21
21
that has merged into it. As the first step of a merge, pull, or
22
22
branch operation we copy history from the source into the destination
25
The copying is done in a slightly complicated order. We don't want to
26
add a revision to the store until everything it refers to is also
27
stored, so that if a revision is present we can totally recreate it.
28
However, we can't know what files are included in a revision until we
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.
28
from bzrlib.lazy_import import lazy_import
29
lazy_import(globals(), """
36
import bzrlib.errors as errors
37
from bzrlib.errors import InstallFailed
38
from bzrlib.progress import ProgressPhase
40
39
from bzrlib.revision import NULL_REVISION
40
from bzrlib.tsort import topo_sort
41
41
from bzrlib.trace import mutter
43
from bzrlib.versionedfile import FulltextContentFactory
45
# TODO: Avoid repeatedly opening weaves so many times.
47
# XXX: This doesn't handle ghost (not present in branch) revisions at
48
# all yet. I'm not sure they really should be supported.
50
# NOTE: This doesn't copy revisions which may be present but not
51
# merged into the last revision. I'm not sure we want to do that.
53
# - get a list of revisions that need to be pulled in
54
# - for each one, pull in that revision file
55
# and get the inventory, and store the inventory with right
57
# - and get the ancestry, and store that with right parents too
58
# - and keep a note of all file ids and version seen
59
# - then go through all files; for each one get the weave,
60
# and add in all file versions
44
63
class RepoFetcher(object):
45
64
"""Pull revisions and texts from one repository to another.
67
if set, try to limit to the data this revision references.
47
71
This should not be used directly, it's essential a object to encapsulate
48
72
the logic in InterRepository.fetch().
51
def __init__(self, to_repository, from_repository, last_revision=None,
52
find_ghosts=True, fetch_spec=None):
75
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
53
77
"""Create a repo fetcher.
55
:param last_revision: If set, try to limit to the data this revision
57
79
:param find_ghosts: If True search the entire history for ghosts.
80
:param _write_group_acquired_callable: Don't use; this parameter only
81
exists to facilitate a hack done in InterPackRepo.fetch. We would
82
like to remove this parameter.
59
# repository.fetch has the responsibility for short-circuiting
60
# attempts to copy between a repository and itself.
86
self.failed_revisions = []
87
if to_repository.has_same_location(from_repository):
88
# repository.fetch should be taking care of this case.
89
raise errors.BzrError('RepoFetcher run '
90
'between two objects at the same location: '
91
'%r and %r' % (to_repository, from_repository))
61
92
self.to_repository = to_repository
62
93
self.from_repository = from_repository
63
94
self.sink = to_repository._get_sink()
64
95
# must not mutate self._last_revision as its potentially a shared instance
65
96
self._last_revision = last_revision
66
self._fetch_spec = fetch_spec
67
97
self.find_ghosts = find_ghosts
99
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
100
self.nested_pb = self.pb
103
self.nested_pb = None
68
104
self.from_repository.lock_read()
69
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
70
self.from_repository, self.from_repository._format,
71
self.to_repository, self.to_repository._format)
109
if self.nested_pb is not None:
110
self.nested_pb.finished()
75
112
self.from_repository.unlock()
88
125
# assert not missing
89
126
self.count_total = 0
90
127
self.file_ids_names = {}
91
pb = ui.ui_factory.nested_progress_bar()
92
pb.show_pct = pb.show_count = False
128
pp = ProgressPhase('Transferring', 4, self.pb)
94
pb.update("Finding revisions", 0, 2)
95
131
search = self._revids_to_fetch()
96
132
if search is None:
98
pb.update("Fetching revisions", 1, 2)
99
self._fetch_everything_for_search(search)
134
self._fetch_everything_for_search(search, pp)
103
def _fetch_everything_for_search(self, search):
138
def _fetch_everything_for_search(self, search, pp):
104
139
"""Fetch all data for the given set of revisions."""
105
140
# The first phase is "file". We pass the progress bar for it directly
106
141
# into item_keys_introduced_by, which has more information about how
115
150
raise errors.IncompatibleRepositories(
116
151
self.from_repository, self.to_repository,
117
152
"different rich-root support")
118
pb = ui.ui_factory.nested_progress_bar()
153
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
120
pb.update("Get stream source")
121
155
source = self.from_repository._get_source(
122
156
self.to_repository._format)
123
157
stream = source.get_stream(search)
124
158
from_format = self.from_repository._format
125
pb.update("Inserting stream")
126
159
resume_tokens, missing_keys = self.sink.insert_stream(
127
160
stream, from_format, [])
128
if self.to_repository._fallback_repositories:
130
self._parent_inventories(search.get_keys()))
132
pb.update("Missing keys")
133
162
stream = source.get_stream_for_missing_keys(missing_keys)
134
pb.update("Inserting missing keys")
135
163
resume_tokens, missing_keys = self.sink.insert_stream(
136
164
stream, from_format, resume_tokens)
154
183
If no revisions need to be fetched, then this just returns None.
