21
21
that has merged into it. As the first step of a merge, pull, or
22
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.
28
from bzrlib.lazy_import import lazy_import
29
lazy_import(globals(), """
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
40
from bzrlib.revision import NULL_REVISION
44
41
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
45
class RepoFetcher(object):
67
46
"""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
48
This should not be used directly, it's essential a object to encapsulate
76
49
the logic in InterRepository.fetch().
79
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
52
def __init__(self, to_repository, from_repository, last_revision=None,
53
pb=None, find_ghosts=True, fetch_spec=None):
81
54
"""Create a repo fetcher.
56
:param last_revision: If set, try to limit to the data this revision
83
58
:param find_ghosts: If True search the entire history for ghosts.
59
:param pb: ProgressBar object to use; deprecated and ignored.
60
This method will just create one on top of the stack.
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))
63
symbol_versioning.warn(
64
symbol_versioning.deprecated_in((1, 14, 0))
65
% "pb parameter to RepoFetcher.__init__")
66
# and for simplicity it is in fact ignored
67
# repository.fetch has the responsibility for short-circuiting
68
# attempts to copy between a repository and itself.
93
69
self.to_repository = to_repository
94
70
self.from_repository = from_repository
71
self.sink = to_repository._get_sink()
95
72
# must not mutate self._last_revision as its potentially a shared instance
96
73
self._last_revision = last_revision
74
self._fetch_spec = fetch_spec
97
75
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
76
self.from_repository.lock_read()
77
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
78
self.from_repository, self.from_repository._format,
79
self.to_repository, self.to_repository._format)
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
83
self.from_repository.unlock()
125
85
def __fetch(self):
126
86
"""Primary worker function.
128
This initialises all the needed variables, and then fetches the
88
This initialises all the needed variables, and then fetches the
129
89
requested revisions, finally clearing the progress bar.
91
# Roughly this is what we're aiming for fetch to become:
93
# missing = self.sink.insert_stream(self.source.get_stream(search))
95
# missing = self.sink.insert_stream(self.source.get_items(missing))
131
97
self.count_total = 0
132
98
self.file_ids_names = {}
133
pp = ProgressPhase('Transferring', 4, self.pb)
99
pb = bzrlib.ui.ui_factory.nested_progress_bar()
100
pb.show_pct = pb.show_count = False
102
pb.update("Finding revisions", 0, 2)
136
103
search = self._revids_to_fetch()
137
104
if search is None:
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)
106
pb.update("Fetching revisions", 1, 2)
107
self._fetch_everything_for_search(search)
147
def _fetch_everything_for_search(self, search, pp):
111
def _fetch_everything_for_search(self, search):
148
112
"""Fetch all data for the given set of revisions."""
149
113
# The first phase is "file". We pass the progress bar for it directly
150
114
# into item_keys_introduced_by, which has more information about how
154
118
# item_keys_introduced_by should have a richer API than it does at the
155
119
# moment, so that it can feed the progress information back to this
121
if (self.from_repository._format.rich_root_data and
122
not self.to_repository._format.rich_root_data):
123
raise errors.IncompatibleRepositories(
124
self.from_repository, self.to_repository,
125
"different rich-root support")
158
126
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, self.to_repository._fetch_order,
183
not self.to_repository._fetch_uses_deltas))
184
# Cause an error if a text occurs after we have done the
187
# Before we process the inventory we generate the root
188
# texts (if necessary) so that the inventories references
190
self._generate_root_texts(revs)
191
# NB: This currently reopens the inventory weave in source;
192
# using a single stream interface instead would avoid this.
193
self._fetch_inventory_weave(revs, pb)
194
elif knit_kind == "signatures":
195
# Nothing to do here; this will be taken care of when
196
# _fetch_revision_texts happens.
198
elif knit_kind == "revisions":
199
self._fetch_revision_texts(revs, pb)
201
raise AssertionError("Unknown knit kind %r" % knit_kind)
202
if self.to_repository._fetch_reconcile:
203
self.to_repository.reconcile()
128
pb.update("Get stream source")
129
source = self.from_repository._get_source(
130
self.to_repository._format)
131
stream = source.get_stream(search)
132
from_format = self.from_repository._format
133
pb.update("Inserting stream")
134
resume_tokens, missing_keys = self.sink.insert_stream(
135
stream, from_format, [])
136
if self.to_repository._fallback_repositories:
138
self._parent_inventories(search.get_keys()))
140
pb.update("Missing keys")
141
stream = source.get_stream_for_missing_keys(missing_keys)
142
pb.update("Inserting missing keys")
143
resume_tokens, missing_keys = self.sink.insert_stream(
144
stream, from_format, resume_tokens)
146
raise AssertionError(
147
"second push failed to complete a fetch %r." % (
150
raise AssertionError(
151
"second push failed to commit the fetch %r." % (
153
pb.update("Finishing stream")
207
self.count_copied += len(revs)
209
158
def _revids_to_fetch(self):
210
159
"""Determines the exact revisions needed from self.from_repository to
211
160
install self._last_revision in self.to_repository.
