82
73
count_copied -- number of revisions copied
84
This should not be used directory, its essential a object to encapsulate
75
This should not be used directly, it's essential a object to encapsulate
85
76
the logic in InterRepository.fetch().
87
def __init__(self, to_repository, from_repository, last_revision=None, pb=None):
79
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
81
"""Create a repo fetcher.
83
:param find_ghosts: If True search the entire history for ghosts.
84
:param _write_group_acquired_callable: Don't use; this parameter only
85
exists to facilitate a hack done in InterPackRepo.fetch. We would
86
like to remove this parameter.
88
88
# result variables.
89
89
self.failed_revisions = []
90
90
self.count_copied = 0
91
if to_repository.control_files._transport.base == from_repository.control_files._transport.base:
92
# check that last_revision is in 'from' and then return a no-operation.
93
if last_revision not in (None, NULL_REVISION):
94
from_repository.get_revision(last_revision)
91
if to_repository.has_same_location(from_repository):
92
# repository.fetch should be taking care of this case.
93
raise errors.BzrError('RepoFetcher run '
94
'between two objects at the same location: '
95
'%r and %r' % (to_repository, from_repository))
96
96
self.to_repository = to_repository
97
97
self.from_repository = from_repository
98
98
# must not mutate self._last_revision as its potentially a shared instance
99
99
self._last_revision = last_revision
100
self.find_ghosts = find_ghosts
101
102
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
102
103
self.nested_pb = self.pb
121
131
This initialises all the needed variables, and then fetches the
122
132
requested revisions, finally clearing the progress bar.
124
self.to_weaves = self.to_repository.weave_store
125
self.to_control = self.to_repository.control_weaves
126
self.from_weaves = self.from_repository.weave_store
127
self.from_control = self.from_repository.control_weaves
128
134
self.count_total = 0
129
135
self.file_ids_names = {}
130
pp = ProgressPhase('Fetch phase', 4, self.pb)
136
pp = ProgressPhase('Transferring', 4, self.pb)
133
revs = self._revids_to_fetch()
137
self._fetch_weave_texts(revs)
139
self._fetch_inventory_weave(revs)
141
self._fetch_revision_texts(revs)
142
self.count_copied += len(revs)
139
search = self._revids_to_fetch()
142
if getattr(self, '_fetch_everything_for_search', None) is not None:
143
self._fetch_everything_for_search(search, pp)
145
# backward compatibility
146
self._fetch_everything_for_revisions(search.get_keys, pp)
150
def _fetch_everything_for_search(self, search, pp):
151
"""Fetch all data for the given set of revisions."""
152
# The first phase is "file". We pass the progress bar for it directly
153
# into item_keys_introduced_by, which has more information about how
154
# that phase is progressing than we do. Progress updates for the other
155
# phases are taken care of in this function.
156
# XXX: there should be a clear owner of the progress reporting. Perhaps
157
# item_keys_introduced_by should have a richer API than it does at the
158
# moment, so that it can feed the progress information back to this
161
pb = bzrlib.ui.ui_factory.nested_progress_bar()
163
revs = search.get_keys()
164
graph = self.from_repository.get_graph()
165
revs = list(graph.iter_topo_order(revs))
166
data_to_fetch = self.from_repository.item_keys_introduced_by(revs,
169
for knit_kind, file_id, revisions in data_to_fetch:
170
if knit_kind != phase:
172
# Make a new progress bar for this phase
175
pb = bzrlib.ui.ui_factory.nested_progress_bar()
176
if knit_kind == "file":
177
# Accumulate file texts
178
text_keys.extend([(file_id, revision) for revision in
180
elif knit_kind == "inventory":
181
# Now copy the file texts.
182
to_texts = self.to_repository.texts
183
from_texts = self.from_repository.texts
184
to_texts.insert_record_stream(from_texts.get_record_stream(
185
text_keys, self.to_repository._fetch_order,
186
not self.to_repository._fetch_uses_deltas))
187
# Cause an error if a text occurs after we have done the
190
# Before we process the inventory we generate the root
191
# texts (if necessary) so that the inventories references
193
self._generate_root_texts(revs)
194
# NB: This currently reopens the inventory weave in source;
195
# using a single stream interface instead would avoid this.
196
self._fetch_inventory_weave(revs, pb)
197
elif knit_kind == "signatures":
198
# Nothing to do here; this will be taken care of when
199
# _fetch_revision_texts happens.
201
elif knit_kind == "revisions":
202
self._fetch_revision_texts(revs, pb)
204
raise AssertionError("Unknown knit kind %r" % knit_kind)
205
if self.to_repository._fetch_reconcile:
206
self.to_repository.reconcile()
210
self.count_copied += len(revs)
146
212
def _revids_to_fetch(self):
213
"""Determines the exact revisions needed from self.from_repository to
214
install self._last_revision in self.to_repository.
