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.
36
import bzrlib.errors as errors
37
33
from bzrlib.errors import InstallFailed
38
34
from bzrlib.progress import ProgressPhase
39
35
from bzrlib.revision import NULL_REVISION
40
36
from bzrlib.tsort import topo_sort
41
37
from bzrlib.trace import mutter
43
from bzrlib.versionedfile import filter_absent, 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
39
from bzrlib.versionedfile import FulltextContentFactory
63
42
class RepoFetcher(object):
64
43
"""Pull revisions and texts from one repository to another.
67
if set, try to limit to the data this revision references.
70
count_copied -- number of revisions copied
72
45
This should not be used directly, it's essential a object to encapsulate
73
46
the logic in InterRepository.fetch().
76
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
49
def __init__(self, to_repository, from_repository, last_revision=None,
50
pb=None, find_ghosts=True, fetch_spec=None):
78
51
"""Create a repo fetcher.
53
:param last_revision: If set, try to limit to the data this revision
80
55
:param find_ghosts: If True search the entire history for ghosts.
81
56
:param _write_group_acquired_callable: Don't use; this parameter only
82
57
exists to facilitate a hack done in InterPackRepo.fetch. We would
83
58
like to remove this parameter.
59
:param pb: ProgressBar object to use; deprecated and ignored.
60
This method will just create one on top of the stack.
86
self.failed_revisions = []
63
symbol_versioning.warn(
64
symbol_versioning.deprecated_in((1, 14, 0))
65
% "pb parameter to RepoFetcher.__init__")
66
# and for simplicity it is in fact ignored
88
67
if to_repository.has_same_location(from_repository):
89
68
# repository.fetch should be taking care of this case.
90
69
raise errors.BzrError('RepoFetcher run '
95
74
self.sink = to_repository._get_sink()
96
75
# must not mutate self._last_revision as its potentially a shared instance
97
76
self._last_revision = last_revision
77
self._fetch_spec = fetch_spec
98
78
self.find_ghosts = find_ghosts
100
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
101
self.nested_pb = self.pb
104
self.nested_pb = None
105
79
self.from_repository.lock_read()
80
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
81
self.from_repository, self.from_repository._format,
82
self.to_repository, self.to_repository._format)
107
self.to_repository.lock_write()
109
self.to_repository.start_write_group()
113
self.to_repository.abort_write_group(suppress_errors=True)
116
self.to_repository.commit_write_group()
119
if self.nested_pb is not None:
120
self.nested_pb.finished()
122
self.to_repository.unlock()
124
86
self.from_repository.unlock()
126
88
def __fetch(self):
127
89
"""Primary worker function.
129
This initialises all the needed variables, and then fetches the
91
This initialises all the needed variables, and then fetches the
130
92
requested revisions, finally clearing the progress bar.
132
94
# Roughly this is what we're aiming for fetch to become:
137
99
# assert not missing
138
100
self.count_total = 0
139
101
self.file_ids_names = {}
140
pp = ProgressPhase('Transferring', 4, self.pb)
102
pb = bzrlib.ui.ui_factory.nested_progress_bar()
103
pb.show_pct = pb.show_count = False
105
pb.update("Finding revisions", 0, 2)
143
106
search = self._revids_to_fetch()
144
107
if search is None:
146
self._fetch_everything_for_search(search, pp)
109
pb.update("Fetching revisions", 1, 2)
110
self._fetch_everything_for_search(search)
150
def _fetch_everything_for_search(self, search, pp):
114
def _fetch_everything_for_search(self, search):
151
115
"""Fetch all data for the given set of revisions."""
152
116
# The first phase is "file". We pass the progress bar for it directly
153
117
# into item_keys_introduced_by, which has more information about how
157
121
# item_keys_introduced_by should have a richer API than it does at the
158
122
# moment, so that it can feed the progress information back to this
160
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
124
if (self.from_repository._format.rich_root_data and
125
not self.to_repository._format.rich_root_data):
126
raise errors.IncompatibleRepositories(
127
self.from_repository, self.to_repository,
128
"different rich-root support")
129
pb = bzrlib.ui.ui_factory.nested_progress_bar()
131
pb.update("Get stream source")
132
source = self.from_repository._get_source(
133
self.to_repository._format)
134
stream = source.get_stream(search)
162
135
from_format = self.from_repository._format
163
stream = self.get_stream(search, pp)
164
self.sink.insert_stream(stream, from_format)
136
pb.update("Inserting stream")
137
resume_tokens, missing_keys = self.sink.insert_stream(
138
stream, from_format, [])
140
pb.update("Missing keys")
141
stream = source.get_stream_for_missing_keys(missing_keys)
142
pb.update("Inserting missing keys")
143
resume_tokens, missing_keys = self.sink.insert_stream(
144
stream, from_format, resume_tokens)
146
raise AssertionError(
147
"second push failed to complete a fetch %r." % (
150
raise AssertionError(
151
"second push failed to commit the fetch %r." % (
153
pb.update("Finishing stream")
165
154
self.sink.finished()
167
if self.pb is not None:
170
def get_stream(self, search, pp):
172
revs = search.get_keys()
173
graph = self.from_repository.get_graph()
174
revs = list(graph.iter_topo_order(revs))
175
data_to_fetch = self.from_repository.item_keys_introduced_by(
178
for knit_kind, file_id, revisions in data_to_fetch:
179
if knit_kind != phase:
181
# Make a new progress bar for this phase
184
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
185
if knit_kind == "file":
186
# Accumulate file texts
187
text_keys.extend([(file_id, revision) for revision in
189
elif knit_kind == "inventory":
