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(), """
40
from bzrlib.revision import NULL_REVISION
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
41
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
44
66
class RepoFetcher(object):
45
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
47
75
This should not be used directly, it's essential a object to encapsulate
48
76
the logic in InterRepository.fetch().
51
def __init__(self, to_repository, from_repository, last_revision=None,
52
find_ghosts=True, fetch_spec=None):
79
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
53
81
"""Create a repo fetcher.
55
:param last_revision: If set, try to limit to the data this revision
57
83
:param find_ghosts: If True search the entire history for ghosts.
59
# repository.fetch has the responsibility for short-circuiting
60
# attempts to copy between a repository and itself.
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))
61
93
self.to_repository = to_repository
62
94
self.from_repository = from_repository
63
self.sink = to_repository._get_sink()
64
95
# must not mutate self._last_revision as its potentially a shared instance
65
96
self._last_revision = last_revision
66
self._fetch_spec = fetch_spec
67
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
68
104
self.from_repository.lock_read()
69
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
70
self.from_repository, self.from_repository._format,
71
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()
75
123
self.from_repository.unlock()
77
125
def __fetch(self):
78
126
"""Primary worker function.
80
This initialises all the needed variables, and then fetches the
128
This initialises all the needed variables, and then fetches the
81
129
requested revisions, finally clearing the progress bar.
83
# Roughly this is what we're aiming for fetch to become:
85
# missing = self.sink.insert_stream(self.source.get_stream(search))
87
# missing = self.sink.insert_stream(self.source.get_items(missing))
89
131
self.count_total = 0
90
132
self.file_ids_names = {}
91
pb = ui.ui_factory.nested_progress_bar()
92
pb.show_pct = pb.show_count = False
133
pp = ProgressPhase('Transferring', 4, self.pb)
94
pb.update("Finding revisions", 0, 2)
95
136
search = self._revids_to_fetch()
96
137
if search is None:
98
pb.update("Fetching revisions", 1, 2)
99
self._fetch_everything_for_search(search)
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)
103
def _fetch_everything_for_search(self, search):
147
def _fetch_everything_for_search(self, search, pp):
104
148
"""Fetch all data for the given set of revisions."""
105
149
# The first phase is "file". We pass the progress bar for it directly
106
150
# into item_keys_introduced_by, which has more information about how
110
154
# item_keys_introduced_by should have a richer API than it does at the
111
155
# moment, so that it can feed the progress information back to this
113
if (self.from_repository._format.rich_root_data and
114
not self.to_repository._format.rich_root_data):
115
raise errors.IncompatibleRepositories(
116
self.from_repository, self.to_repository,
117
"different rich-root support")
118
pb = ui.ui_factory.nested_progress_bar()
158
pb = bzrlib.ui.ui_factory.nested_progress_bar()
120
pb.update("Get stream source")
121
source = self.from_repository._get_source(
122
self.to_repository._format)
123
stream = source.get_stream(search)
124
from_format = self.from_repository._format
125
pb.update("Inserting stream")
126
resume_tokens, missing_keys = self.sink.insert_stream(
127
stream, from_format, [])
128
if self.to_repository._fallback_repositories:
130
self._parent_inventories(search.get_keys()))
132
pb.update("Missing keys")
133
stream = source.get_stream_for_missing_keys(missing_keys)
134
pb.update("Inserting missing keys")
135
resume_tokens, missing_keys = self.sink.insert_stream(
136
stream, from_format, resume_tokens)
138
raise AssertionError(
139
"second push failed to complete a fetch %r." % (
142
raise AssertionError(
143
"second push failed to commit the fetch %r." % (
145
pb.update("Finishing stream")
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)
150
206
def _revids_to_fetch(self):
151
207
"""Determines the exact revisions needed from self.from_repository to
152
208
install self._last_revision in self.to_repository.
154
210
If no revisions need to be fetched, then this just returns None.
