120
128
This initialises all the needed variables, and then fetches the
121
129
requested revisions, finally clearing the progress bar.
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
127
131
self.count_total = 0
128
132
self.file_ids_names = {}
129
pp = ProgressPhase('Fetch phase', 4, self.pb)
133
pp = ProgressPhase('Transferring', 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)
136
search = self._revids_to_fetch()
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)
147
def _fetch_everything_for_search(self, search, pp):
148
"""Fetch all data for the given set of revisions."""
149
# The first phase is "file". We pass the progress bar for it directly
150
# into item_keys_introduced_by, which has more information about how
151
# that phase is progressing than we do. Progress updates for the other
152
# phases are taken care of in this function.
153
# XXX: there should be a clear owner of the progress reporting. Perhaps
154
# item_keys_introduced_by should have a richer API than it does at the
155
# moment, so that it can feed the progress information back to this
158
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
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()
207
self.count_copied += len(revs)
145
209
def _revids_to_fetch(self):
210
"""Determines the exact revisions needed from self.from_repository to
211
install self._last_revision in self.to_repository.
213
If no revisions need to be fetched, then this just returns None.
146
215
mutter('fetch up to rev {%s}', self._last_revision)
147
216
if self._last_revision is NULL_REVISION:
148
217
# explicit limit of no revisions needed
150
if (self._last_revision != None and
219
if (self._last_revision is not None and
151
220
self.to_repository.has_revision(self._last_revision)):
155
return self.to_repository.missing_revision_ids(self.from_repository,
157
except errors.NoSuchRevision:
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:
158
227
raise InstallFailed([self._last_revision])
160
def _fetch_weave_texts(self, revs):
161
texts_pb = bzrlib.ui.ui_factory.nested_progress_bar()
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()
163
file_ids = self.from_repository.fileids_altered_by_revision_ids(revs)
165
num_file_ids = len(file_ids)
166
for file_id, required_versions in file_ids.items():
167
texts_pb.update("fetch texts", count, num_file_ids)
169
to_weave = self.to_weaves.get_weave_or_empty(file_id,
170
self.to_repository.get_transaction())
171
from_weave = self.from_weaves.get_weave(file_id,
172
self.from_repository.get_transaction())
173
# we fetch all the texts, because texts do
174
# not reference anything, and its cheap enough
175
to_weave.join(from_weave, version_ids=required_versions)
176
# we don't need *all* of this data anymore, but we dont know
177
# what we do. This cache clearing will result in a new read
178
# of the knit data when we do the checkout, but probably we
179
# want to emit the needed data on the fly rather than at the
181
# the from weave should know not to cache data being joined,
182
# but its ok to ask it to clear.
183
from_weave.clear_cache()
184
to_weave.clear_cache()
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
self.to_repository._fetch_uses_deltas))
188
def _fetch_inventory_weave(self, revs):
189
pb = bzrlib.ui.ui_factory.nested_progress_bar()
191
pb.update("fetch inventory", 0, 2)
192
to_weave = self.to_control.get_weave('inventory',
193
self.to_repository.get_transaction())
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,
258
self.to_repository._fetch_uses_deltas)))
259
self._fetch_just_revision_texts(revs)
261
def _fetch_just_revision_texts(self, version_ids):
262
to_rf = self.to_repository.revisions
263
from_rf = self.from_repository.revisions
264
to_rf.insert_record_stream(from_rf.get_record_stream(
265
[(rev_id,) for rev_id in version_ids],
266
self.to_repository._fetch_order,
267
self.to_repository._fetch_uses_deltas))
269
def _generate_root_texts(self, revs):
270
"""This will be called by __fetch between fetching weave texts and
271
fetching the inventory weave.
273
Subclasses should override this if they need to generate root texts
274
after fetching weave texts.
