27
27
stored, so that if a revision is present we can totally recreate it.
28
28
However, we can't know what files are included in a revision until we
29
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 then return to the inventories.
30
the ids we need, and then pull those ids and finally actually join
36
35
import bzrlib.errors as errors
37
from bzrlib.errors import InstallFailed
36
from bzrlib.errors import (InstallFailed,
38
38
from bzrlib.progress import ProgressPhase
39
39
from bzrlib.revision import is_null, NULL_REVISION
40
40
from bzrlib.symbol_versioning import (deprecated_function,
43
from bzrlib.tsort import topo_sort
44
43
from bzrlib.trace import mutter
46
from bzrlib.versionedfile import filter_absent, FulltextContentFactory
46
from bzrlib.lazy_import import lazy_import
48
48
# TODO: Avoid repeatedly opening weaves so many times.
128
119
This initialises all the needed variables, and then fetches the
129
120
requested revisions, finally clearing the progress bar.
122
self.to_weaves = self.to_repository.weave_store
123
self.from_weaves = self.from_repository.weave_store
131
124
self.count_total = 0
132
125
self.file_ids_names = {}
133
pp = ProgressPhase('Transferring', 4, self.pb)
126
pp = ProgressPhase('Fetch phase', 4, self.pb)
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)
129
revs = self._revids_to_fetch()
130
self._fetch_everything_for_revisions(revs, pp)
147
def _fetch_everything_for_search(self, search, pp):
134
def _fetch_everything_for_revisions(self, revs, pp):
148
135
"""Fetch all data for the given set of revisions."""
149
138
# The first phase is "file". We pass the progress bar for it directly
150
139
# into item_keys_introduced_by, which has more information about how
151
140
# that phase is progressing than we do. Progress updates for the other
172
156
pb = bzrlib.ui.ui_factory.nested_progress_bar()
173
157
if knit_kind == "file":
174
# Accumulate file texts
175
text_keys.extend([(file_id, revision) for revision in
158
self._fetch_weave_text(file_id, revisions)
177
159
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
not 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
161
# Once we've processed all the files, then we generate the root
162
# texts (if necessary), then we process the inventory. It's a
163
# bit distasteful to have knit_kind == "inventory" mean this,
164
# perhaps it should happen on the first non-"file" knit, in case
165
# it's not always inventory?
190
166
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
167
self._fetch_inventory_weave(revs, pb)
194
168
elif knit_kind == "signatures":
195
169
# Nothing to do here; this will be taken care of when
219
191
if (self._last_revision is not None and
220
192
self.to_repository.has_revision(self._last_revision)):
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:
196
return self.to_repository.missing_revision_ids(self.from_repository,
198
except errors.NoSuchRevision:
227
199
raise InstallFailed([self._last_revision])
201
def _fetch_weave_text(self, file_id, required_versions):
202
to_weave = self.to_weaves.get_weave_or_empty(file_id,
203
self.to_repository.get_transaction())
204
from_weave = self.from_weaves.get_weave(file_id,
205
self.from_repository.get_transaction())
206
# we fetch all the texts, because texts do
207
# not reference anything, and its cheap enough
208
to_weave.join(from_weave, version_ids=required_versions)
209
# we don't need *all* of this data anymore, but we dont know
210
# what we do. This cache clearing will result in a new read
211
# of the knit data when we do the checkout, but probably we
212
# want to emit the needed data on the fly rather than at the
214
# the from weave should know not to cache data being joined,
215
# but its ok to ask it to clear.
216
from_weave.clear_cache()
217
to_weave.clear_cache()
229
219
def _fetch_inventory_weave(self, revs, pb):
230
220
pb.update("fetch inventory", 0, 2)
231
to_weave = self.to_repository.inventories
221
to_weave = self.to_repository.get_inventory_weave()
232
222
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
234
224
# just merge, this is optimisable and its means we don't
235
225
# copy unreferenced data such as not-needed inventories.
236
226
pb.update("fetch inventory", 1, 3)
237
from_weave = self.from_repository.inventories
227
from_weave = self.from_repository.get_inventory_weave()
238
228
pb.update("fetch inventory", 2, 3)
239
229
# we fetch only the referenced inventories because we do not
240
230
# know for unselected inventories whether all their required
241
231
# 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
not self.to_repository._fetch_uses_deltas))
233
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
235
from_weave.clear_cache()
248
237
child_pb.finished()
