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
43
39
from bzrlib.revision import NULL_REVISION
40
from bzrlib.tsort import topo_sort
44
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
47
63
class RepoFetcher(object):
48
64
"""Pull revisions and texts from one repository to another.
67
if set, try to limit to the data this revision references.
50
69
This should not be used directly, it's essential a object to encapsulate
51
70
the logic in InterRepository.fetch().
54
def __init__(self, to_repository, from_repository, last_revision=None,
55
find_ghosts=True, fetch_spec=None):
73
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
56
75
"""Create a repo fetcher.
58
:param last_revision: If set, try to limit to the data this revision
60
77
:param find_ghosts: If True search the entire history for ghosts.
78
:param _write_group_acquired_callable: Don't use; this parameter only
79
exists to facilitate a hack done in InterPackRepo.fetch. We would
80
like to remove this parameter.
62
# repository.fetch has the responsibility for short-circuiting
63
# attempts to copy between a repository and itself.
82
if to_repository.has_same_location(from_repository):
83
# repository.fetch should be taking care of this case.
84
raise errors.BzrError('RepoFetcher run '
85
'between two objects at the same location: '
86
'%r and %r' % (to_repository, from_repository))
64
87
self.to_repository = to_repository
65
88
self.from_repository = from_repository
66
89
self.sink = to_repository._get_sink()
67
90
# must not mutate self._last_revision as its potentially a shared instance
68
91
self._last_revision = last_revision
69
self._fetch_spec = fetch_spec
70
92
self.find_ghosts = find_ghosts
94
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
95
self.nested_pb = self.pb
71
99
self.from_repository.lock_read()
72
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
73
self.from_repository, self.from_repository._format,
74
self.to_repository, self.to_repository._format)
104
if self.nested_pb is not None:
105
self.nested_pb.finished()
78
107
self.from_repository.unlock()
91
120
# assert not missing
92
121
self.count_total = 0
93
122
self.file_ids_names = {}
94
pb = ui.ui_factory.nested_progress_bar()
95
pb.show_pct = pb.show_count = False
123
pp = ProgressPhase('Transferring', 4, self.pb)
97
pb.update("Finding revisions", 0, 2)
98
126
search = self._revids_to_fetch()
99
127
if search is None:
101
pb.update("Fetching revisions", 1, 2)
102
self._fetch_everything_for_search(search)
129
self._fetch_everything_for_search(search, pp)
106
def _fetch_everything_for_search(self, search):
133
def _fetch_everything_for_search(self, search, pp):
107
134
"""Fetch all data for the given set of revisions."""
108
135
# The first phase is "file". We pass the progress bar for it directly
109
136
# into item_keys_introduced_by, which has more information about how
118
145
raise errors.IncompatibleRepositories(
119
146
self.from_repository, self.to_repository,
120
147
"different rich-root support")
121
pb = ui.ui_factory.nested_progress_bar()
148
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
123
pb.update("Get stream source")
124
150
source = self.from_repository._get_source(
125
151
self.to_repository._format)
126
152
stream = source.get_stream(search)
127
153
from_format = self.from_repository._format
128
pb.update("Inserting stream")
129
154
resume_tokens, missing_keys = self.sink.insert_stream(
130
155
stream, from_format, [])
131
if self.to_repository._fallback_repositories:
133
self._parent_inventories(search.get_keys()))
135
pb.update("Missing keys")
136
157
stream = source.get_stream_for_missing_keys(missing_keys)
137
pb.update("Inserting missing keys")
138
158
resume_tokens, missing_keys = self.sink.insert_stream(
139
159
stream, from_format, resume_tokens)
157
177
If no revisions need to be fetched, then this just returns None.
159
if self._fetch_spec is not None:
160
return self._fetch_spec
161
179
mutter('fetch up to rev {%s}', self._last_revision)
162
180
if self._last_revision is NULL_REVISION:
163
181
# explicit limit of no revisions needed
165
return self.to_repository.search_missing_revision_ids(
166
self.from_repository, self._last_revision,
167
find_ghosts=self.find_ghosts)
169
def _parent_inventories(self, revision_ids):
170
# Find all the parent revisions referenced by the stream, but
171
# not present in the stream, and make sure we send their
173
parent_maps = self.to_repository.get_parent_map(revision_ids)
175
map(parents.update, parent_maps.itervalues())
176
parents.discard(NULL_REVISION)
177
parents.difference_update(revision_ids)
178
missing_keys = set(('inventories', rev_id) for rev_id in parents)
183
if (self._last_revision is not None and
184
self.to_repository.has_revision(self._last_revision)):
187
return self.to_repository.search_missing_revision_ids(
188
self.from_repository, self._last_revision,
189
find_ghosts=self.find_ghosts)
190
except errors.NoSuchRevision, e:
191
raise InstallFailed([self._last_revision])
182
194
class Inter1and2Helper(object):
