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.trace import mutter
39
from bzrlib.progress import ProgressPhase
40
40
from bzrlib.revision import NULL_REVISION
41
from bzrlib.trace import mutter
41
from bzrlib.symbol_versioning import (deprecated_function,
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
@deprecated_function(zero_eight)
66
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
67
"""Legacy API, please see branch.fetch(from_branch, last_revision, pb)."""
68
f = Fetcher(to_branch, from_branch, revision, pb)
69
return f.count_copied, f.failed_revisions
45
74
class RepoFetcher(object):
46
75
"""Pull revisions and texts from one repository to another.
48
This should not be used directly, it's essential a object to encapsulate
78
if set, try to limit to the data this revision references.
81
count_copied -- number of revisions copied
83
This should not be used directory, its essential a object to encapsulate
49
84
the logic in InterRepository.fetch().
52
def __init__(self, to_repository, from_repository, last_revision=None,
53
pb=None, find_ghosts=True, fetch_spec=None):
54
"""Create a repo fetcher.
56
:param last_revision: If set, try to limit to the data this revision
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.
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.
86
def __init__(self, to_repository, from_repository, last_revision=None, pb=None):
88
self.failed_revisions = []
90
if to_repository.control_files._transport.base == from_repository.control_files._transport.base:
91
# check that last_revision is in 'from' and then return a no-operation.
92
if last_revision not in (None, NULL_REVISION):
93
from_repository.get_revision(last_revision)
69
95
self.to_repository = to_repository
70
96
self.from_repository = from_repository
71
self.sink = to_repository._get_sink()
72
97
# must not mutate self._last_revision as its potentially a shared instance
73
98
self._last_revision = last_revision
74
self._fetch_spec = fetch_spec
75
self.find_ghosts = find_ghosts
100
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
101
self.nested_pb = self.pb
104
self.nested_pb = None
76
105
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)
107
self.to_repository.lock_write()
111
if self.nested_pb is not None:
112
self.nested_pb.finished()
113
self.to_repository.unlock()
83
115
self.from_repository.unlock()
85
117
def __fetch(self):
86
118
"""Primary worker function.
88
This initialises all the needed variables, and then fetches the
120
This initialises all the needed variables, and then fetches the
89
121
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))
123
self.to_weaves = self.to_repository.weave_store
124
self.to_control = self.to_repository.control_weaves
125
self.from_weaves = self.from_repository.weave_store
126
self.from_control = self.from_repository.control_weaves
97
127
self.count_total = 0
98
128
self.file_ids_names = {}
99
pb = bzrlib.ui.ui_factory.nested_progress_bar()
100
pb.show_pct = pb.show_count = False
102
pb.update("Finding revisions", 0, 2)
103
search = self._revids_to_fetch()
106
pb.update("Fetching revisions", 1, 2)
107
self._fetch_everything_for_search(search)
111
def _fetch_everything_for_search(self, search):
112
"""Fetch all data for the given set of revisions."""
113
# The first phase is "file". We pass the progress bar for it directly
114
# into item_keys_introduced_by, which has more information about how
115
# that phase is progressing than we do. Progress updates for the other
116
# phases are taken care of in this function.
117
# XXX: there should be a clear owner of the progress reporting. Perhaps
118
# item_keys_introduced_by should have a richer API than it does at the
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")
126
pb = bzrlib.ui.ui_factory.nested_progress_bar()
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")
129
pp = ProgressPhase('Fetch phase', 4, self.pb)
132
revs = self._revids_to_fetch()
136
self._fetch_weave_texts(revs)
138
self._fetch_inventory_weave(revs)
140
self._fetch_revision_texts(revs)
141
self.count_copied += len(revs)
158
145
def _revids_to_fetch(self):
159
"""Determines the exact revisions needed from self.from_repository to
160
install self._last_revision in self.to_repository.