156
if self._fetch_spec is not None:
157
return self._fetch_spec
158
212
mutter('fetch up to rev {%s}', self._last_revision)
159
213
if self._last_revision is NULL_REVISION:
160
214
# explicit limit of no revisions needed
162
return self.to_repository.search_missing_revision_ids(
163
self.from_repository, self._last_revision,
164
find_ghosts=self.find_ghosts)
166
def _parent_inventories(self, revision_ids):
167
# Find all the parent revisions referenced by the stream, but
168
# not present in the stream, and make sure we send their
170
parent_maps = self.to_repository.get_parent_map(revision_ids)
172
map(parents.update, parent_maps.itervalues())
173
parents.discard(NULL_REVISION)
174
parents.difference_update(revision_ids)
175
missing_keys = set(('inventories', rev_id) for rev_id in parents)
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))
179
315
class Inter1and2Helper(object):
180
316
"""Helper for operations that convert data from model 1 and 2
182
318
This is for use by fetchers and converters.
185
def __init__(self, source):
321
def __init__(self, source, target):
188
324
:param source: The repository data comes from
325
:param target: The repository data goes to
190
327
self.source = source
192
330
def iter_rev_trees(self, revs):
193
331
"""Iterate through RevisionTrees efficiently.
243
385
# yet, and are unlikely to in non-rich-root environments anyway.
244
386
root_id_order.sort(key=operator.itemgetter(0))
245
387
# Create a record stream containing the roots to create.
247
# XXX: not covered by tests, should have a flag to always run
248
# this. -- mbp 20100129
249
graph = self.source_repo.get_known_graph_ancestry(revs)
250
new_roots_stream = _new_root_data_stream(
251
root_id_order, rev_id_to_root_id, parent_map, self.source, graph)
252
return [('texts', new_roots_stream)]
255
def _get_rich_root_heads_graph(source_repo, revision_ids):
256
"""Get a Graph object suitable for asking heads() for new rich roots."""
260
def _new_root_data_stream(
261
root_keys_to_create, rev_id_to_root_id_map, parent_map, repo, graph=None):
262
"""Generate a texts substream of synthesised root entries.
264
Used in fetches that do rich-root upgrades.
266
:param root_keys_to_create: iterable of (root_id, rev_id) pairs describing
267
the root entries to create.
268
:param rev_id_to_root_id_map: dict of known rev_id -> root_id mappings for
269
calculating the parents. If a parent rev_id is not found here then it
270
will be recalculated.
271
:param parent_map: a parent map for all the revisions in
273
:param graph: a graph to use instead of repo.get_graph().
275
for root_key in root_keys_to_create:
276
root_id, rev_id = root_key
277
parent_keys = _parent_keys_for_root_version(
278
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph)
279
yield versionedfile.FulltextContentFactory(
280
root_key, parent_keys, None, '')
283
def _parent_keys_for_root_version(
284
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph=None):
285
"""Get the parent keys for a given root id.
287
A helper function for _new_root_data_stream.
289
# Include direct parents of the revision, but only if they used the same
290
# root_id and are heads.
291
rev_parents = parent_map[rev_id]
293
for parent_id in rev_parents:
294
if parent_id == NULL_REVISION:
296
if parent_id not in rev_id_to_root_id_map:
297
# We probably didn't read this revision, go spend the extra effort
300
tree = repo.revision_tree(parent_id)
301
except errors.NoSuchRevision:
302
# Ghost, fill out rev_id_to_root_id in case we encounter this
304
# But set parent_root_id to None since we don't really know
305
parent_root_id = None
307
parent_root_id = tree.get_root_id()
308
rev_id_to_root_id_map[parent_id] = None
310
# rev_id_to_root_id_map[parent_id] = parent_root_id
311
# memory consumption maybe?
313
parent_root_id = rev_id_to_root_id_map[parent_id]
314
if root_id == parent_root_id:
315
# With stacking we _might_ want to refer to a non-local revision,
316
# but this code path only applies when we have the full content
317
# available, so ghosts really are ghosts, not just the edge of
319
parent_ids.append(parent_id)
321
# root_id may be in the parent anyway.
323
tree = repo.revision_tree(parent_id)
324
except errors.NoSuchRevision:
325
# ghost, can't refer to it.
329
parent_ids.append(tree.inventory[root_id].revision)
330
except errors.NoSuchId:
333
# Drop non-head parents
335
graph = repo.get_graph()
336
heads = graph.heads(parent_ids)
338
for parent_id in parent_ids:
339
if parent_id in heads and parent_id not in selected_ids:
340
selected_ids.append(parent_id)
341
parent_keys = [(root_id, parent_id) for parent_id in selected_ids]
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
