21
34
that has merged into it. As the first step of a merge, pull, or
22
35
branch operation we copy history from the source into the destination
38
The copying is done in a slightly complicated order. We don't want to
39
add a revision to the store until everything it refers to is also
40
stored, so that if a revision is present we can totally recreate it.
41
However, we can't know what files are included in a revision until we
42
read its inventory. Therefore, we first pull the XML and hold it in
43
memory until we've updated all of the files referenced.
28
from bzrlib.lazy_import import lazy_import
29
lazy_import(globals(), """
40
from bzrlib.revision import NULL_REVISION
41
from bzrlib.trace import mutter
45
class RepoFetcher(object):
46
"""Pull revisions and texts from one repository to another.
48
This should not be used directly, it's essential a object to encapsulate
49
the logic in InterRepository.fetch().
46
# TODO: Avoid repeatedly opening weaves so many times.
48
# XXX: This doesn't handle ghost (not present in branch) revisions at
49
# all yet. I'm not sure they really should be supported.
51
# NOTE: This doesn't copy revisions which may be present but not
52
# merged into the last revision. I'm not sure we want to do that.
54
# - get a list of revisions that need to be pulled in
55
# - for each one, pull in that revision file
56
# and get the inventory, and store the inventory with right
58
# - and get the ancestry, and store that with right parents too
59
# - and keep a note of all file ids and version seen
60
# - then go through all files; for each one get the weave,
61
# and add in all file versions
65
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
66
f = Fetcher(to_branch, from_branch, revision, pb)
67
return f.count_copied, f.failed_revisions
71
class Fetcher(object):
72
"""Pull revisions and texts from one branch to another.
74
This doesn't update the destination's history; that can be done
75
separately if desired.
78
If set, pull only up to this revision_id.
82
last_revision -- if last_revision
83
is given it will be that, otherwise the last revision of
86
count_copied -- number of revisions copied
88
count_weaves -- number of file weaves copied
52
def __init__(self, to_repository, from_repository, last_revision=None,
53
pb=None, find_ghosts=True, fetch_spec=None):
54
"""Create a repo fetcher.
56
:param last_revision: If set, try to limit to the data this revision
58
:param find_ghosts: If True search the entire history for ghosts.
59
:param pb: ProgressBar object to use; deprecated and ignored.
60
This method will just create one on top of the stack.
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
67
# repository.fetch has the responsibility for short-circuiting
68
# attempts to copy between a repository and itself.
69
self.to_repository = to_repository
70
self.from_repository = from_repository
71
self.sink = to_repository._get_sink()
72
# must not mutate self._last_revision as its potentially a shared instance
73
self._last_revision = last_revision
74
self._fetch_spec = fetch_spec
75
self.find_ghosts = find_ghosts
76
self.from_repository.lock_read()
77
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
78
self.from_repository, self.from_repository._format,
79
self.to_repository, self.to_repository._format)
83
self.from_repository.unlock()
86
"""Primary worker function.
88
This initialises all the needed variables, and then fetches the
89
requested revisions, finally clearing the progress bar.
91
# Roughly this is what we're aiming for fetch to become:
93
# missing = self.sink.insert_stream(self.source.get_stream(search))
95
# missing = self.sink.insert_stream(self.source.get_items(missing))
90
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
91
if to_branch == from_branch:
92
raise Exception("can't fetch from a branch to itself")
93
self.to_branch = to_branch
94
self.to_weaves = to_branch.weave_store
95
self.to_control = to_branch.control_weaves
96
self.from_branch = from_branch
97
self.from_weaves = from_branch.weave_store
98
self.from_control = from_branch.control_weaves
99
self.failed_revisions = []
100
self.count_copied = 0
97
101
self.count_total = 0
98
self.file_ids_names = {}
99
pb = bzrlib.ui.ui_factory.nested_progress_bar()
100
pb.show_pct = pb.show_count = False
102
pb.update("Finding revisions", 0, 2)
103
search = self._revids_to_fetch()
106
pb.update("Fetching revisions", 1, 2)
107
self._fetch_everything_for_search(search)
111
def _fetch_everything_for_search(self, search):
112
"""Fetch all data for the given set of revisions."""