213
162
If no revisions need to be fetched, then this just returns None.
164
if self._fetch_spec is not None:
165
return self._fetch_spec
215
166
mutter('fetch up to rev {%s}', self._last_revision)
216
167
if self._last_revision is NULL_REVISION:
217
168
# explicit limit of no revisions needed
219
if (self._last_revision is not None and
220
self.to_repository.has_revision(self._last_revision)):
223
return self.to_repository.search_missing_revision_ids(
224
self.from_repository, self._last_revision,
225
find_ghosts=self.find_ghosts)
226
except errors.NoSuchRevision, e:
227
raise InstallFailed([self._last_revision])
229
def _fetch_inventory_weave(self, revs, pb):
230
pb.update("fetch inventory", 0, 2)
231
to_weave = self.to_repository.inventories
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.inventories
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.insert_record_stream(from_weave.get_record_stream(
244
[(rev_id,) for rev_id in revs],
245
self.to_repository._fetch_order,
246
not self.to_repository._fetch_uses_deltas))
250
def _fetch_revision_texts(self, revs, pb):
251
# may need to be a InterRevisionStore call here.
252
to_sf = self.to_repository.signatures
253
from_sf = self.from_repository.signatures
254
# A missing signature is just skipped.
255
to_sf.insert_record_stream(filter_absent(from_sf.get_record_stream(
256
[(rev_id,) for rev_id in revs],
257
self.to_repository._fetch_order,
259
# Bug #261339, some knit repositories accidentally had deltas in their
260
# revision stream, when you weren't ever supposed to have deltas.
261
# So we now *force* fulltext copying for signatures and revisions
262
self._fetch_just_revision_texts(revs)
264
def _fetch_just_revision_texts(self, version_ids):
265
to_rf = self.to_repository.revisions
266
from_rf = self.from_repository.revisions
267
to_rf.insert_record_stream(from_rf.get_record_stream(
268
[(rev_id,) for rev_id in version_ids],
269
self.to_repository._fetch_order,
271
# Bug #261339, some knit repositories accidentally had deltas in their
272
# revision stream, when you weren't ever supposed to have deltas.
273
# So we now *force* fulltext copying for signatures and revisions
275
def _generate_root_texts(self, revs):
276
"""This will be called by __fetch between fetching weave texts and
277
fetching the inventory weave.
279
Subclasses should override this if they need to generate root texts
280
after fetching weave texts.
170
return self.to_repository.search_missing_revision_ids(
171
self.from_repository, self._last_revision,
172
find_ghosts=self.find_ghosts)
174
def _parent_inventories(self, revision_ids):
175
# Find all the parent revisions referenced by the stream, but
176
# not present in the stream, and make sure we send their
178
parent_maps = self.to_repository.get_parent_map(revision_ids)
180
map(parents.update, parent_maps.itervalues())
181
parents.discard(NULL_REVISION)
182
parents.difference_update(revision_ids)
183
missing_keys = set(('inventories', rev_id) for rev_id in parents)
285
187
class Inter1and2Helper(object):
286
188
"""Helper for operations that convert data from model 1 and 2
288
190
This is for use by fetchers and converters.
291
def __init__(self, source, target):
193
def __init__(self, source):
294
196
:param source: The repository data comes from
295
:param target: The repository data goes to
297
198
self.source = source
300
200
def iter_rev_trees(self, revs):
301
201
"""Iterate through RevisionTrees efficiently.
355
251
# yet, and are unlikely to in non-rich-root environments anyway.
356
252
root_id_order.sort(key=operator.itemgetter(0))
357
253
# Create a record stream containing the roots to create.
359
for key in root_id_order:
360
root_id, rev_id = key
361
rev_parents = parent_map[rev_id]
362
# We drop revision parents with different file-ids, because
363
# that represents a rename of the root to a different location
364
# - its not actually a parent for us. (We could look for that
365
# file id in the revision tree at considerably more expense,
366
# but for now this is sufficient (and reconcile will catch and
367
# correct this anyway).
368
# When a parent revision is a ghost, we guess that its root id
369
# was unchanged (rather than trimming it from the parent list).
370
parent_keys = tuple((root_id, parent) for parent in rev_parents
371
if parent != NULL_REVISION and
372
rev_id_to_root_id.get(parent, root_id) == root_id)
373
yield FulltextContentFactory(key, parent_keys, None, '')
374
to_texts.insert_record_stream(yield_roots())
376
def regenerate_inventory(self, revs):
377
"""Generate a new inventory versionedfile in target, convertin data.
379
The inventory is retrieved from the source, (deserializing it), and
380
stored in the target (reserializing it in a different format).