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,
259
# Bug #261339, some knit repositories accidentally had deltas in their
260
# revision stream, when you weren't ever supposed to have deltas.
261
# So we now *force* fulltext copying for signatures and revisions
262
self._fetch_just_revision_texts(revs)
264
def _fetch_just_revision_texts(self, version_ids):
265
to_rf = self.to_repository.revisions
266
from_rf = self.from_repository.revisions
267
to_rf.insert_record_stream(from_rf.get_record_stream(
268
[(rev_id,) for rev_id in version_ids],
269
self.to_repository._fetch_order,
271
# Bug #261339, some knit repositories accidentally had deltas in their
272
# revision stream, when you weren't ever supposed to have deltas.
273
# So we now *force* fulltext copying for signatures and revisions
275
239
def _generate_root_texts(self, revs):
276
240
"""This will be called by __fetch between fetching weave texts and
277
241
fetching the inventory weave.
249
class GenericRepoFetcher(RepoFetcher):
250
"""This is a generic repo to repo fetcher.
252
This makes minimal assumptions about repo layout and contents.
253
It triggers a reconciliation after fetching to ensure integrity.
256
def _fetch_revision_texts(self, revs, pb):
257
"""Fetch revision object texts"""
258
to_txn = self.to_transaction = self.to_repository.get_transaction()
261
to_store = self.to_repository._revision_store
263
pb.update('copying revisions', count, total)
265
sig_text = self.from_repository.get_signature_text(rev)
266
to_store.add_revision_signature_text(rev, sig_text, to_txn)
267
except errors.NoSuchRevision:
270
to_store.add_revision(self.from_repository.get_revision(rev),
273
# fixup inventory if needed:
274
# this is expensive because we have no inverse index to current ghosts.
275
# but on local disk its a few seconds and sftp push is already insane.
277
# FIXME: repository should inform if this is needed.
278
self.to_repository.reconcile()
281
class KnitRepoFetcher(RepoFetcher):
282
"""This is a knit format repository specific fetcher.
284
This differs from the GenericRepoFetcher by not doing a
285
reconciliation after copying, and using knit joining to
289
def _fetch_revision_texts(self, revs, pb):
290
# may need to be a InterRevisionStore call here.
291
from_transaction = self.from_repository.get_transaction()
292
to_transaction = self.to_repository.get_transaction()
293
to_sf = self.to_repository._revision_store.get_signature_file(
295
from_sf = self.from_repository._revision_store.get_signature_file(
297
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
298
to_rf = self.to_repository._revision_store.get_revision_file(
300
from_rf = self.from_repository._revision_store.get_revision_file(
302
to_rf.join(from_rf, version_ids=revs)
285
305
class Inter1and2Helper(object):
286
306
"""Helper for operations that convert data from model 1 and 2
317
335
revs = revs[100:]
319
def _find_root_ids(self, revs, parent_map, graph):
321
planned_versions = {}
322
for tree in self.iter_rev_trees(revs):
323
revision_id = tree.inventory.root.revision
324
root_id = tree.get_root_id()
325
planned_versions.setdefault(root_id, []).append(revision_id)
326
revision_root[revision_id] = root_id
327
# Find out which parents we don't already know root ids for
329
for revision_parents in parent_map.itervalues():
330
parents.update(revision_parents)
331
parents.difference_update(revision_root.keys() + [NULL_REVISION])
332
# Limit to revisions present in the versionedfile
333
parents = graph.get_parent_map(parents).keys()
334
for tree in self.iter_rev_trees(parents):
335
root_id = tree.get_root_id()
336
revision_root[tree.get_revision_id()] = root_id
337
return revision_root, planned_versions
339
337
def generate_root_texts(self, revs):
340
338
"""Generate VersionedFiles for all root ids.
342
340
:param revs: the revisions to include
344
to_texts = self.target.texts
345
graph = self.source.get_graph()
346
parent_map = graph.get_parent_map(revs)
347
rev_order = topo_sort(parent_map)
348
rev_id_to_root_id, root_id_to_rev_ids = self._find_root_ids(
349
revs, parent_map, graph)
350
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
352
# Guaranteed stable, this groups all the file id operations together
353
# retaining topological order within the revisions of a file id.
354
# File id splits and joins would invalidate this, but they don't exist
355
# yet, and are unlikely to in non-rich-root environments anyway.
356
root_id_order.sort(key=operator.itemgetter(0))
