63
59
# and add in all file versions
66
class RepoFetcher(object):
67
"""Pull revisions and texts from one repository to another.
70
if set, try to limit to the data this revision references.
63
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
64
f = Fetcher(to_branch, from_branch, revision, pb)
65
return f.count_copied, f.failed_revisions
69
class Fetcher(object):
70
"""Pull revisions and texts from one branch to another.
72
This doesn't update the destination's history; that can be done
73
separately if desired.
76
If set, pull only up to this revision_id.
80
last_revision -- if last_revision
81
is given it will be that, otherwise the last revision of
73
84
count_copied -- number of revisions copied
75
This should not be used directly, it's essential a object to encapsulate
76
the logic in InterRepository.fetch().
86
count_weaves -- number of file weaves copied
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.
88
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
89
if to_branch == from_branch:
90
raise Exception("can't fetch from a branch to itself")
91
self.to_branch = to_branch
92
self.to_weaves = to_branch.weave_store
93
self.to_control = to_branch.control_weaves
94
self.from_branch = from_branch
95
self.from_weaves = from_branch.weave_store
96
self.from_control = from_branch.control_weaves
86
97
self.failed_revisions = []
87
98
self.count_copied = 0
88
if to_repository.has_same_location(from_repository):
89
# repository.fetch should be taking care of this case.
90
raise errors.BzrError('RepoFetcher run '
91
'between two objects at the same location: '
92
'%r and %r' % (to_repository, from_repository))
93
self.to_repository = to_repository
94
self.from_repository = from_repository
95
# must not mutate self._last_revision as its potentially a shared instance
96
self._last_revision = last_revision
97
self.find_ghosts = find_ghosts
100
self.count_weaves = 0
101
self.copied_file_ids = set()
99
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
100
self.nested_pb = self.pb
103
self.pb = bzrlib.ui.ui_factory.progress_bar()
103
self.nested_pb = None
104
self.from_repository.lock_read()
106
self.to_repository.lock_write()
108
self.to_repository.start_write_group()
112
self.to_repository.abort_write_group()
115
self.to_repository.commit_write_group()
118
if self.nested_pb is not None:
119
self.nested_pb.finished()
121
self.to_repository.unlock()
123
self.from_repository.unlock()
126
"""Primary worker function.
128
This initialises all the needed variables, and then fetches the
129
requested revisions, finally clearing the progress bar.
132
self.file_ids_names = {}
133
pp = ProgressPhase('Transferring', 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)
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
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
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)
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.
215
mutter('fetch up to rev {%s}', self._last_revision)
216
if self._last_revision is NULL_REVISION:
217
# explicit limit of no revisions needed
219
if (self._last_revision is not None and
220
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:
227
raise InstallFailed([self._last_revision])
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()
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
not self.to_repository._fetch_uses_deltas))
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
def _generate_root_texts(self, revs):
276
"""This will be called by __fetch between fetching weave texts and
277
fetching the inventory weave.
279
Subclasses should override this if they need to generate root texts
280
after fetching weave texts.
285
class Inter1and2Helper(object):
286
"""Helper for operations that convert data from model 1 and 2
288
This is for use by fetchers and converters.
291
def __init__(self, source, target):
294
:param source: The repository data comes from
295
:param target: The repository data goes to
300
def iter_rev_trees(self, revs):
301
"""Iterate through RevisionTrees efficiently.
303
Additionally, the inventory's revision_id is set if unset.
305
Trees are retrieved in batches of 100, and then yielded in the order
308
:param revs: A list of revision ids
310
# In case that revs is not a list.
313
for tree in self.source.revision_trees(revs[:100]):
314
if tree.inventory.revision_id is None:
315
tree.inventory.revision_id = tree.get_revision_id()
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
def generate_root_texts(self, revs):
340
"""Generate VersionedFiles for all root ids.
342
: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())
376
def regenerate_inventory(self, revs):
377
"""Generate a new inventory versionedfile in target, convertin data.
379
The inventory is retrieved from the source, (deserializing it), and
380
stored in the target (reserializing it in a different format).