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
166
146
mutter('fetch up to rev {%s}', self._last_revision)
167
147
if self._last_revision is NULL_REVISION:
168
148
# explicit limit of no revisions needed
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)
150
if (self._last_revision is not None and
151
self.to_repository.has_revision(self._last_revision)):
155
return self.to_repository.missing_revision_ids(self.from_repository,
157
except errors.NoSuchRevision:
158
raise InstallFailed([self._last_revision])
160
def _fetch_weave_texts(self, revs):
161
texts_pb = bzrlib.ui.ui_factory.nested_progress_bar()
163
# fileids_altered_by_revision_ids requires reading the inventory
164
# weave, we will need to read the inventory weave again when
165
# all this is done, so enable caching for that specific weave
166
inv_w = self.from_repository.get_inventory_weave()
168
file_ids = self.from_repository.fileids_altered_by_revision_ids(revs)
170
num_file_ids = len(file_ids)
171
for file_id, required_versions in file_ids.items():
172
texts_pb.update("fetch texts", count, num_file_ids)
174
to_weave = self.to_weaves.get_weave_or_empty(file_id,
175
self.to_repository.get_transaction())
176
from_weave = self.from_weaves.get_weave(file_id,
177
self.from_repository.get_transaction())
178
# we fetch all the texts, because texts do
179
# not reference anything, and its cheap enough
180
to_weave.join(from_weave, version_ids=required_versions)
181
# we don't need *all* of this data anymore, but we dont know
182
# what we do. This cache clearing will result in a new read
183
# of the knit data when we do the checkout, but probably we
184
# want to emit the needed data on the fly rather than at the
186
# the from weave should know not to cache data being joined,
187
# but its ok to ask it to clear.
188
from_weave.clear_cache()
189
to_weave.clear_cache()
193
def _fetch_inventory_weave(self, revs):
194
pb = bzrlib.ui.ui_factory.nested_progress_bar()
196
pb.update("fetch inventory", 0, 2)
197
to_weave = self.to_control.get_weave('inventory',
198
self.to_repository.get_transaction())
200
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
202
# just merge, this is optimisable and its means we don't
203
# copy unreferenced data such as not-needed inventories.
204
pb.update("fetch inventory", 1, 3)
205
from_weave = self.from_repository.get_inventory_weave()
206
pb.update("fetch inventory", 2, 3)
207
# we fetch only the referenced inventories because we do not
208
# know for unselected inventories whether all their required
209
# texts are present in the other repository - it could be
211
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
213
from_weave.clear_cache()
220
class GenericRepoFetcher(RepoFetcher):
221
"""This is a generic repo to repo fetcher.
223
This makes minimal assumptions about repo layout and contents.
224
It triggers a reconciliation after fetching to ensure integrity.
227
def _fetch_revision_texts(self, revs):
228
"""Fetch revision object texts"""
229
rev_pb = bzrlib.ui.ui_factory.nested_progress_bar()
231
to_txn = self.to_transaction = self.to_repository.get_transaction()
234
to_store = self.to_repository._revision_store
236
pb = bzrlib.ui.ui_factory.nested_progress_bar()
238
pb.update('copying revisions', count, total)
240
sig_text = self.from_repository.get_signature_text(rev)
241
to_store.add_revision_signature_text(rev, sig_text, to_txn)
242
except errors.NoSuchRevision:
245
to_store.add_revision(self.from_repository.get_revision(rev),
250
# fixup inventory if needed:
251
# this is expensive because we have no inverse index to current ghosts.
252
# but on local disk its a few seconds and sftp push is already insane.
254
# FIXME: repository should inform if this is needed.
255
self.to_repository.reconcile()
260
class KnitRepoFetcher(RepoFetcher):
261
"""This is a knit format repository specific fetcher.
263
This differs from the GenericRepoFetcher by not doing a
264
reconciliation after copying, and using knit joining to
268
def _fetch_revision_texts(self, revs):
269
# may need to be a InterRevisionStore call here.
270
from_transaction = self.from_repository.get_transaction()
271
to_transaction = self.to_repository.get_transaction()
272
to_sf = self.to_repository._revision_store.get_signature_file(
274
from_sf = self.from_repository._revision_store.get_signature_file(
276
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
277
to_rf = self.to_repository._revision_store.get_revision_file(
279
from_rf = self.from_repository._revision_store.get_revision_file(
281
to_rf.join(from_rf, version_ids=revs)
187
284
class Inter1and2Helper(object):
188
285
"""Helper for operations that convert data from model 1 and 2
190
287
This is for use by fetchers and converters.