357
# Create a record stream containing the roots to create.
359
for key in root_id_order:
360
root_id, rev_id = key
361
rev_parents = parent_map[rev_id]
362
# We drop revision parents with different file-ids, because
363
# that represents a rename of the root to a different location
364
# - its not actually a parent for us. (We could look for that
365
# file id in the revision tree at considerably more expense,
366
# but for now this is sufficient (and reconcile will catch and
367
# correct this anyway).
368
# When a parent revision is a ghost, we guess that its root id
369
# was unchanged (rather than trimming it from the parent list).
370
parent_keys = tuple((root_id, parent) for parent in rev_parents
371
if parent != NULL_REVISION and
372
rev_id_to_root_id.get(parent, root_id) == root_id)
373
yield FulltextContentFactory(key, parent_keys, None, '')
374
to_texts.insert_record_stream(yield_roots())
342
inventory_weave = self.source.get_inventory_weave()
345
to_store = self.target.weave_store
346
for tree in self.iter_rev_trees(revs):
347
revision_id = tree.inventory.root.revision
348
root_id = tree.inventory.root.file_id
349
parents = inventory_weave.get_parents(revision_id)
350
if root_id not in versionedfile:
351
versionedfile[root_id] = to_store.get_weave_or_empty(root_id,
352
self.target.get_transaction())
353
_, _, parent_texts[root_id] = versionedfile[root_id].add_lines(
354
revision_id, parents, [], parent_texts)
376
356
def regenerate_inventory(self, revs):
377
357
"""Generate a new inventory versionedfile in target, convertin data.
380
360
stored in the target (reserializing it in a different format).
381
361
:param revs: The revisions to include
363
inventory_weave = self.source.get_inventory_weave()
383
364
for tree in self.iter_rev_trees(revs):
384
parents = tree.get_parent_ids()
365
parents = inventory_weave.get_parents(tree.get_revision_id())
385
366
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
388
def fetch_revisions(self, revision_ids):
389
# TODO: should this batch them up rather than requesting 10,000
391
for revision in self.source.get_revisions(revision_ids):
392
self.target.add_revision(revision.revision_id, revision)
395
class Model1toKnit2Fetcher(RepoFetcher):
370
class Model1toKnit2Fetcher(GenericRepoFetcher):
396
371
"""Fetch from a Model1 repository into a Knit2 repository
398
def __init__(self, to_repository, from_repository, last_revision=None,
399
pb=None, find_ghosts=True):
373
def __init__(self, to_repository, from_repository, last_revision=None,
400
375
self.helper = Inter1and2Helper(from_repository, to_repository)
401
RepoFetcher.__init__(self, to_repository, from_repository,
402
last_revision, pb, find_ghosts)
376
GenericRepoFetcher.__init__(self, to_repository, from_repository,
404
379
def _generate_root_texts(self, revs):
405
380
self.helper.generate_root_texts(revs)
407
382
def _fetch_inventory_weave(self, revs, pb):
408
383
self.helper.regenerate_inventory(revs)
410
def _fetch_revision_texts(self, revs, pb):
411
"""Fetch revision object texts"""
415
pb.update('copying revisions', count, total)
417
sig_text = self.from_repository.get_signature_text(rev)
418
self.to_repository.add_signature_text(rev, sig_text)
419
except errors.NoSuchRevision:
422
self._copy_revision(rev)
425
def _copy_revision(self, rev):
426
self.helper.fetch_revisions([rev])
429
class Knit1to2Fetcher(RepoFetcher):
386
class Knit1to2Fetcher(KnitRepoFetcher):
430
387
"""Fetch from a Knit1 repository into a Knit2 repository"""
432
def __init__(self, to_repository, from_repository, last_revision=None,
433
pb=None, find_ghosts=True):
389
def __init__(self, to_repository, from_repository, last_revision=None,
434
391
self.helper = Inter1and2Helper(from_repository, to_repository)
435
RepoFetcher.__init__(self, to_repository, from_repository,
436
last_revision, pb, find_ghosts)
392
KnitRepoFetcher.__init__(self, to_repository, from_repository,
438
395
def _generate_root_texts(self, revs):
439
396
self.helper.generate_root_texts(revs)
441
398
def _fetch_inventory_weave(self, revs, pb):
442
399
self.helper.regenerate_inventory(revs)
444
def _fetch_just_revision_texts(self, version_ids):
445
self.helper.fetch_revisions(version_ids)