381
:param revs: The revisions to include
383
for tree in self.iter_rev_trees(revs):
384
parents = tree.get_parent_ids()
385
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
388
def fetch_revisions(self, revision_ids):
389
# TODO: should this batch them up rather than requesting 10,000
391
for revision in self.source.get_revisions(revision_ids):
392
self.target.add_revision(revision.revision_id, revision)
395
class Model1toKnit2Fetcher(RepoFetcher):
396
"""Fetch from a Model1 repository into a Knit2 repository
398
def __init__(self, to_repository, from_repository, last_revision=None,
399
pb=None, find_ghosts=True):
400
self.helper = Inter1and2Helper(from_repository, to_repository)
401
RepoFetcher.__init__(self, to_repository, from_repository,
402
last_revision, pb, find_ghosts)
404
def _generate_root_texts(self, revs):
405
self.helper.generate_root_texts(revs)
407
def _fetch_inventory_weave(self, revs, pb):
408
self.helper.regenerate_inventory(revs)
410
def _fetch_revision_texts(self, revs, pb):
411
"""Fetch revision object texts"""
415
pb.update('copying revisions', count, total)
417
sig_text = self.from_repository.get_signature_text(rev)
418
self.to_repository.add_signature_text(rev, sig_text)
419
except errors.NoSuchRevision:
254
from bzrlib.graph import FrozenHeadsCache
255
graph = FrozenHeadsCache(graph)
256
new_roots_stream = _new_root_data_stream(
257
root_id_order, rev_id_to_root_id, parent_map, self.source, graph)
258
return [('texts', new_roots_stream)]
261
def _new_root_data_stream(
262
root_keys_to_create, rev_id_to_root_id_map, parent_map, repo, graph=None):
263
"""Generate a texts substream of synthesised root entries.
265
Used in fetches that do rich-root upgrades.
267
:param root_keys_to_create: iterable of (root_id, rev_id) pairs describing
268
the root entries to create.
269
:param rev_id_to_root_id_map: dict of known rev_id -> root_id mappings for
270
calculating the parents. If a parent rev_id is not found here then it
271
will be recalculated.
272
:param parent_map: a parent map for all the revisions in
274
:param graph: a graph to use instead of repo.get_graph().
276
for root_key in root_keys_to_create:
277
root_id, rev_id = root_key
278
parent_keys = _parent_keys_for_root_version(
279
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph)
280
yield versionedfile.FulltextContentFactory(
281
root_key, parent_keys, None, '')
284
def _parent_keys_for_root_version(
285
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph=None):
286
"""Get the parent keys for a given root id.
288
A helper function for _new_root_data_stream.
290
# Include direct parents of the revision, but only if they used the same
291
# root_id and are heads.
292
rev_parents = parent_map[rev_id]
294
for parent_id in rev_parents:
295
if parent_id == NULL_REVISION:
297
if parent_id not in rev_id_to_root_id_map:
298
# We probably didn't read this revision, go spend the extra effort
301
tree = repo.revision_tree(parent_id)
302
except errors.NoSuchRevision:
303
# Ghost, fill out rev_id_to_root_id in case we encounter this
305
# But set parent_root_id to None since we don't really know
306
parent_root_id = None
308
parent_root_id = tree.get_root_id()
309
rev_id_to_root_id_map[parent_id] = None
311
# rev_id_to_root_id_map[parent_id] = parent_root_id
312
# memory consumption maybe?
314
parent_root_id = rev_id_to_root_id_map[parent_id]
315
if root_id == parent_root_id:
316
# With stacking we _might_ want to refer to a non-local revision,
317
# but this code path only applies when we have the full content
318
# available, so ghosts really are ghosts, not just the edge of
320
parent_ids.append(parent_id)
322
# root_id may be in the parent anyway.
324
tree = repo.revision_tree(parent_id)
325
except errors.NoSuchRevision:
326
# ghost, can't refer to it.
422
self._copy_revision(rev)
425
def _copy_revision(self, rev):
426
self.helper.fetch_revisions([rev])
429
class Knit1to2Fetcher(RepoFetcher):
430
"""Fetch from a Knit1 repository into a Knit2 repository"""
432
def __init__(self, to_repository, from_repository, last_revision=None,
433
pb=None, find_ghosts=True):
434
self.helper = Inter1and2Helper(from_repository, to_repository)
435
RepoFetcher.__init__(self, to_repository, from_repository,
436
last_revision, pb, find_ghosts)
438
def _generate_root_texts(self, revs):
439
self.helper.generate_root_texts(revs)
441
def _fetch_inventory_weave(self, revs, pb):
442
self.helper.regenerate_inventory(revs)
444
def _fetch_just_revision_texts(self, version_ids):
445
self.helper.fetch_revisions(version_ids)
330
parent_ids.append(tree.inventory[root_id].revision)
331
except errors.NoSuchId:
334
# Drop non-head parents
336
graph = repo.get_graph()
337
heads = graph.heads(parent_ids)
339
for parent_id in parent_ids:
340
if parent_id in heads and parent_id not in selected_ids:
341
selected_ids.append(parent_id)
342
parent_keys = [(root_id, parent_id) for parent_id in selected_ids]