118
131
This initialises all the needed variables, and then fetches the
119
132
requested revisions, finally clearing the progress bar.
121
self.to_weaves = self.to_repository.weave_store
122
self.to_control = self.to_repository.control_weaves
123
self.from_weaves = self.from_repository.weave_store
124
self.from_control = self.from_repository.control_weaves
125
134
self.count_total = 0
126
135
self.file_ids_names = {}
127
pp = ProgressPhase('fetch phase', 4, self.pb)
136
pp = ProgressPhase('Transferring', 4, self.pb)
129
revs = self._revids_to_fetch()
133
self._fetch_weave_texts(revs)
135
self._fetch_inventory_weave(revs)
137
self._fetch_revision_texts(revs)
138
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)
142
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.
143
218
mutter('fetch up to rev {%s}', self._last_revision)
144
219
if self._last_revision is NULL_REVISION:
145
220
# explicit limit of no revisions needed
147
if (self._last_revision != None and
222
if (self._last_revision is not None and
148
223
self.to_repository.has_revision(self._last_revision)):
152
return self.to_repository.missing_revision_ids(self.from_repository,
154
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:
155
230
raise InstallFailed([self._last_revision])
157
def _fetch_weave_texts(self, revs):
158
texts_pb = bzrlib.ui.ui_factory.nested_progress_bar()
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()
160
file_ids = self.from_repository.fileids_altered_by_revision_ids(revs)
162
num_file_ids = len(file_ids)
163
for file_id, required_versions in file_ids.items():
164
texts_pb.update("fetch texts", count, num_file_ids)
166
to_weave = self.to_weaves.get_weave_or_empty(file_id,
167
self.to_repository.get_transaction())
168
from_weave = self.from_weaves.get_weave(file_id,
169
self.from_repository.get_transaction())
170
# we fetch all the texts, because texts do
171
# not reference anything, and its cheap enough
172
to_weave.join(from_weave, version_ids=required_versions)
173
# we don't need *all* of this data anymore, but we dont know
174
# what we do. This cache clearing will result in a new read
175
# of the knit data when we do the checkout, but probably we
176
# want to emit the needed data on the fly rather than at the
178
# the from weave should know not to cache data being joined,
179
# but its ok to ask it to clear.
180
from_weave.clear_cache()
181
to_weave.clear_cache()
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))
185
def _fetch_inventory_weave(self, revs):
186
pb = bzrlib.ui.ui_factory.nested_progress_bar()
188
pb.update("fetch inventory", 0, 2)
189
to_weave = self.to_control.get_weave('inventory',
190
self.to_repository.get_transaction())
253
def _fetch_revision_texts(self, revs, pb):
254
# fetch signatures first and then the revision texts
255
# may need to be a InterRevisionStore call here.
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.
286
class Inter1and2Helper(object):
287
"""Helper for operations that convert data from model 1 and 2
192
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
194
# just merge, this is optimisable and its means we don't
195
# copy unreferenced data such as not-needed inventories.
196
pb.update("fetch inventory", 1, 3)
197
from_weave = self.from_repository.get_inventory_weave()
198
pb.update("fetch inventory", 2, 3)
199
# we fetch only the referenced inventories because we do not
200
# know for unselected inventories whether all their required
201
# texts are present in the other repository - it could be
203
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
211
class GenericRepoFetcher(RepoFetcher):
212
"""This is a generic repo to repo fetcher.
214
This makes minimal assumptions about repo layout and contents.
215
It triggers a reconciliation after fetching to ensure integrity.
218
def _fetch_revision_texts(self, revs):
289
This is for use by fetchers and converters.
292
def __init__(self, source, target):
295
:param source: The repository data comes from
296
:param target: The repository data goes to
301
def iter_rev_trees(self, revs):
302
"""Iterate through RevisionTrees efficiently.
304
Additionally, the inventory's revision_id is set if unset.