113
# The first phase is "file". We pass the progress bar for it directly
114
# into item_keys_introduced_by, which has more information about how
115
# that phase is progressing than we do. Progress updates for the other
116
# phases are taken care of in this function.
117
# XXX: there should be a clear owner of the progress reporting. Perhaps
118
# item_keys_introduced_by should have a richer API than it does at the
119
# moment, so that it can feed the progress information back to this
121
if (self.from_repository._format.rich_root_data and
122
not self.to_repository._format.rich_root_data):
123
raise errors.IncompatibleRepositories(
124
self.from_repository, self.to_repository,
125
"different rich-root support")
126
pb = bzrlib.ui.ui_factory.nested_progress_bar()
128
pb.update("Get stream source")
129
source = self.from_repository._get_source(
130
self.to_repository._format)
131
stream = source.get_stream(search)
132
from_format = self.from_repository._format
133
pb.update("Inserting stream")
134
resume_tokens, missing_keys = self.sink.insert_stream(
135
stream, from_format, [])
136
if self.to_repository._fallback_repositories:
138
self._parent_inventories(search.get_keys()))
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")
158
def _revids_to_fetch(self):
159
"""Determines the exact revisions needed from self.from_repository to
160
install self._last_revision in self.to_repository.
162
If no revisions need to be fetched, then this just returns None.
164
if self._fetch_spec is not None:
165
return self._fetch_spec
166
mutter('fetch up to rev {%s}', self._last_revision)
167
if self._last_revision is NULL_REVISION:
168
# explicit limit of no revisions needed
170
return self.to_repository.search_missing_revision_ids(
171
self.from_repository, self._last_revision,
172
find_ghosts=self.find_ghosts)
174
def _parent_inventories(self, revision_ids):
175
# Find all the parent revisions referenced by the stream, but
176
# not present in the stream, and make sure we send their
178
parent_maps = self.to_repository.get_parent_map(revision_ids)
180
map(parents.update, parent_maps.itervalues())
181
parents.discard(NULL_REVISION)
182
parents.difference_update(revision_ids)
183
missing_keys = set(('inventories', rev_id) for rev_id in parents)
187
class Inter1and2Helper(object):
188
"""Helper for operations that convert data from model 1 and 2
190
This is for use by fetchers and converters.
193
def __init__(self, source):
196
:param source: The repository data comes from
200
def iter_rev_trees(self, revs):
201
"""Iterate through RevisionTrees efficiently.
203
Additionally, the inventory's revision_id is set if unset.
205
Trees are retrieved in batches of 100, and then yielded in the order
208
:param revs: A list of revision ids
210
# In case that revs is not a list.
213
for tree in self.source.revision_trees(revs[:100]):
214
if tree.inventory.revision_id is None:
215
tree.inventory.revision_id = tree.get_revision_id()
219
def _find_root_ids(self, revs, parent_map, graph):
221
for tree in self.iter_rev_trees(revs):
222
revision_id = tree.inventory.root.revision
223
root_id = tree.get_root_id()
224
revision_root[revision_id] = root_id
225
# Find out which parents we don't already know root ids for
227
for revision_parents in parent_map.itervalues():
228
parents.update(revision_parents)
229
parents.difference_update(revision_root.keys() + [NULL_REVISION])
230
# Limit to revisions present in the versionedfile
231
parents = graph.get_parent_map(parents).keys()
232
for tree in self.iter_rev_trees(parents):
233
root_id = tree.get_root_id()
234
revision_root[tree.get_revision_id()] = root_id
237
def generate_root_texts(self, revs):
238
"""Generate VersionedFiles for all root ids.
