63
58
# 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.
62
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
63
f = Fetcher(to_branch, from_branch, revision, pb)
64
return f.count_copied, f.failed_revisions
68
class Fetcher(object):
69
"""Pull revisions and texts from one branch to another.
71
This doesn't update the destination's history; that can be done
72
separately if desired.
75
If set, pull only up to this revision_id.
79
last_revision -- if last_revision
80
is given it will be that, otherwise the last revision of
73
83
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().
85
count_texts -- number of file texts 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.
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.
87
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
88
if to_branch == from_branch:
89
raise Exception("can't fetch from a branch to itself")
90
self.to_branch = to_branch
91
self.to_weaves = to_branch.weave_store
92
self.to_control = to_branch.control_weaves
93
self.from_branch = from_branch
94
self.from_weaves = from_branch.weave_store
95
self.from_control = from_branch.control_weaves
89
96
self.failed_revisions = []
90
97
self.count_copied = 0
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
self.to_repository = to_repository
97
self.from_repository = from_repository
98
# must not mutate self._last_revision as its potentially a shared instance
99
self._last_revision = last_revision
100
self.find_ghosts = find_ghosts
102
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
103
self.nested_pb = self.pb
101
self.pb = bzrlib.ui.ui_factory.progress_bar()
106
self.nested_pb = None
107
self.from_repository.lock_read()
109
self.to_repository.lock_write()
111
self.to_repository.start_write_group()
115
self.to_repository.abort_write_group(suppress_errors=True)
118
self.to_repository.commit_write_group()
121
if self.nested_pb is not None:
122
self.nested_pb.finished()
124
self.to_repository.unlock()
126
self.from_repository.unlock()
129
"""Primary worker function.
131
This initialises all the needed variables, and then fetches the
132
requested revisions, finally clearing the progress bar.
135
self.file_ids_names = {}
136
pp = ProgressPhase('Transferring', 4, self.pb)
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)
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.
218
mutter('fetch up to rev {%s}', self._last_revision)
219
if self._last_revision is NULL_REVISION:
220
# explicit limit of no revisions needed
222
if (self._last_revision is not None and
223
self.to_repository.has_revision(self._last_revision)):
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:
230
raise InstallFailed([self._last_revision])
232
def _fetch_inventory_weave(self, revs, pb):
233
pb.update("fetch inventory", 0, 2)
234
to_weave = self.to_repository.inventories
235
# just merge, this is optimisable and its means we don't
236
# copy unreferenced data such as not-needed inventories.
237
pb.update("fetch inventory", 1, 3)
238
from_weave = self.from_repository.inventories
239
pb.update("fetch inventory", 2, 3)
240
# we fetch only the referenced inventories because we do not
241
# know for unselected inventories whether all their required
242
# texts are present in the other repository - it could be
244
to_weave.insert_record_stream(from_weave.get_record_stream(
245
[(rev_id,) for rev_id in revs],
246
self.to_repository._fetch_order,
247
not self.to_repository._fetch_uses_deltas))
249
def _fetch_revision_texts(self, revs, pb):
250
# fetch signatures first and then the revision texts
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
not 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
# If a revision has a delta, this is actually expanded inside the
265
# insert_record_stream code now, which is an alternate fix for
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,
270
not self.to_repository._fetch_uses_deltas))
272
def _generate_root_texts(self, revs):
273
"""This will be called by __fetch between fetching weave texts and
274
fetching the inventory weave.
276
Subclasses should override this if they need to generate root texts
277
after fetching weave texts.
282
class Inter1and2Helper(object):
283
"""Helper for operations that convert data from model 1 and 2
285
This is for use by fetchers and converters.
288
def __init__(self, source, target):
291
:param source: The repository data comes from
292
:param target: The repository data goes to
297
def iter_rev_trees(self, revs):
298
"""Iterate through RevisionTrees efficiently.
300
Additionally, the inventory's revision_id is set if unset.
302
Trees are retrieved in batches of 100, and then yielded in the order
305
:param revs: A list of revision ids
307
# In case that revs is not a list.
310
for tree in self.source.revision_trees(revs[:100]):
311
if tree.inventory.revision_id is None:
312
tree.inventory.revision_id = tree.get_revision_id()
316
def _find_root_ids(self, revs, parent_map, graph):
318
planned_versions = {}
319
for tree in self.iter_rev_trees(revs):
320
revision_id = tree.inventory.root.revision
321
root_id = tree.get_root_id()
322
planned_versions.setdefault(root_id, []).append(revision_id)
323
revision_root[revision_id] = root_id
324
# Find out which parents we don't already know root ids for
326
for revision_parents in parent_map.itervalues():
327
parents.update(revision_parents)
328
parents.difference_update(revision_root.keys() + [NULL_REVISION])
329
# Limit to revisions present in the versionedfile
330
parents = graph.get_parent_map(parents).keys()
331
for tree in self.iter_rev_trees(parents):
332
root_id = tree.get_root_id()
333
revision_root[tree.get_revision_id()] = root_id
334
return revision_root, planned_versions
336
def generate_root_texts(self, revs):
337
"""Generate VersionedFiles for all root ids.