190
# Now copy the file texts.
191
to_texts = self.to_repository.texts
192
from_texts = self.from_repository.texts
193
yield ('texts', from_texts.get_record_stream(
194
text_keys, self.to_repository._fetch_order,
195
not self.to_repository._fetch_uses_deltas))
196
# Cause an error if a text occurs after we have done the
199
# Before we process the inventory we generate the root
200
# texts (if necessary) so that the inventories references
202
for _ in self._generate_root_texts(revs):
204
# NB: This currently reopens the inventory weave in source;
205
# using a single stream interface instead would avoid this.
206
self.pb.update("fetch inventory", 0, 1)
207
from_weave = self.from_repository.inventories
208
# we fetch only the referenced inventories because we do not
209
# know for unselected inventories whether all their required
210
# texts are present in the other repository - it could be
212
yield ('inventories', from_weave.get_record_stream(
213
[(rev_id,) for rev_id in revs],
214
self.inventory_fetch_order(),
215
not self.delta_on_metadata()))
216
elif knit_kind == "signatures":
217
# Nothing to do here; this will be taken care of when
218
# _fetch_revision_texts happens.
220
elif knit_kind == "revisions":
221
for _ in self._fetch_revision_texts(revs, self.pb):
224
raise AssertionError("Unknown knit kind %r" % knit_kind)
225
self.count_copied += len(revs)
227
158
def _revids_to_fetch(self):
228
159
"""Determines the exact revisions needed from self.from_repository to
244
177
except errors.NoSuchRevision, e:
245
178
raise InstallFailed([self._last_revision])
247
def _fetch_revision_texts(self, revs, pb):
248
# fetch signatures first and then the revision texts
249
# may need to be a InterRevisionStore call here.
250
from_sf = self.from_repository.signatures
251
# A missing signature is just skipped.
252
keys = [(rev_id,) for rev_id in revs]
253
signatures = filter_absent(from_sf.get_record_stream(
255
self.to_repository._fetch_order,
256
not self.to_repository._fetch_uses_deltas))
257
# If a revision has a delta, this is actually expanded inside the
258
# insert_record_stream code now, which is an alternate fix for
260
from_rf = self.from_repository.revisions
261
revisions = from_rf.get_record_stream(
263
self.to_repository._fetch_order,
264
not self.delta_on_metadata())
265
return [('signatures', signatures), ('revisions', revisions)]
267
def _generate_root_texts(self, revs):
268
"""This will be called by __fetch between fetching weave texts and
269
fetching the inventory weave.
271
Subclasses should override this if they need to generate root texts
272
after fetching weave texts.
276
def inventory_fetch_order(self):
277
return self.to_repository._fetch_order
279
def delta_on_metadata(self):
280
src_serializer = self.from_repository._format._serializer
281
target_serializer = self.to_repository._format._serializer
282
return (self.to_repository._fetch_uses_deltas and
283
src_serializer == target_serializer)
286
181
class Inter1and2Helper(object):
287
182
"""Helper for operations that convert data from model 1 and 2
289
184
This is for use by fetchers and converters.
370
265
rev_id_to_root_id.get(parent, root_id) == root_id)
371
266
yield FulltextContentFactory(key, parent_keys, None, '')
372
267
return [('texts', yield_roots())]
375
class Model1toKnit2Fetcher(RepoFetcher):
376
"""Fetch from a Model1 repository into a Knit2 repository
378
def __init__(self, to_repository, from_repository, last_revision=None,
379
pb=None, find_ghosts=True):
380
self.helper = Inter1and2Helper(from_repository)
381
RepoFetcher.__init__(self, to_repository, from_repository,
382
last_revision, pb, find_ghosts)
384
def _generate_root_texts(self, revs):
385
return self.helper.generate_root_texts(revs)
387
def inventory_fetch_order(self):
390
Knit1to2Fetcher = Model1toKnit2Fetcher