279
class Inter1and2Helper(object):
280
"""Helper for operations that convert data from model 1 and 2
195
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
197
# just merge, this is optimisable and its means we don't
198
# copy unreferenced data such as not-needed inventories.
199
pb.update("fetch inventory", 1, 3)
200
from_weave = self.from_repository.get_inventory_weave()
201
pb.update("fetch inventory", 2, 3)
202
# we fetch only the referenced inventories because we do not
203
# know for unselected inventories whether all their required
204
# texts are present in the other repository - it could be
206
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
214
class GenericRepoFetcher(RepoFetcher):
215
"""This is a generic repo to repo fetcher.
217
This makes minimal assumptions about repo layout and contents.
218
It triggers a reconciliation after fetching to ensure integrity.
221
def _fetch_revision_texts(self, revs):
282
This is for use by fetchers and converters.
285
def __init__(self, source, target):
288
:param source: The repository data comes from
289
:param target: The repository data goes to
294
def iter_rev_trees(self, revs):
295
"""Iterate through RevisionTrees efficiently.
297
Additionally, the inventory's revision_id is set if unset.
299
Trees are retrieved in batches of 100, and then yielded in the order
302
:param revs: A list of revision ids
304
# In case that revs is not a list.
307
for tree in self.source.revision_trees(revs[:100]):
308
if tree.inventory.revision_id is None:
309
tree.inventory.revision_id = tree.get_revision_id()
313
def _find_root_ids(self, revs, parent_map, graph):
315
planned_versions = {}
316
for tree in self.iter_rev_trees(revs):
317
revision_id = tree.inventory.root.revision
318
root_id = tree.get_root_id()
319
planned_versions.setdefault(root_id, []).append(revision_id)
320
revision_root[revision_id] = root_id
321
# Find out which parents we don't already know root ids for
323
for revision_parents in parent_map.itervalues():
324
parents.update(revision_parents)
325
parents.difference_update(revision_root.keys() + [NULL_REVISION])
326
# Limit to revisions present in the versionedfile
327
parents = graph.get_parent_map(parents).keys()
328
for tree in self.iter_rev_trees(parents):
329
root_id = tree.get_root_id()
330
revision_root[tree.get_revision_id()] = root_id
331
return revision_root, planned_versions
333
def generate_root_texts(self, revs):
334
"""Generate VersionedFiles for all root ids.
336
:param revs: the revisions to include
338
to_texts = self.target.texts
339
graph = self.source.get_graph()
340
parent_map = graph.get_parent_map(revs)
341
rev_order = topo_sort(parent_map)
342
rev_id_to_root_id, root_id_to_rev_ids = self._find_root_ids(
343
revs, parent_map, graph)
344
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
346
# Guaranteed stable, this groups all the file id operations together
347
# retaining topological order within the revisions of a file id.
348
# File id splits and joins would invalidate this, but they don't exist
349
# yet, and are unlikely to in non-rich-root environments anyway.
350
root_id_order.sort(key=operator.itemgetter(0))
351
# Create a record stream containing the roots to create.
353
for key in root_id_order:
354
root_id, rev_id = key
355
rev_parents = parent_map[rev_id]
356
# We drop revision parents with different file-ids, because
357
# that represents a rename of the root to a different location
358
# - its not actually a parent for us. (We could look for that
359
# file id in the revision tree at considerably more expense,
360
# but for now this is sufficient (and reconcile will catch and
361
# correct this anyway).
362
# When a parent revision is a ghost, we guess that its root id
363
# was unchanged (rather than trimming it from the parent list).
364
parent_keys = tuple((root_id, parent) for parent in rev_parents
365
if parent != NULL_REVISION and
366
rev_id_to_root_id.get(parent, root_id) == root_id)
367
yield FulltextContentFactory(key, parent_keys, None, '')
368
to_texts.insert_record_stream(yield_roots())
370
def regenerate_inventory(self, revs):
371
"""Generate a new inventory versionedfile in target, convertin data.
373
The inventory is retrieved from the source, (deserializing it), and
374
stored in the target (reserializing it in a different format).