216
If no revisions need to be fetched, then this just returns None.
147
218
mutter('fetch up to rev {%s}', self._last_revision)
148
219
if self._last_revision is NULL_REVISION:
149
220
# explicit limit of no revisions needed
151
222
if (self._last_revision is not None and
152
223
self.to_repository.has_revision(self._last_revision)):
156
return self.to_repository.missing_revision_ids(self.from_repository,
158
except errors.NoSuchRevision:
226
return self.to_repository.search_missing_revision_ids(
227
self.from_repository, self._last_revision,
228
find_ghosts=self.find_ghosts)
229
except errors.NoSuchRevision, e:
159
230
raise InstallFailed([self._last_revision])
161
def _fetch_weave_texts(self, revs):
162
texts_pb = bzrlib.ui.ui_factory.nested_progress_bar()
164
# fileids_altered_by_revision_ids requires reading the inventory
165
# weave, we will need to read the inventory weave again when
166
# all this is done, so enable caching for that specific weave
167
inv_w = self.from_repository.get_inventory_weave()
169
file_ids = self.from_repository.fileids_altered_by_revision_ids(revs)
171
num_file_ids = len(file_ids)
172
for file_id, required_versions in file_ids.items():
173
texts_pb.update("fetch texts", count, num_file_ids)
175
to_weave = self.to_weaves.get_weave_or_empty(file_id,
176
self.to_repository.get_transaction())
177
from_weave = self.from_weaves.get_weave(file_id,
178
self.from_repository.get_transaction())
179
# we fetch all the texts, because texts do
180
# not reference anything, and its cheap enough
181
to_weave.join(from_weave, version_ids=required_versions)
182
# we don't need *all* of this data anymore, but we dont know
183
# what we do. This cache clearing will result in a new read
184
# of the knit data when we do the checkout, but probably we
185
# want to emit the needed data on the fly rather than at the
187
# the from weave should know not to cache data being joined,
188
# but its ok to ask it to clear.
189
from_weave.clear_cache()
190
to_weave.clear_cache()
194
def _fetch_inventory_weave(self, revs):
195
pb = bzrlib.ui.ui_factory.nested_progress_bar()
197
pb.update("fetch inventory", 0, 2)
198
to_weave = self.to_control.get_weave('inventory',
199
self.to_repository.get_transaction())
201
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
203
# just merge, this is optimisable and its means we don't
204
# copy unreferenced data such as not-needed inventories.
205
pb.update("fetch inventory", 1, 3)
206
from_weave = self.from_repository.get_inventory_weave()
207
pb.update("fetch inventory", 2, 3)
208
# we fetch only the referenced inventories because we do not
209
# know for unselected inventories whether all their required
210
# texts are present in the other repository - it could be
212
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
214
from_weave.clear_cache()
221
class GenericRepoFetcher(RepoFetcher):
222
"""This is a generic repo to repo fetcher.
224
This makes minimal assumptions about repo layout and contents.
225
It triggers a reconciliation after fetching to ensure integrity.
228
def _fetch_revision_texts(self, revs):
229
"""Fetch revision object texts"""
230
rev_pb = bzrlib.ui.ui_factory.nested_progress_bar()
232
to_txn = self.to_transaction = self.to_repository.get_transaction()
235
to_store = self.to_repository._revision_store
237
pb = bzrlib.ui.ui_factory.nested_progress_bar()
239
pb.update('copying revisions', count, total)
241
sig_text = self.from_repository.get_signature_text(rev)
242
to_store.add_revision_signature_text(rev, sig_text, to_txn)
243
except errors.NoSuchRevision:
246
to_store.add_revision(self.from_repository.get_revision(rev),
251
# fixup inventory if needed:
252
# this is expensive because we have no inverse index to current ghosts.
253
# but on local disk its a few seconds and sftp push is already insane.
255
# FIXME: repository should inform if this is needed.
256
self.to_repository.reconcile()
261
class KnitRepoFetcher(RepoFetcher):
262
"""This is a knit format repository specific fetcher.
