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 finally actually join
28
from bzrlib.lazy_import import lazy_import
29
lazy_import(globals(), """
35
import bzrlib.errors as errors
36
from bzrlib.errors import (InstallFailed,
38
from bzrlib.progress import ProgressPhase
40
39
from bzrlib.revision import NULL_REVISION
40
from bzrlib.symbol_versioning import (deprecated_function,
41
44
from bzrlib.trace import mutter
47
from bzrlib.lazy_import import lazy_import
48
lazy_import(globals(), '''
49
from bzrlib.remote import RemoteRepository
52
# TODO: Avoid repeatedly opening weaves so many times.
54
# XXX: This doesn't handle ghost (not present in branch) revisions at
55
# all yet. I'm not sure they really should be supported.
57
# NOTE: This doesn't copy revisions which may be present but not
58
# merged into the last revision. I'm not sure we want to do that.
60
# - get a list of revisions that need to be pulled in
61
# - for each one, pull in that revision file
62
# and get the inventory, and store the inventory with right
64
# - and get the ancestry, and store that with right parents too
65
# - and keep a note of all file ids and version seen
66
# - then go through all files; for each one get the weave,
67
# and add in all file versions
70
@deprecated_function(zero_eight)
71
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
72
"""Legacy API, please see branch.fetch(from_branch, last_revision, pb)."""
73
f = Fetcher(to_branch, from_branch, revision, pb)
74
return f.count_copied, f.failed_revisions
44
79
class RepoFetcher(object):
45
80
"""Pull revisions and texts from one repository to another.
47
This should not be used directly, it's essential a object to encapsulate
83
if set, try to limit to the data this revision references.
86
count_copied -- number of revisions copied
88
This should not be used directory, its essential a object to encapsulate
48
89
the logic in InterRepository.fetch().
51
def __init__(self, to_repository, from_repository, last_revision=None,
52
find_ghosts=True, fetch_spec=None):
53
"""Create a repo fetcher.
55
:param last_revision: If set, try to limit to the data this revision
57
: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.
91
def __init__(self, to_repository, from_repository, last_revision=None, pb=None):
93
self.failed_revisions = []
95
if isinstance(to_repository, RemoteRepository):
96
to_repository._ensure_real()
97
to_repository = to_repository._real_repository
98
if isinstance(from_repository, RemoteRepository):
99
from_repository._ensure_real()
100
from_repository = from_repository._real_repository
101
if to_repository.control_files._transport.base == from_repository.control_files._transport.base:
102
# check that last_revision is in 'from' and then return a no-operation.
103
if last_revision not in (None, NULL_REVISION):
104
from_repository.get_revision(last_revision)
61
106
self.to_repository = to_repository
62
107
self.from_repository = from_repository
63
self.sink = to_repository._get_sink()
64
108
# must not mutate self._last_revision as its potentially a shared instance
65
109
self._last_revision = last_revision
66
self._fetch_spec = fetch_spec
67
self.find_ghosts = find_ghosts
111
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
112
self.nested_pb = self.pb
115
self.nested_pb = None
68
116
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)
118
self.to_repository.lock_write()
122
if self.nested_pb is not None:
123
self.nested_pb.finished()
124
self.to_repository.unlock()
75
126
self.from_repository.unlock()
77
128
def __fetch(self):
78
129
"""Primary worker function.
80
This initialises all the needed variables, and then fetches the
131
This initialises all the needed variables, and then fetches the
81
132
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))
134
self.to_weaves = self.to_repository.weave_store
135
self.to_control = self.to_repository.control_weaves
136
self.from_weaves = self.from_repository.weave_store
137
self.from_control = self.from_repository.control_weaves
89
138
self.count_total = 0
90
139
self.file_ids_names = {}
91
pb = ui.ui_factory.nested_progress_bar()
92
pb.show_pct = pb.show_count = False
94
pb.update("Finding revisions", 0, 2)
95
search = self._revids_to_fetch()
98
pb.update("Fetching revisions", 1, 2)
99
self._fetch_everything_for_search(search)
103
def _fetch_everything_for_search(self, search):
104
"""Fetch all data for the given set of revisions."""
105
# The first phase is "file". We pass the progress bar for it directly
106
# into item_keys_introduced_by, which has more information about how
107
# that phase is progressing than we do. Progress updates for the other
108
# phases are taken care of in this function.
109
# XXX: there should be a clear owner of the progress reporting. Perhaps
110
# item_keys_introduced_by should have a richer API than it does at the
111
# 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()
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")
140
pp = ProgressPhase('Fetch phase', 4, self.pb)
143
revs = self._revids_to_fetch()
147
self._fetch_weave_texts(revs)
149
self._fetch_inventory_weave(revs)
151
self._fetch_revision_texts(revs)
152
self.count_copied += len(revs)
150
156
def _revids_to_fetch(self):
151
"""Determines the exact revisions needed from self.from_repository to
152
install self._last_revision in self.to_repository.