339
:param revs: the revisions to include
341
to_texts = self.target.texts
342
graph = self.source.get_graph()
343
parent_map = graph.get_parent_map(revs)
344
rev_order = topo_sort(parent_map)
345
rev_id_to_root_id, root_id_to_rev_ids = self._find_root_ids(
346
revs, parent_map, graph)
347
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
349
# Guaranteed stable, this groups all the file id operations together
350
# retaining topological order within the revisions of a file id.
351
# File id splits and joins would invalidate this, but they don't exist
352
# yet, and are unlikely to in non-rich-root environments anyway.
353
root_id_order.sort(key=operator.itemgetter(0))
354
# Create a record stream containing the roots to create.
356
for key in root_id_order:
357
root_id, rev_id = key
358
rev_parents = parent_map[rev_id]
359
# We drop revision parents with different file-ids, because
360
# that represents a rename of the root to a different location
361
# - its not actually a parent for us. (We could look for that
362
# file id in the revision tree at considerably more expense,
363
# but for now this is sufficient (and reconcile will catch and
364
# correct this anyway).
365
# When a parent revision is a ghost, we guess that its root id
366
# was unchanged (rather than trimming it from the parent list).
367
parent_keys = tuple((root_id, parent) for parent in rev_parents
368
if parent != NULL_REVISION and
369
rev_id_to_root_id.get(parent, root_id) == root_id)
370
yield FulltextContentFactory(key, parent_keys, None, '')
371
to_texts.insert_record_stream(yield_roots())
373
def regenerate_inventory(self, revs):
374
"""Generate a new inventory versionedfile in target, convertin data.
376
The inventory is retrieved from the source, (deserializing it), and
377
stored in the target (reserializing it in a different format).
378
:param revs: The revisions to include
380
for tree in self.iter_rev_trees(revs):
381
parents = tree.get_parent_ids()
382
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
385
def fetch_revisions(self, revision_ids):
386
# TODO: should this batch them up rather than requesting 10,000
388
for revision in self.source.get_revisions(revision_ids):
389
self.target.add_revision(revision.revision_id, revision)
392
class Model1toKnit2Fetcher(RepoFetcher):
393
"""Fetch from a Model1 repository into a Knit2 repository
395
def __init__(self, to_repository, from_repository, last_revision=None,
396
pb=None, find_ghosts=True):
397
self.helper = Inter1and2Helper(from_repository, to_repository)
398
RepoFetcher.__init__(self, to_repository, from_repository,
399
last_revision, pb, find_ghosts)
401
def _generate_root_texts(self, revs):
402
self.helper.generate_root_texts(revs)
404
def _fetch_inventory_weave(self, revs, pb):
405
self.helper.regenerate_inventory(revs)
407
def _fetch_revision_texts(self, revs, pb):
408
"""Fetch revision object texts"""
412
pb.update('copying revisions', count, total)
414
sig_text = self.from_repository.get_signature_text(rev)
415
self.to_repository.add_signature_text(rev, sig_text)
416
except errors.NoSuchRevision:
419
self._copy_revision(rev)
422
def _copy_revision(self, rev):
423
self.helper.fetch_revisions([rev])
426
class Knit1to2Fetcher(RepoFetcher):
427
"""Fetch from a Knit1 repository into a Knit2 repository"""
429
def __init__(self, to_repository, from_repository, last_revision=None,
430
pb=None, find_ghosts=True):
431
self.helper = Inter1and2Helper(from_repository, to_repository)
432
RepoFetcher.__init__(self, to_repository, from_repository,
433
last_revision, pb, find_ghosts)
435
def _generate_root_texts(self, revs):
436
self.helper.generate_root_texts(revs)
438
def _fetch_inventory_weave(self, revs, pb):
439
self.helper.regenerate_inventory(revs)
441
def _fetch_just_revision_texts(self, version_ids):
442
self.helper.fetch_revisions(version_ids)
105
self.last_revision = self._find_last_revision(last_revision)
106
except NoSuchRevision, e:
107
mutter('failed getting last revision: %s', e)
108
raise InstallFailed([last_revision])
109
mutter('fetch up to rev {%s}', self.last_revision)
111
revs_to_fetch = self._compare_ancestries()
113
raise InstallFailed([self.last_revision])
114
self._copy_revisions(revs_to_fetch)
115
self.new_ancestry = revs_to_fetch
118
def _find_last_revision(self, last_revision):
119
"""Find the limiting source revision.