264
This differs from the GenericRepoFetcher by not doing a
265
reconciliation after copying, and using knit joining to
269
def _fetch_revision_texts(self, revs):
232
def _fetch_inventory_weave(self, revs, pb):
233
pb.update("fetch inventory", 0, 2)
234
to_weave = self.to_repository.inventories
235
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
237
# just merge, this is optimisable and its means we don't
238
# copy unreferenced data such as not-needed inventories.
239
pb.update("fetch inventory", 1, 3)
240
from_weave = self.from_repository.inventories
241
pb.update("fetch inventory", 2, 3)
242
# we fetch only the referenced inventories because we do not
243
# know for unselected inventories whether all their required
244
# texts are present in the other repository - it could be
246
to_weave.insert_record_stream(from_weave.get_record_stream(
247
[(rev_id,) for rev_id in revs],
248
self.to_repository._fetch_order,
249
not self.to_repository._fetch_uses_deltas))
253
def _fetch_revision_texts(self, revs, pb):
254
# fetch signatures first and then the revision texts
270
255
# may need to be a InterRevisionStore call here.
271
from_transaction = self.from_repository.get_transaction()
272
to_transaction = self.to_repository.get_transaction()
273
to_sf = self.to_repository._revision_store.get_signature_file(
275
from_sf = self.from_repository._revision_store.get_signature_file(
277
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
278
to_rf = self.to_repository._revision_store.get_revision_file(
280
from_rf = self.from_repository._revision_store.get_revision_file(
282
to_rf.join(from_rf, version_ids=revs)
256
to_sf = self.to_repository.signatures
257
from_sf = self.from_repository.signatures
258
# A missing signature is just skipped.
259
to_sf.insert_record_stream(filter_absent(from_sf.get_record_stream(
260
[(rev_id,) for rev_id in revs],
261
self.to_repository._fetch_order,
262
not self.to_repository._fetch_uses_deltas)))
263
self._fetch_just_revision_texts(revs)
265
def _fetch_just_revision_texts(self, version_ids):
266
to_rf = self.to_repository.revisions
267
from_rf = self.from_repository.revisions
268
# If a revision has a delta, this is actually expanded inside the
269
# insert_record_stream code now, which is an alternate fix for
271
to_rf.insert_record_stream(from_rf.get_record_stream(
272
[(rev_id,) for rev_id in version_ids],
273
self.to_repository._fetch_order,
274
not self.to_repository._fetch_uses_deltas))
276
def _generate_root_texts(self, revs):
277
"""This will be called by __fetch between fetching weave texts and
278
fetching the inventory weave.
280
Subclasses should override this if they need to generate root texts
281
after fetching weave texts.
285
286
class Inter1and2Helper(object):
315
318
revs = revs[100:]
320
def _find_root_ids(self, revs, parent_map, graph):
322
planned_versions = {}
323
for tree in self.iter_rev_trees(revs):
324
revision_id = tree.inventory.root.revision
325
root_id = tree.get_root_id()
326
planned_versions.setdefault(root_id, []).append(revision_id)
327
revision_root[revision_id] = root_id
328
# Find out which parents we don't already know root ids for
330
for revision_parents in parent_map.itervalues():
331
parents.update(revision_parents)
332
parents.difference_update(revision_root.keys() + [NULL_REVISION])
333
# Limit to revisions present in the versionedfile
334
parents = graph.get_parent_map(parents).keys()
335
for tree in self.iter_rev_trees(parents):
336
root_id = tree.get_root_id()
337
revision_root[tree.get_revision_id()] = root_id
338
return revision_root, planned_versions
317
340
def generate_root_texts(self, revs):
318
341
"""Generate VersionedFiles for all root ids.
320
343
:param revs: the revisions to include
322
inventory_weave = self.source.get_inventory_weave()
325
to_store = self.target.weave_store
326
for tree in self.iter_rev_trees(revs):
327
revision_id = tree.inventory.root.revision
328
root_id = tree.inventory.root.file_id
329
parents = inventory_weave.get_parents(revision_id)
330
if root_id not in versionedfile:
331
versionedfile[root_id] = to_store.get_weave_or_empty(root_id,
332
self.target.get_transaction())
333
parent_texts[root_id] = versionedfile[root_id].add_lines(
334
revision_id, parents, [], parent_texts)
345
to_texts = self.target.texts
346
graph = self.source.get_graph()
347
parent_map = graph.get_parent_map(revs)
348
rev_order = topo_sort(parent_map)
349
rev_id_to_root_id, root_id_to_rev_ids = self._find_root_ids(
350
revs, parent_map, graph)