306
Trees are retrieved in batches of 100, and then yielded in the order
309
:param revs: A list of revision ids
311
# In case that revs is not a list.
314
for tree in self.source.revision_trees(revs[:100]):
315
if tree.inventory.revision_id is None:
316
tree.inventory.revision_id = tree.get_revision_id()
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
340
def generate_root_texts(self, revs):
341
"""Generate VersionedFiles for all root ids.
343
:param revs: the revisions to include
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())
377
def regenerate_inventory(self, revs):
378
"""Generate a new inventory versionedfile in target, convertin data.
380
The inventory is retrieved from the source, (deserializing it), and
381
stored in the target (reserializing it in a different format).
382
:param revs: The revisions to include
384
for tree in self.iter_rev_trees(revs):
385
parents = tree.get_parent_ids()
386
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
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):
397
"""Fetch from a Model1 repository into a Knit2 repository
399
def __init__(self, to_repository, from_repository, last_revision=None,
400
pb=None, find_ghosts=True):
401
self.helper = Inter1and2Helper(from_repository, to_repository)
402
RepoFetcher.__init__(self, to_repository, from_repository,
403
last_revision, pb, find_ghosts)
405
def _generate_root_texts(self, revs):
406
self.helper.generate_root_texts(revs)
408
def _fetch_inventory_weave(self, revs, pb):
409
self.helper.regenerate_inventory(revs)
411
def _fetch_revision_texts(self, revs, pb):
219
412
"""Fetch revision object texts"""
220
rev_pb = bzrlib.ui.ui_factory.nested_progress_bar()
222
to_txn = self.to_transaction = self.to_repository.get_transaction()
225
to_store = self.to_repository._revision_store
227
pb = bzrlib.ui.ui_factory.nested_progress_bar()
229
pb.update('copying revisions', count, total)
231
sig_text = self.from_repository.get_signature_text(rev)
232
to_store.add_revision_signature_text(rev, sig_text, to_txn)
233
except errors.NoSuchRevision:
236
to_store.add_revision(self.from_repository.get_revision(rev),
241
# fixup inventory if needed:
242
# this is expensive because we have no inverse index to current ghosts.
243
# but on local disk its a few seconds and sftp push is already insane.
245
# FIXME: repository should inform if this is needed.
246
self.to_repository.reconcile()
251
class KnitRepoFetcher(RepoFetcher):
252
"""This is a knit format repository specific fetcher.
254
This differs from the GenericRepoFetcher by not doing a
255
reconciliation after copying, and using knit joining to
259
def _fetch_revision_texts(self, revs):
260
# may need to be a InterRevisionStore call here.
261
from_transaction = self.from_repository.get_transaction()
262
to_transaction = self.to_repository.get_transaction()
263
to_sf = self.to_repository._revision_store.get_signature_file(
265
from_sf = self.from_repository._revision_store.get_signature_file(
267
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
268
to_rf = self.to_repository._revision_store.get_revision_file(
270
from_rf = self.from_repository._revision_store.get_revision_file(
272
to_rf.join(from_rf, version_ids=revs)
275
class Fetcher(object):
276
"""Backwards compatibility glue for branch.fetch()."""
278
@deprecated_method(zero_eight)
279
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
280
"""Please see branch.fetch()."""
281
to_branch.fetch(from_branch, last_revision, pb)
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):
431
"""Fetch from a Knit1 repository into a Knit2 repository"""
433
def __init__(self, to_repository, from_repository, last_revision=None,
434
pb=None, find_ghosts=True):
435
self.helper = Inter1and2Helper(from_repository, to_repository)
436
RepoFetcher.__init__(self, to_repository, from_repository,
437
last_revision, pb, find_ghosts)
439
def _generate_root_texts(self, revs):
440
self.helper.generate_root_texts(revs)
442
def _fetch_inventory_weave(self, revs, pb):
443
self.helper.regenerate_inventory(revs)
445
def _fetch_just_revision_texts(self, version_ids):
446
self.helper.fetch_revisions(version_ids)