121
Every ancestor of that revision will be merged across.
123
Returns the revision_id, or returns None if there's no history
124
in the source branch."""
125
self.pb.update('get source history')
126
from_history = self.from_branch.revision_history()
127
self.pb.update('get destination history')
129
self.from_branch.get_revision(last_revision)
132
return from_history[-1]
134
return None # no history in the source branch
137
def _compare_ancestries(self):
138
"""Get a list of revisions that must be copied.
140
That is, every revision that's in the ancestry of the source
141
branch and not in the destination branch."""
142
self.pb.update('get source ancestry')
143
self.from_ancestry = self.from_branch.get_ancestry(self.last_revision)
145
dest_last_rev = self.to_branch.last_revision()
146
self.pb.update('get destination ancestry')
148
dest_ancestry = self.to_branch.get_ancestry(dest_last_rev)
151
ss = set(dest_ancestry)
153
for rev_id in self.from_ancestry:
155
to_fetch.append(rev_id)
156
mutter('need to get revision {%s}', rev_id)
157
mutter('need to get %d revisions in total', len(to_fetch))
158
self.count_total = len(to_fetch)
161
def _copy_revisions(self, revs_to_fetch):
163
for rev_id in revs_to_fetch:
167
if self.to_branch.has_revision(rev_id):
169
self.pb.update('fetch revision', i, self.count_total)
170
self._copy_one_revision(rev_id)
171
self.count_copied += 1
174
def _copy_one_revision(self, rev_id):
175
"""Copy revision and everything referenced by it."""
176
mutter('copying revision {%s}', rev_id)
177
rev_xml = self.from_branch.get_revision_xml(rev_id)
178
inv_xml = self.from_branch.get_inventory_xml(rev_id)
179
rev = serializer_v5.read_revision_from_string(rev_xml)
180
inv = serializer_v5.read_inventory_from_string(inv_xml)
181
assert rev.revision_id == rev_id
182
assert rev.inventory_sha1 == sha_string(inv_xml)
183
mutter(' commiter %s, %d parents',
186
self._copy_new_texts(rev_id, inv)
187
parents = rev.parent_ids
188
for parent in parents:
189
if not self.to_branch.has_revision(parent):
190
parents.pop(parents.index(parent))
191
self._copy_inventory(rev_id, inv_xml, parents)
192
self._copy_ancestry(rev_id, parents)
193
self.to_branch.revision_store.add(StringIO(rev_xml), rev_id)
196
def _copy_inventory(self, rev_id, inv_xml, parent_ids):
197
self.to_control.add_text('inventory', rev_id,
198
split_lines(inv_xml), parent_ids)
201
def _copy_ancestry(self, rev_id, parent_ids):
202
ancestry_lines = self.from_control.get_lines('ancestry', rev_id)
203
self.to_control.add_text('ancestry', rev_id, ancestry_lines,
207
def _copy_new_texts(self, rev_id, inv):
208
"""Copy any new texts occuring in this revision."""
209
# TODO: Rather than writing out weaves every time, hold them
210
# in memory until everything's done? But this way is nicer
211
# if it's interrupted.
212
for path, ie in inv.iter_entries():
213
if ie.kind != 'file':
215
if ie.revision != rev_id:
217
mutter('%s {%s} is changed in this revision',
219
self._copy_one_text(rev_id, ie.file_id)
222
def _copy_one_text(self, rev_id, file_id):
223
"""Copy one file text."""
224
mutter('copy text version {%s} of file {%s}',
226
from_weave = self.from_weaves.get_weave(file_id)
227
from_idx = from_weave.lookup(rev_id)
228
from_parents = map(from_weave.idx_to_name, from_weave.parents(from_idx))
229
text_lines = from_weave.get(from_idx)
230
to_weave = self.to_weaves.get_weave_or_empty(file_id)
231
to_parents = map(to_weave.lookup, from_parents)
232
# it's ok to add even if the text is already there
233
to_weave.add(rev_id, to_parents, text_lines)
234
self.to_weaves.put_weave(file_id, to_weave)
235
self.count_texts += 1
added
removed
bzrlib/fetch.py