381
:param revs: The revisions to include
383
for tree in self.iter_rev_trees(revs):
384
parents = tree.get_parent_ids()
385
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):
396
"""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):
400
self.helper = Inter1and2Helper(from_repository, to_repository)
401
RepoFetcher.__init__(self, to_repository, from_repository,
402
last_revision, pb, find_ghosts)
404
def _generate_root_texts(self, revs):
405
self.helper.generate_root_texts(revs)
407
def _fetch_inventory_weave(self, revs, pb):
408
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):
430
"""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):
434
self.helper = Inter1and2Helper(from_repository, to_repository)
435
RepoFetcher.__init__(self, to_repository, from_repository,
436
last_revision, pb, find_ghosts)
438
def _generate_root_texts(self, revs):
439
self.helper.generate_root_texts(revs)
441
def _fetch_inventory_weave(self, revs, pb):
442
self.helper.regenerate_inventory(revs)
444
def _fetch_just_revision_texts(self, version_ids):
445
self.helper.fetch_revisions(version_ids)
107
self.last_revision = self._find_last_revision(last_revision)
108
except NoSuchRevision, e:
109
mutter('failed getting last revision: %s', e)
110
raise InstallFailed([last_revision])
111
mutter('fetch up to rev {%s}', self.last_revision)
113
revs_to_fetch = self._compare_ancestries()
115
raise InstallFailed([self.last_revision])
116
self._copy_revisions(revs_to_fetch)
117
self.new_ancestry = revs_to_fetch
120
def _find_last_revision(self, last_revision):
121
"""Find the limiting source revision.
123
Every ancestor of that revision will be merged across.
125
Returns the revision_id, or returns None if there's no history
126
in the source branch."""
127
self.pb.update('get source history')
128
from_history = self.from_branch.revision_history()
129
self.pb.update('get destination history')
131
self.from_branch.get_revision(last_revision)
134
return from_history[-1]
136
return None # no history in the source branch
139
def _compare_ancestries(self):
140
"""Get a list of revisions that must be copied.
142
That is, every revision that's in the ancestry of the source
143
branch and not in the destination branch."""
144
self.pb.update('get source ancestry')
145
self.from_ancestry = self.from_branch.get_ancestry(self.last_revision)
147
dest_last_rev = self.to_branch.last_revision()
148
self.pb.update('get destination ancestry')
150
dest_ancestry = self.to_branch.get_ancestry(dest_last_rev)
153
ss = set(dest_ancestry)
155
for rev_id in self.from_ancestry:
157
to_fetch.append(rev_id)
158
mutter('need to get revision {%s}', rev_id)
159
mutter('need to get %d revisions in total', len(to_fetch))
160
self.count_total = len(to_fetch)
163
def _copy_revisions(self, revs_to_fetch):
165
for rev_id in revs_to_fetch:
169
if self.to_branch.has_revision(rev_id):
171
self.pb.update('fetch revision', i, self.count_total)
172
self._copy_one_revision(rev_id)
173
self.count_copied += 1
176
def _copy_one_revision(self, rev_id):
177
"""Copy revision and everything referenced by it."""
178
mutter('copying revision {%s}', rev_id)
179
rev_xml = self.from_branch.get_revision_xml(rev_id)
180
inv_xml = self.from_branch.get_inventory_xml(rev_id)
181
rev = serializer_v5.read_revision_from_string(rev_xml)
182
inv = serializer_v5.read_inventory_from_string(inv_xml)
183
assert rev.revision_id == rev_id
184
assert rev.inventory_sha1 == sha_string(inv_xml)
185
mutter(' commiter %s, %d parents',
188
self._copy_new_texts(rev_id, inv)
189
parents = rev.parent_ids
190
for parent in parents:
191
if not self.to_branch.has_revision(parent):
192
parents.pop(parents.index(parent))
193
self._copy_inventory(rev_id, inv_xml, parents)
194
self.to_branch.revision_store.add(StringIO(rev_xml), rev_id)
195
mutter('copied revision %s', rev_id)
198
def _copy_inventory(self, rev_id, inv_xml, parent_ids):
199
self.to_control.add_text('inventory', rev_id,
200
split_lines(inv_xml), parent_ids,
201
self.to_branch.get_transaction())
203
def _copy_new_texts(self, rev_id, inv):
204
"""Copy any new texts occuring in this revision."""
205
# TODO: Rather than writing out weaves every time, hold them
206
# in memory until everything's done? But this way is nicer
207
# if it's interrupted.
208
for path, ie in inv.iter_entries():
209
if ie.revision != rev_id:
211
mutter('%s {%s} is changed in this revision',
213
self._copy_one_weave(rev_id, ie.file_id)
216
def _copy_one_weave(self, rev_id, file_id):
217
"""Copy one file weave."""
218
mutter('copy file {%s} modified in {%s}', file_id, rev_id)
219
if file_id in self.copied_file_ids:
220
mutter('file {%s} already copied', file_id)
222
from_weave = self.from_weaves.get_weave(file_id,
223
self.from_branch.get_transaction())
224
to_weave = self.to_weaves.get_weave_or_empty(file_id,
225
self.to_branch.get_transaction())
227
to_weave.join(from_weave)
228
except errors.WeaveParentMismatch:
229
to_weave.reweave(from_weave)
230
self.to_weaves.put_weave(file_id, to_weave,
231
self.to_branch.get_transaction())
232
self.count_weaves += 1
233
self.copied_file_ids.add(file_id)
234
mutter('copied file {%s}', file_id)
added
removed
bzrlib/fetch.py