193
def __init__(self, source):
290
def __init__(self, source, target):
196
293
:param source: The repository data comes from
294
:param target: The repository data goes to
198
296
self.source = source
200
299
def iter_rev_trees(self, revs):
201
300
"""Iterate through RevisionTrees efficiently.
217
314
revs = revs[100:]
219
def _find_root_ids(self, revs, parent_map, graph):
221
for tree in self.iter_rev_trees(revs):
222
revision_id = tree.inventory.root.revision
223
root_id = tree.get_root_id()
224
revision_root[revision_id] = root_id
225
# Find out which parents we don't already know root ids for
227
for revision_parents in parent_map.itervalues():
228
parents.update(revision_parents)
229
parents.difference_update(revision_root.keys() + [NULL_REVISION])
230
# Limit to revisions present in the versionedfile
231
parents = graph.get_parent_map(parents).keys()
232
for tree in self.iter_rev_trees(parents):
233
root_id = tree.get_root_id()
234
revision_root[tree.get_revision_id()] = root_id
237
316
def generate_root_texts(self, revs):
238
317
"""Generate VersionedFiles for all root ids.
240
319
:param revs: the revisions to include
242
graph = self.source.get_graph()
243
parent_map = graph.get_parent_map(revs)
244
rev_order = tsort.topo_sort(parent_map)
245
rev_id_to_root_id = self._find_root_ids(revs, parent_map, graph)
246
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
248
# Guaranteed stable, this groups all the file id operations together
249
# retaining topological order within the revisions of a file id.
250
# File id splits and joins would invalidate this, but they don't exist
251
# yet, and are unlikely to in non-rich-root environments anyway.
252
root_id_order.sort(key=operator.itemgetter(0))
253
# Create a record stream containing the roots to create.
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.
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]
321
inventory_weave = self.source.get_inventory_weave()
324
to_store = self.target.weave_store
325
for tree in self.iter_rev_trees(revs):
326
revision_id = tree.inventory.root.revision
327
root_id = tree.inventory.root.file_id
328
parents = inventory_weave.get_parents(revision_id)
329
if root_id not in versionedfile:
330
versionedfile[root_id] = to_store.get_weave_or_empty(root_id,
331
self.target.get_transaction())
332
parent_texts[root_id] = versionedfile[root_id].add_lines(
333
revision_id, parents, [], parent_texts)
335
def regenerate_inventory(self, revs):
336
"""Generate a new inventory versionedfile in target, convertin data.
338
The inventory is retrieved from the source, (deserializing it), and
339
stored in the target (reserializing it in a different format).
340
:param revs: The revisions to include
342
inventory_weave = self.source.get_inventory_weave()
343
for tree in self.iter_rev_trees(revs):
344
parents = inventory_weave.get_parents(tree.get_revision_id())
345
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
349
class Model1toKnit2Fetcher(GenericRepoFetcher):
350
"""Fetch from a Model1 repository into a Knit2 repository
352
def __init__(self, to_repository, from_repository, last_revision=None,
354
self.helper = Inter1and2Helper(from_repository, to_repository)
355
GenericRepoFetcher.__init__(self, to_repository, from_repository,
358
def _fetch_weave_texts(self, revs):
359
GenericRepoFetcher._fetch_weave_texts(self, revs)
360
# Now generate a weave for the tree root
361
self.helper.generate_root_texts(revs)
363
def _fetch_inventory_weave(self, revs):
364
self.helper.regenerate_inventory(revs)
367
class Knit1to2Fetcher(KnitRepoFetcher):
368
"""Fetch from a Knit1 repository into a Knit2 repository"""
370
def __init__(self, to_repository, from_repository, last_revision=None,
372
self.helper = Inter1and2Helper(from_repository, to_repository)
373
KnitRepoFetcher.__init__(self, to_repository, from_repository,
376
def _fetch_weave_texts(self, revs):
377
KnitRepoFetcher._fetch_weave_texts(self, revs)
378
# Now generate a weave for the tree root
379
self.helper.generate_root_texts(revs)
381
def _fetch_inventory_weave(self, revs):
382
self.helper.regenerate_inventory(revs)
385
class Fetcher(object):
386
"""Backwards compatibility glue for branch.fetch()."""
388
@deprecated_method(zero_eight)
389
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
390
"""Please see branch.fetch()."""
391
to_branch.fetch(from_branch, last_revision, pb)