246
261
# yet, and are unlikely to in non-rich-root environments anyway.
247
262
root_id_order.sort(key=operator.itemgetter(0))
248
263
# Create a record stream containing the roots to create.
250
# XXX: not covered by tests, should have a flag to always run
251
# this. -- mbp 20100129
252
graph = _get_rich_root_heads_graph(self.source, revs)
253
new_roots_stream = _new_root_data_stream(
254
root_id_order, rev_id_to_root_id, parent_map, self.source, graph)
255
return [('texts', new_roots_stream)]
258
def _get_rich_root_heads_graph(source_repo, revision_ids):
259
"""Get a Graph object suitable for asking heads() for new rich roots."""
260
st = static_tuple.StaticTuple
261
revision_keys = [st(r_id).intern() for r_id in revision_ids]
262
known_graph = source_repo.revisions.get_known_graph_ancestry(
264
return _mod_graph.GraphThunkIdsToKeys(known_graph)
267
def _new_root_data_stream(
268
root_keys_to_create, rev_id_to_root_id_map, parent_map, repo, graph=None):
269
"""Generate a texts substream of synthesised root entries.
271
Used in fetches that do rich-root upgrades.
273
:param root_keys_to_create: iterable of (root_id, rev_id) pairs describing
274
the root entries to create.
275
:param rev_id_to_root_id_map: dict of known rev_id -> root_id mappings for
276
calculating the parents. If a parent rev_id is not found here then it
277
will be recalculated.
278
:param parent_map: a parent map for all the revisions in
280
:param graph: a graph to use instead of repo.get_graph().
282
for root_key in root_keys_to_create:
283
root_id, rev_id = root_key
284
parent_keys = _parent_keys_for_root_version(
285
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph)
286
yield versionedfile.FulltextContentFactory(
287
root_key, parent_keys, None, '')
290
def _parent_keys_for_root_version(
291
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph=None):
292
"""Get the parent keys for a given root id.
294
A helper function for _new_root_data_stream.
296
# Include direct parents of the revision, but only if they used the same
297
# root_id and are heads.
298
rev_parents = parent_map[rev_id]
300
for parent_id in rev_parents:
301
if parent_id == NULL_REVISION:
303
if parent_id not in rev_id_to_root_id_map:
304
# We probably didn't read this revision, go spend the extra effort
307
tree = repo.revision_tree(parent_id)
308
except errors.NoSuchRevision:
309
# Ghost, fill out rev_id_to_root_id in case we encounter this
311
# But set parent_root_id to None since we don't really know
312
parent_root_id = None
314
parent_root_id = tree.get_root_id()
315
rev_id_to_root_id_map[parent_id] = None
317
# rev_id_to_root_id_map[parent_id] = parent_root_id
318
# memory consumption maybe?
320
parent_root_id = rev_id_to_root_id_map[parent_id]
321
if root_id == parent_root_id:
322
# With stacking we _might_ want to refer to a non-local revision,
323
# but this code path only applies when we have the full content
324
# available, so ghosts really are ghosts, not just the edge of
326
parent_ids.append(parent_id)
328
# root_id may be in the parent anyway.
330
tree = repo.revision_tree(parent_id)
331
except errors.NoSuchRevision:
332
# ghost, can't refer to it.
336
parent_ids.append(tree.inventory[root_id].revision)
337
except errors.NoSuchId:
340
# Drop non-head parents
342
graph = repo.get_graph()
343
heads = graph.heads(parent_ids)
345
for parent_id in parent_ids:
346
if parent_id in heads and parent_id not in selected_ids:
347
selected_ids.append(parent_id)
348
parent_keys = [(root_id, parent_id) for parent_id in selected_ids]
265
for key in root_id_order:
266
root_id, rev_id = key
267
rev_parents = parent_map[rev_id]
268
# We drop revision parents with different file-ids, because
269
# that represents a rename of the root to a different location
270
# - its not actually a parent for us. (We could look for that
271
# file id in the revision tree at considerably more expense,
272
# but for now this is sufficient (and reconcile will catch and
273
# correct this anyway).
274
# When a parent revision is a ghost, we guess that its root id
275
# was unchanged (rather than trimming it from the parent list).
276
parent_keys = tuple((root_id, parent) for parent in rev_parents
277
if parent != NULL_REVISION and
278
rev_id_to_root_id.get(parent, root_id) == root_id)
279
yield FulltextContentFactory(key, parent_keys, None, '')
280
return [('texts', yield_roots())]
added
removed
bzrlib/fetch.py