156
if self._fetch_spec is not None:
157
return self._fetch_spec
158
185
mutter('fetch up to rev {%s}', self._last_revision)
159
186
if self._last_revision is NULL_REVISION:
160
187
# explicit limit of no revisions needed
162
return self.to_repository.search_missing_revision_ids(
163
self.from_repository, self._last_revision,
164
find_ghosts=self.find_ghosts)
166
def _parent_inventories(self, revision_ids):
167
# Find all the parent revisions referenced by the stream, but
168
# not present in the stream, and make sure we send their
170
parent_maps = self.to_repository.get_parent_map(revision_ids)
172
map(parents.update, parent_maps.itervalues())
173
parents.discard(NULL_REVISION)
174
parents.difference_update(revision_ids)
175
missing_keys = set(('inventories', rev_id) for rev_id in parents)
189
if (self._last_revision is not None and
190
self.to_repository.has_revision(self._last_revision)):
193
return self.to_repository.search_missing_revision_ids(
194
self.from_repository, self._last_revision,
195
find_ghosts=self.find_ghosts)
196
except errors.NoSuchRevision, e:
197
raise InstallFailed([self._last_revision])
179
200
class Inter1and2Helper(object):
243
267
# yet, and are unlikely to in non-rich-root environments anyway.
244
268
root_id_order.sort(key=operator.itemgetter(0))
245
269
# Create a record stream containing the roots to create.
247
# XXX: not covered by tests, should have a flag to always run
248
# this. -- mbp 20100129
249
graph = self.source_repo.get_known_graph_ancestry(revs)
250
new_roots_stream = _new_root_data_stream(
251
root_id_order, rev_id_to_root_id, parent_map, self.source, graph)
252
return [('texts', new_roots_stream)]
255
def _get_rich_root_heads_graph(source_repo, revision_ids):
256
"""Get a Graph object suitable for asking heads() for new rich roots."""
260
def _new_root_data_stream(
261
root_keys_to_create, rev_id_to_root_id_map, parent_map, repo, graph=None):
262
"""Generate a texts substream of synthesised root entries.
264
Used in fetches that do rich-root upgrades.
266
:param root_keys_to_create: iterable of (root_id, rev_id) pairs describing
267
the root entries to create.
268
:param rev_id_to_root_id_map: dict of known rev_id -> root_id mappings for
269
calculating the parents. If a parent rev_id is not found here then it
270
will be recalculated.
271
:param parent_map: a parent map for all the revisions in
273
:param graph: a graph to use instead of repo.get_graph().
275
for root_key in root_keys_to_create:
276
root_id, rev_id = root_key
277
parent_keys = _parent_keys_for_root_version(
278
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph)
279
yield versionedfile.FulltextContentFactory(
280
root_key, parent_keys, None, '')
283
def _parent_keys_for_root_version(
284
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph=None):
285
"""Get the parent keys for a given root id.
287
A helper function for _new_root_data_stream.
289
# Include direct parents of the revision, but only if they used the same
290
# root_id and are heads.
291
rev_parents = parent_map[rev_id]
293
for parent_id in rev_parents:
294
if parent_id == NULL_REVISION:
296
if parent_id not in rev_id_to_root_id_map:
297
# We probably didn't read this revision, go spend the extra effort
300
tree = repo.revision_tree(parent_id)
301
except errors.NoSuchRevision:
302
# Ghost, fill out rev_id_to_root_id in case we encounter this
304
# But set parent_root_id to None since we don't really know
305
parent_root_id = None
307
parent_root_id = tree.get_root_id()
308
rev_id_to_root_id_map[parent_id] = None
310
# rev_id_to_root_id_map[parent_id] = parent_root_id
311
# memory consumption maybe?
313
parent_root_id = rev_id_to_root_id_map[parent_id]
314
if root_id == parent_root_id:
315
# With stacking we _might_ want to refer to a non-local revision,
316
# but this code path only applies when we have the full content
317
# available, so ghosts really are ghosts, not just the edge of
319
parent_ids.append(parent_id)
321
# root_id may be in the parent anyway.
323
tree = repo.revision_tree(parent_id)
324
except errors.NoSuchRevision:
325
# ghost, can't refer to it.
329
parent_ids.append(tree.inventory[root_id].revision)
330
except errors.NoSuchId:
333
# Drop non-head parents
335
graph = repo.get_graph()
336
heads = graph.heads(parent_ids)
338
for parent_id in parent_ids:
339
if parent_id in heads and parent_id not in selected_ids:
340
selected_ids.append(parent_id)
341
parent_keys = [(root_id, parent_id) for parent_id in selected_ids]
271
for key in root_id_order:
272
root_id, rev_id = key
273
rev_parents = parent_map[rev_id]
274
# We drop revision parents with different file-ids, because
275
# that represents a rename of the root to a different location
276
# - its not actually a parent for us. (We could look for that
277
# file id in the revision tree at considerably more expense,
278
# but for now this is sufficient (and reconcile will catch and
279
# correct this anyway).
280
# When a parent revision is a ghost, we guess that its root id
281
# was unchanged (rather than trimming it from the parent list).
282
parent_keys = tuple((root_id, parent) for parent in rev_parents
283
if parent != NULL_REVISION and
284
rev_id_to_root_id.get(parent, root_id) == root_id)
285
yield FulltextContentFactory(key, parent_keys, None, '')
286
return [('texts', yield_roots())]
added
removed
bzrlib/fetch.py