240
:param revs: the revisions to include
242
graph = self.source.get_graph()
243
parent_map = graph.get_parent_map(revs)
244
rev_order = tsort.topo_sort(parent_map)
245
rev_id_to_root_id = self._find_root_ids(revs, parent_map, graph)
246
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
248
# Guaranteed stable, this groups all the file id operations together
249
# retaining topological order within the revisions of a file id.
250
# File id splits and joins would invalidate this, but they don't exist
251
# yet, and are unlikely to in non-rich-root environments anyway.
252
root_id_order.sort(key=operator.itemgetter(0))
253
# Create a record stream containing the roots to create.
254
from bzrlib.graph import FrozenHeadsCache
255
graph = FrozenHeadsCache(graph)
256
new_roots_stream = _new_root_data_stream(
257
root_id_order, rev_id_to_root_id, parent_map, self.source, graph)
258
return [('texts', new_roots_stream)]
261
def _new_root_data_stream(
262
root_keys_to_create, rev_id_to_root_id_map, parent_map, repo, graph=None):
263
"""Generate a texts substream of synthesised root entries.
265
Used in fetches that do rich-root upgrades.
267
:param root_keys_to_create: iterable of (root_id, rev_id) pairs describing
268
the root entries to create.
269
:param rev_id_to_root_id_map: dict of known rev_id -> root_id mappings for
270
calculating the parents. If a parent rev_id is not found here then it
271
will be recalculated.
272
:param parent_map: a parent map for all the revisions in
274
:param graph: a graph to use instead of repo.get_graph().
276
for root_key in root_keys_to_create:
277
root_id, rev_id = root_key
278
parent_keys = _parent_keys_for_root_version(
279
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph)
280
yield versionedfile.FulltextContentFactory(
281
root_key, parent_keys, None, '')
284
def _parent_keys_for_root_version(
285
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph=None):
286
"""Get the parent keys for a given root id.
288
A helper function for _new_root_data_stream.
290
# Include direct parents of the revision, but only if they used the same
291
# root_id and are heads.
292
rev_parents = parent_map[rev_id]
294
for parent_id in rev_parents:
295
if parent_id == NULL_REVISION:
297
if parent_id not in rev_id_to_root_id_map:
298
# We probably didn't read this revision, go spend the extra effort
301
tree = repo.revision_tree(parent_id)
302
except errors.NoSuchRevision:
303
# Ghost, fill out rev_id_to_root_id in case we encounter this
305
# But set parent_root_id to None since we don't really know
306
parent_root_id = None
308
parent_root_id = tree.get_root_id()
309
rev_id_to_root_id_map[parent_id] = None
311
# rev_id_to_root_id_map[parent_id] = parent_root_id
312
# memory consumption maybe?
314
parent_root_id = rev_id_to_root_id_map[parent_id]
315
if root_id == parent_root_id:
316
# With stacking we _might_ want to refer to a non-local revision,
317
# but this code path only applies when we have the full content
318
# available, so ghosts really are ghosts, not just the edge of
320
parent_ids.append(parent_id)
322
# root_id may be in the parent anyway.
324
tree = repo.revision_tree(parent_id)
325
except errors.NoSuchRevision:
326
# ghost, can't refer to it.
330
parent_ids.append(tree.inventory[root_id].revision)
331
except errors.NoSuchId:
334
# Drop non-head parents
336
graph = repo.get_graph()
337
heads = graph.heads(parent_ids)
339
for parent_id in parent_ids:
340
if parent_id in heads and parent_id not in selected_ids:
341
selected_ids.append(parent_id)
342
parent_keys = [(root_id, parent_id) for parent_id in selected_ids]
102
self.count_weaves = 0
103
self.copied_file_ids = set()
105
self.pb = bzrlib.ui.ui_factory.progress_bar()
108
self.from_branch.lock_read()
110
self._fetch_revisions(last_revision)
112
self.from_branch.unlock()
115
def _fetch_revisions(self, last_revision):
116
self.last_revision = self._find_last_revision(last_revision)
117
mutter('fetch up to rev {%s}', self.last_revision)
118
if (self.last_revision is not None and
119
self.to_branch.has_revision(self.last_revision)):
122
revs_to_fetch = self._compare_ancestries()
124
raise InstallFailed([self.last_revision])
125
self._copy_revisions(revs_to_fetch)
126
self.new_ancestry = revs_to_fetch
128
def _find_last_revision(self, last_revision):
129
"""Find the limiting source revision.