154
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
157
mutter('fetch up to rev {%s}', self._last_revision)
159
158
if self._last_revision is NULL_REVISION:
160
159
# 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)
161
if (self._last_revision is not None and
162
self.to_repository.has_revision(self._last_revision)):
166
return self.to_repository.missing_revision_ids(self.from_repository,
168
except errors.NoSuchRevision:
169
raise InstallFailed([self._last_revision])
171
def _fetch_weave_texts(self, revs):
172
texts_pb = bzrlib.ui.ui_factory.nested_progress_bar()
174
# fileids_altered_by_revision_ids requires reading the inventory
175
# weave, we will need to read the inventory weave again when
176
# all this is done, so enable caching for that specific weave
177
inv_w = self.from_repository.get_inventory_weave()
179
file_ids = self.from_repository.fileids_altered_by_revision_ids(revs)
181
num_file_ids = len(file_ids)
182
for file_id, required_versions in file_ids.items():
183
texts_pb.update("fetch texts", count, num_file_ids)
185
to_weave = self.to_weaves.get_weave_or_empty(file_id,
186
self.to_repository.get_transaction())
187
from_weave = self.from_weaves.get_weave(file_id,
188
self.from_repository.get_transaction())
189
# we fetch all the texts, because texts do
190
# not reference anything, and its cheap enough
191
to_weave.join(from_weave, version_ids=required_versions)
192
# we don't need *all* of this data anymore, but we dont know
193
# what we do. This cache clearing will result in a new read
194
# of the knit data when we do the checkout, but probably we
195
# want to emit the needed data on the fly rather than at the
197
# the from weave should know not to cache data being joined,
198
# but its ok to ask it to clear.
199
from_weave.clear_cache()
200
to_weave.clear_cache()
204
def _fetch_inventory_weave(self, revs):
205
pb = bzrlib.ui.ui_factory.nested_progress_bar()
207
pb.update("fetch inventory", 0, 2)
208
to_weave = self.to_control.get_weave('inventory',
209
self.to_repository.get_transaction())
211
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
213
# just merge, this is optimisable and its means we don't
214
# copy unreferenced data such as not-needed inventories.
215
pb.update("fetch inventory", 1, 3)
216
from_weave = self.from_repository.get_inventory_weave()
217
pb.update("fetch inventory", 2, 3)
218
# we fetch only the referenced inventories because we do not
219
# know for unselected inventories whether all their required
220
# texts are present in the other repository - it could be
222
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
224
from_weave.clear_cache()
231
class GenericRepoFetcher(RepoFetcher):
232
"""This is a generic repo to repo fetcher.
234
This makes minimal assumptions about repo layout and contents.
235
It triggers a reconciliation after fetching to ensure integrity.
238
def _fetch_revision_texts(self, revs):
239
"""Fetch revision object texts"""
240
rev_pb = bzrlib.ui.ui_factory.nested_progress_bar()
242
to_txn = self.to_transaction = self.to_repository.get_transaction()
245
to_store = self.to_repository._revision_store
247
pb = bzrlib.ui.ui_factory.nested_progress_bar()
249
pb.update('copying revisions', count, total)
251
sig_text = self.from_repository.get_signature_text(rev)
252
to_store.add_revision_signature_text(rev, sig_text, to_txn)
253
except errors.NoSuchRevision:
256
to_store.add_revision(self.from_repository.get_revision(rev),
261
# fixup inventory if needed:
262
# this is expensive because we have no inverse index to current ghosts.
263
# but on local disk its a few seconds and sftp push is already insane.
265
# FIXME: repository should inform if this is needed.
266
self.to_repository.reconcile()
271
class KnitRepoFetcher(RepoFetcher):
272
"""This is a knit format repository specific fetcher.
274
This differs from the GenericRepoFetcher by not doing a
275
reconciliation after copying, and using knit joining to
279
def _fetch_revision_texts(self, revs):
280
# may need to be a InterRevisionStore call here.
281
from_transaction = self.from_repository.get_transaction()
282
to_transaction = self.to_repository.get_transaction()
283
to_sf = self.to_repository._revision_store.get_signature_file(
285
from_sf = self.from_repository._revision_store.get_signature_file(
287
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
288
to_rf = self.to_repository._revision_store.get_revision_file(
290
from_rf = self.from_repository._revision_store.get_revision_file(
292
to_rf.join(from_rf, version_ids=revs)
179
295
class Inter1and2Helper(object):
180
296
"""Helper for operations that convert data from model 1 and 2
182
298
This is for use by fetchers and converters.