351
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
353
# Guaranteed stable, this groups all the file id operations together
354
# retaining topological order within the revisions of a file id.
355
# File id splits and joins would invalidate this, but they don't exist
356
# yet, and are unlikely to in non-rich-root environments anyway.
357
root_id_order.sort(key=operator.itemgetter(0))
358
# Create a record stream containing the roots to create.
360
for key in root_id_order:
361
root_id, rev_id = key
362
rev_parents = parent_map[rev_id]
363
# We drop revision parents with different file-ids, because
364
# that represents a rename of the root to a different location
365
# - its not actually a parent for us. (We could look for that
366
# file id in the revision tree at considerably more expense,
367
# but for now this is sufficient (and reconcile will catch and
368
# correct this anyway).
369
# When a parent revision is a ghost, we guess that its root id
370
# was unchanged (rather than trimming it from the parent list).
371
parent_keys = tuple((root_id, parent) for parent in rev_parents
372
if parent != NULL_REVISION and
373
rev_id_to_root_id.get(parent, root_id) == root_id)
374
yield FulltextContentFactory(key, parent_keys, None, '')
375
to_texts.insert_record_stream(yield_roots())
336
377
def regenerate_inventory(self, revs):
337
378
"""Generate a new inventory versionedfile in target, convertin data.
340
381
stored in the target (reserializing it in a different format).
341
382
:param revs: The revisions to include
343
inventory_weave = self.source.get_inventory_weave()
344
384
for tree in self.iter_rev_trees(revs):
345
parents = inventory_weave.get_parents(tree.get_revision_id())
385
parents = tree.get_parent_ids()
346
386
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
350
class Model1toKnit2Fetcher(GenericRepoFetcher):
389
def fetch_revisions(self, revision_ids):
390
# TODO: should this batch them up rather than requesting 10,000
392
for revision in self.source.get_revisions(revision_ids):
393
self.target.add_revision(revision.revision_id, revision)
396
class Model1toKnit2Fetcher(RepoFetcher):
351
397
"""Fetch from a Model1 repository into a Knit2 repository
353
def __init__(self, to_repository, from_repository, last_revision=None,
399
def __init__(self, to_repository, from_repository, last_revision=None,
400
pb=None, find_ghosts=True):
355
401
self.helper = Inter1and2Helper(from_repository, to_repository)
356
GenericRepoFetcher.__init__(self, to_repository, from_repository,
402
RepoFetcher.__init__(self, to_repository, from_repository,
403
last_revision, pb, find_ghosts)
359
def _fetch_weave_texts(self, revs):
360
GenericRepoFetcher._fetch_weave_texts(self, revs)
361
# Now generate a weave for the tree root
405
def _generate_root_texts(self, revs):
362
406
self.helper.generate_root_texts(revs)
364
def _fetch_inventory_weave(self, revs):
408
def _fetch_inventory_weave(self, revs, pb):
365
409
self.helper.regenerate_inventory(revs)
368
class Knit1to2Fetcher(KnitRepoFetcher):
411
def _fetch_revision_texts(self, revs, pb):
412
"""Fetch revision object texts"""
416
pb.update('copying revisions', count, total)
418
sig_text = self.from_repository.get_signature_text(rev)
419
self.to_repository.add_signature_text(rev, sig_text)
420
except errors.NoSuchRevision:
423
self._copy_revision(rev)
426
def _copy_revision(self, rev):
427
self.helper.fetch_revisions([rev])
430
class Knit1to2Fetcher(RepoFetcher):
369
431
"""Fetch from a Knit1 repository into a Knit2 repository"""
371
def __init__(self, to_repository, from_repository, last_revision=None,
433
def __init__(self, to_repository, from_repository, last_revision=None,
434
pb=None, find_ghosts=True):
373
435
self.helper = Inter1and2Helper(from_repository, to_repository)
374
KnitRepoFetcher.__init__(self, to_repository, from_repository,
436
RepoFetcher.__init__(self, to_repository, from_repository,
437
last_revision, pb, find_ghosts)
377
def _fetch_weave_texts(self, revs):
378
KnitRepoFetcher._fetch_weave_texts(self, revs)
379
# Now generate a weave for the tree root
439
def _generate_root_texts(self, revs):
380
440
self.helper.generate_root_texts(revs)
382
def _fetch_inventory_weave(self, revs):
442
def _fetch_inventory_weave(self, revs, pb):
383
443
self.helper.regenerate_inventory(revs)
386
class Fetcher(object):
387
"""Backwards compatibility glue for branch.fetch()."""
389
@deprecated_method(zero_eight)
390
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
391
"""Please see branch.fetch()."""
392
to_branch.fetch(from_branch, last_revision, pb)
445
def _fetch_just_revision_texts(self, version_ids):
446
self.helper.fetch_revisions(version_ids)