131
Every ancestor of that revision will be merged across.
133
Returns the revision_id, or returns None if there's no history
134
in the source branch."""
137
self.pb.update('get source history')
138
from_history = self.from_branch.revision_history()
139
self.pb.update('get destination history')
141
return from_history[-1]
143
return None # no history in the source branch
146
def _compare_ancestries(self):
147
"""Get a list of revisions that must be copied.
149
That is, every revision that's in the ancestry of the source
150
branch and not in the destination branch."""
151
self.pb.update('get source ancestry')
152
self.from_ancestry = self.from_branch.get_ancestry(self.last_revision)
154
dest_last_rev = self.to_branch.last_revision()
155
self.pb.update('get destination ancestry')
157
dest_ancestry = self.to_branch.get_ancestry(dest_last_rev)
160
ss = set(dest_ancestry)
162
for rev_id in self.from_ancestry:
164
to_fetch.append(rev_id)
165
mutter('need to get revision {%s}', rev_id)
166
mutter('need to get %d revisions in total', len(to_fetch))
167
self.count_total = len(to_fetch)
170
def _copy_revisions(self, revs_to_fetch):
172
for rev_id in revs_to_fetch:
176
if self.to_branch.has_revision(rev_id):
178
self.pb.update('fetch revision', i, self.count_total)
179
self._copy_one_revision(rev_id)
180
self.count_copied += 1
183
def _copy_one_revision(self, rev_id):
184
"""Copy revision and everything referenced by it."""
185
mutter('copying revision {%s}', rev_id)
186
rev_xml = self.from_branch.get_revision_xml(rev_id)
187
inv_xml = self.from_branch.get_inventory_xml(rev_id)
188
rev = serializer_v5.read_revision_from_string(rev_xml)
189
inv = serializer_v5.read_inventory_from_string(inv_xml)
190
assert rev.revision_id == rev_id
191
assert rev.inventory_sha1 == sha_string(inv_xml)
192
mutter(' commiter %s, %d parents',
195
self._copy_new_texts(rev_id, inv)
196
parents = rev.parent_ids
197
new_parents = copy(parents)
198
for parent in parents:
199
if not self.to_branch.has_revision(parent):
200
new_parents.pop(new_parents.index(parent))
201
self._copy_inventory(rev_id, inv_xml, new_parents)
202
self.to_branch.revision_store.add(StringIO(rev_xml), rev_id)
203
mutter('copied revision %s', rev_id)
205
def _copy_inventory(self, rev_id, inv_xml, parent_ids):
206
self.to_control.add_text('inventory', rev_id,
207
split_lines(inv_xml), parent_ids,
208
self.to_branch.get_transaction())
210
def _copy_new_texts(self, rev_id, inv):
211
"""Copy any new texts occuring in this revision."""
212
# TODO: Rather than writing out weaves every time, hold them
213
# in memory until everything's done? But this way is nicer
214
# if it's interrupted.
215
for path, ie in inv.iter_entries():
216
self._copy_one_weave(rev_id, ie.file_id, ie.revision)
218
def _copy_one_weave(self, rev_id, file_id, text_revision):
219
"""Copy one file weave, esuring the result contains text_revision."""
220
to_weave = self.to_weaves.get_weave_or_empty(file_id,
221
self.to_branch.get_transaction())
222
if text_revision in to_weave:
224
from_weave = self.from_weaves.get_weave(file_id,
225
self.from_branch.get_transaction())
226
if text_revision not in from_weave:
227
raise MissingText(self.from_branch, text_revision, file_id)
228
mutter('copy file {%s} modified in {%s}', file_id, rev_id)
230
to_weave.join(from_weave)
231
except errors.WeaveParentMismatch:
232
to_weave.reweave(from_weave)
233
self.to_weaves.put_weave(file_id, to_weave,
234
self.to_branch.get_transaction())
235
self.count_weaves += 1
236
self.copied_file_ids.add(file_id)
237
mutter('copied file {%s}', file_id)
added
removed
bzrlib/fetch.py