375
:param revs: The revisions to include
377
for tree in self.iter_rev_trees(revs):
378
parents = tree.get_parent_ids()
379
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
382
def fetch_revisions(self, revision_ids):
383
for revision in self.source.get_revisions(revision_ids):
384
self.target.add_revision(revision.revision_id, revision)
387
class Model1toKnit2Fetcher(RepoFetcher):
388
"""Fetch from a Model1 repository into a Knit2 repository
390
def __init__(self, to_repository, from_repository, last_revision=None,
391
pb=None, find_ghosts=True):
392
self.helper = Inter1and2Helper(from_repository, to_repository)
393
RepoFetcher.__init__(self, to_repository, from_repository,
394
last_revision, pb, find_ghosts)
396
def _generate_root_texts(self, revs):
397
self.helper.generate_root_texts(revs)
399
def _fetch_inventory_weave(self, revs, pb):
400
self.helper.regenerate_inventory(revs)
402
def _fetch_revision_texts(self, revs, pb):
222
403
"""Fetch revision object texts"""
223
rev_pb = bzrlib.ui.ui_factory.nested_progress_bar()
225
to_txn = self.to_transaction = self.to_repository.get_transaction()
228
to_store = self.to_repository._revision_store
230
pb = bzrlib.ui.ui_factory.nested_progress_bar()
232
pb.update('copying revisions', count, total)
234
sig_text = self.from_repository.get_signature_text(rev)
235
to_store.add_revision_signature_text(rev, sig_text, to_txn)
236
except errors.NoSuchRevision:
239
to_store.add_revision(self.from_repository.get_revision(rev),
244
# fixup inventory if needed:
245
# this is expensive because we have no inverse index to current ghosts.
246
# but on local disk its a few seconds and sftp push is already insane.
248
# FIXME: repository should inform if this is needed.
249
self.to_repository.reconcile()
254
class KnitRepoFetcher(RepoFetcher):
255
"""This is a knit format repository specific fetcher.
257
This differs from the GenericRepoFetcher by not doing a
258
reconciliation after copying, and using knit joining to
262
def _fetch_revision_texts(self, revs):
263
# may need to be a InterRevisionStore call here.
264
from_transaction = self.from_repository.get_transaction()
265
to_transaction = self.to_repository.get_transaction()
266
to_sf = self.to_repository._revision_store.get_signature_file(
268
from_sf = self.from_repository._revision_store.get_signature_file(
270
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
271
to_rf = self.to_repository._revision_store.get_revision_file(
273
from_rf = self.from_repository._revision_store.get_revision_file(
275
to_rf.join(from_rf, version_ids=revs)
278
class Fetcher(object):
279
"""Backwards compatibility glue for branch.fetch()."""
281
@deprecated_method(zero_eight)
282
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
283
"""Please see branch.fetch()."""
284
to_branch.fetch(from_branch, last_revision, pb)
407
pb.update('copying revisions', count, total)
409
sig_text = self.from_repository.get_signature_text(rev)
410
self.to_repository.add_signature_text(rev, sig_text)
411
except errors.NoSuchRevision:
414
self._copy_revision(rev)
417
def _copy_revision(self, rev):
418
self.helper.fetch_revisions([rev])
421
class Knit1to2Fetcher(RepoFetcher):
422
"""Fetch from a Knit1 repository into a Knit2 repository"""
424
def __init__(self, to_repository, from_repository, last_revision=None,
425
pb=None, find_ghosts=True):
426
self.helper = Inter1and2Helper(from_repository, to_repository)
427
RepoFetcher.__init__(self, to_repository, from_repository,
428
last_revision, pb, find_ghosts)
430
def _generate_root_texts(self, revs):
431
self.helper.generate_root_texts(revs)
433
def _fetch_inventory_weave(self, revs, pb):
434
self.helper.regenerate_inventory(revs)
436
def _fetch_just_revision_texts(self, version_ids):
437
self.helper.fetch_revisions(version_ids)