185
def __init__(self, source):
301
def __init__(self, source, target):
188
304
:param source: The repository data comes from
305
:param target: The repository data goes to
190
307
self.source = source
192
310
def iter_rev_trees(self, revs):
193
311
"""Iterate through RevisionTrees efficiently.
209
325
revs = revs[100:]
211
def _find_root_ids(self, revs, parent_map, graph):
213
for tree in self.iter_rev_trees(revs):
214
revision_id = tree.inventory.root.revision
215
root_id = tree.get_root_id()
216
revision_root[revision_id] = root_id
217
# Find out which parents we don't already know root ids for
219
for revision_parents in parent_map.itervalues():
220
parents.update(revision_parents)
221
parents.difference_update(revision_root.keys() + [NULL_REVISION])
222
# Limit to revisions present in the versionedfile
223
parents = graph.get_parent_map(parents).keys()
224
for tree in self.iter_rev_trees(parents):
225
root_id = tree.get_root_id()
226
revision_root[tree.get_revision_id()] = root_id
229
327
def generate_root_texts(self, revs):
230
328
"""Generate VersionedFiles for all root ids.
232
330
:param revs: the revisions to include
234
graph = self.source.get_graph()
235
parent_map = graph.get_parent_map(revs)
236
rev_order = tsort.topo_sort(parent_map)
237
rev_id_to_root_id = self._find_root_ids(revs, parent_map, graph)
238
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
240
# Guaranteed stable, this groups all the file id operations together
241
# retaining topological order within the revisions of a file id.
242
# File id splits and joins would invalidate this, but they don't exist
243
# yet, and are unlikely to in non-rich-root environments anyway.
244
root_id_order.sort(key=operator.itemgetter(0))
245
# 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]
332
inventory_weave = self.source.get_inventory_weave()
335
to_store = self.target.weave_store
336
for tree in self.iter_rev_trees(revs):
337
revision_id = tree.inventory.root.revision
338
root_id = tree.inventory.root.file_id
339
parents = inventory_weave.get_parents(revision_id)
340
if root_id not in versionedfile:
341
versionedfile[root_id] = to_store.get_weave_or_empty(root_id,
342
self.target.get_transaction())
343
parent_texts[root_id] = versionedfile[root_id].add_lines(
344
revision_id, parents, [], parent_texts)
346
def regenerate_inventory(self, revs):
347
"""Generate a new inventory versionedfile in target, convertin data.
349
The inventory is retrieved from the source, (deserializing it), and
350
stored in the target (reserializing it in a different format).
351
:param revs: The revisions to include
353
inventory_weave = self.source.get_inventory_weave()
354
for tree in self.iter_rev_trees(revs):
355
parents = inventory_weave.get_parents(tree.get_revision_id())
356
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
360
class Model1toKnit2Fetcher(GenericRepoFetcher):
361
"""Fetch from a Model1 repository into a Knit2 repository
363
def __init__(self, to_repository, from_repository, last_revision=None,
365
self.helper = Inter1and2Helper(from_repository, to_repository)
366
GenericRepoFetcher.__init__(self, to_repository, from_repository,
369
def _fetch_weave_texts(self, revs):
370
GenericRepoFetcher._fetch_weave_texts(self, revs)
371
# Now generate a weave for the tree root
372
self.helper.generate_root_texts(revs)
374
def _fetch_inventory_weave(self, revs):
375
self.helper.regenerate_inventory(revs)
378
class Knit1to2Fetcher(KnitRepoFetcher):
379
"""Fetch from a Knit1 repository into a Knit2 repository"""
381
def __init__(self, to_repository, from_repository, last_revision=None,
383
self.helper = Inter1and2Helper(from_repository, to_repository)
384
KnitRepoFetcher.__init__(self, to_repository, from_repository,
387
def _fetch_weave_texts(self, revs):
388
KnitRepoFetcher._fetch_weave_texts(self, revs)
389
# Now generate a weave for the tree root
390
self.helper.generate_root_texts(revs)
392
def _fetch_inventory_weave(self, revs):
393
self.helper.regenerate_inventory(revs)
396
class Fetcher(object):
397
"""Backwards compatibility glue for branch.fetch()."""
399
@deprecated_method(zero_eight)
400
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
401
"""Please see branch.fetch()."""
402
to_branch.fetch(from_branch, last_revision, pb)