self._last_revision)):
205
return self.to_repository.search_missing_revision_ids(
206
self.from_repository, self._last_revision,
207
find_ghosts=self.find_ghosts)
208
except errors.NoSuchRevision:
209
raise InstallFailed([self._last_revision])
211
def _fetch_weave_text(self, file_id, required_versions):
212
to_weave = self.to_weaves.get_weave_or_empty(file_id,
213
self.to_repository.get_transaction())
214
from_weave = self.from_weaves.get_weave(file_id,
215
self.from_repository.get_transaction())
216
# we fetch all the texts, because texts do
217
# not reference anything, and its cheap enough
218
to_weave.join(from_weave, version_ids=required_versions)
219
# we don't need *all* of this data anymore, but we dont know
220
# what we do. This cache clearing will result in a new read
221
# of the knit data when we do the checkout, but probably we
222
# want to emit the needed data on the fly rather than at the
224
# the from weave should know not to cache data being joined,
225
# but its ok to ask it to clear.
226
from_weave.clear_cache()
227
to_weave.clear_cache()
229
def _fetch_inventory_weave(self, revs, pb):
230
pb.update("fetch inventory", 0, 2)
231
to_weave = self.to_repository.get_inventory_weave()
232
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
234
# just merge, this is optimisable and its means we don't
235
# copy unreferenced data such as not-needed inventories.
236
pb.update("fetch inventory", 1, 3)
237
from_weave = self.from_repository.get_inventory_weave()
238
pb.update("fetch inventory", 2, 3)
239
# we fetch only the referenced inventories because we do not
240
# know for unselected inventories whether all their required
241
# texts are present in the other repository - it could be
243
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
245
from_weave.clear_cache()
249
def _generate_root_texts(self, revs):
250
"""This will be called by __fetch between fetching weave texts and
251
fetching the inventory weave.
253
Subclasses should override this if they need to generate root texts
254
after fetching weave texts.
259
class GenericRepoFetcher(RepoFetcher):
260
"""This is a generic repo to repo fetcher.
262
This makes minimal assumptions about repo layout and contents.
263
It triggers a reconciliation after fetching to ensure integrity.
266
def _fetch_revision_texts(self, revs, pb):
267
"""Fetch revision object texts"""
268
to_txn = self.to_transaction = self.to_repository.get_transaction()
271
to_store = self.to_repository._revision_store
273
pb.update('copying revisions', count, total)
275
sig_text = self.from_repository.get_signature_text(rev)
276
to_store.add_revision_signature_text(rev, sig_text, to_txn)
277
except errors.NoSuchRevision:
280
to_store.add_revision(self.from_repository.get_revision(rev),
283
# fixup inventory if needed:
284
# this is expensive because we have no inverse index to current ghosts.
285
# but on local disk its a few seconds and sftp push is already insane.
287
# FIXME: repository should inform if this is needed.
288
self.to_repository.reconcile()
291
class KnitRepoFetcher(RepoFetcher):
292
"""This is a knit format repository specific fetcher.
294
This differs from the GenericRepoFetcher by not doing a
295
reconciliation after copying, and using knit joining to
299
def _fetch_revision_texts(self, revs, pb):
300
# may need to be a InterRevisionStore call here.
301
from_transaction = self.from_repository.get_transaction()
302
to_transaction = self.to_repository.get_transaction()
303
to_sf = self.to_repository._revision_store.get_signature_file(
305
from_sf = self.from_repository._revision_store.get_signature_file(
307
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
308
to_rf = self.to_repository._revision_store.get_revision_file(
310
from_rf = self.from_repository._revision_store.get_revision_file(
312
to_rf.join(from_rf, version_ids=revs)
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 generate_root_texts(self, revs):
350
"""Generate VersionedFiles for all root ids.
352
:param revs: the revisions to include
354
inventory_weave = self.source.get_inventory_weave()
357
to_store = self.target.weave_store
358
parent_map = self.source.get_graph().get_parent_map(revs)
359
for tree in self.iter_rev_trees(revs):
360
revision_id = tree.inventory.root.revision
361
root_id = tree.get_root_id()
362
parents = parent_map[revision_id]
363
if parents[0] == NULL_REVISION:
365
if root_id not in versionedfile:
366
versionedfile[root_id] = to_store.get_weave_or_empty(root_id,
367
self.target.get_transaction())
368
_, _, parent_texts[root_id] = versionedfile[root_id].add_lines(
369
revision_id, parents, [], parent_texts)
371
def regenerate_inventory(self, revs):
372
"""Generate a new inventory versionedfile in target, convertin data.
374
The inventory is retrieved from the source, (deserializing it), and
375
stored in the target (reserializing it in a different format).
376
:param revs: The revisions to include
378
for tree in self.iter_rev_trees(revs):
379
parents = tree.get_parent_ids()
380
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
384
class Model1toKnit2Fetcher(GenericRepoFetcher):
385
"""Fetch from a Model1 repository into a Knit2 repository
387
def __init__(self, to_repository, from_repository, last_revision=None,
388
pb=None, find_ghosts=True):
389
self.helper = Inter1and2Helper(from_repository, to_repository)
390
GenericRepoFetcher.__init__(self, to_repository, from_repository,
391
last_revision, pb, find_ghosts)
393
def _generate_root_texts(self, revs):
394
self.helper.generate_root_texts(revs)
396
def _fetch_inventory_weave(self, revs, pb):
397
self.helper.regenerate_inventory(revs)
400
class Knit1to2Fetcher(KnitRepoFetcher):
401
"""Fetch from a Knit1 repository into a Knit2 repository"""
403
def __init__(self, to_repository, from_repository, last_revision=None,
404
pb=None, find_ghosts=True):
405
self.helper = Inter1and2Helper(from_repository, to_repository)
406
KnitRepoFetcher.__init__(self, to_repository, from_repository,
407
last_revision, pb, find_ghosts)
409
def _generate_root_texts(self, revs):
410
self.helper.generate_root_texts(revs)
412
def _fetch_inventory_weave(self, revs, pb):
413
self.helper.regenerate_inventory(revs)
416
class RemoteToOtherFetcher(GenericRepoFetcher):
418
def _fetch_everything_for_search(self, search, pp):
419
data_stream = self.from_repository.get_data_stream_for_search(search)
420
self.to_repository.insert_data_stream(data_stream)
added
removed
bzrlib/fetch.py
