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
44
class RepoFetcher(object):
45
"""Pull revisions and texts from one repository to another.
47
This should not be used directly, it's essential a object to encapsulate
48
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
51
def __init__(self, to_repository, from_repository, last_revision=None,
52
find_ghosts=True, fetch_spec=None):
53
"""Create a repo fetcher.
55
:param last_revision: If set, try to limit to the data this revision
57
:param fetch_spec: A SearchResult specifying which revisions to fetch.
58
If set, this overrides last_revision.
59
:param find_ghosts: If True search the entire history for ghosts.
61
# repository.fetch has the responsibility for short-circuiting
62
# attempts to copy between a repository and itself.
63
self.to_repository = to_repository
64
self.from_repository = from_repository
65
self.sink = to_repository._get_sink()
66
# must not mutate self._last_revision as its potentially a shared instance
67
self._last_revision = last_revision
68
self._fetch_spec = fetch_spec
69
self.find_ghosts = find_ghosts
70
self.from_repository.lock_read()
71
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
72
self.from_repository, self.from_repository._format,
73
self.to_repository, self.to_repository._format)
77
self.from_repository.unlock()
80
"""Primary worker function.
82
This initialises all the needed variables, and then fetches the
83
requested revisions, finally clearing the progress bar.
85
# Roughly this is what we're aiming for fetch to become:
87
# missing = self.sink.insert_stream(self.source.get_stream(search))
89
# 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
91
101
self.count_total = 0
102
self.count_weaves = 0
103
self.copied_file_ids = set()
92
104
self.file_ids_names = {}
93
pb = ui.ui_factory.nested_progress_bar()
94
pb.show_pct = pb.show_count = False
96
pb.update("Finding revisions", 0, 2)
97
search_result = self._revids_to_fetch()
98
mutter('fetching: %s', search_result)
99
if search_result.is_empty():
101
pb.update("Fetching revisions", 1, 2)
102
self._fetch_everything_for_search(search_result)
106
def _fetch_everything_for_search(self, search):
107
"""Fetch all data for the given set of revisions."""
108
# The first phase is "file". We pass the progress bar for it directly
109
# into item_keys_introduced_by, which has more information about how
110
# that phase is progressing than we do. Progress updates for the other
111
# phases are taken care of in this function.
112
# XXX: there should be a clear owner of the progress reporting. Perhaps
113
# item_keys_introduced_by should have a richer API than it does at the
114
# moment, so that it can feed the progress information back to this
116
if (self.from_repository._format.rich_root_data and
117
not self.to_repository._format.rich_root_data):
118
raise errors.IncompatibleRepositories(
119
self.from_repository, self.to_repository,
120
"different rich-root support")
121
pb = ui.ui_factory.nested_progress_bar()
123
pb.update("Get stream source")
124
source = self.from_repository._get_source(
125
self.to_repository._format)
126
stream = source.get_stream(search)
127
from_format = self.from_repository._format
128
pb.update("Inserting stream")
129
resume_tokens, missing_keys = self.sink.insert_stream(
130
stream, from_format, [])
132
pb.update("Missing keys")
133
stream = source.get_stream_for_missing_keys(missing_keys)
134
pb.update("Inserting missing keys")
135
resume_tokens, missing_keys = self.sink.insert_stream(
136
stream, from_format, resume_tokens)
138
raise AssertionError(
139
"second push failed to complete a fetch %r." % (
142
raise AssertionError(
143
"second push failed to commit the fetch %r." % (
145
pb.update("Finishing stream")
150
def _revids_to_fetch(self):
151
"""Determines the exact revisions needed from self.from_repository to
152
install self._last_revision in self.to_repository.
154
:returns: A SearchResult of some sort. (Possibly a
155
PendingAncestryResult, EmptySearchResult, etc.)
157
if self._fetch_spec is not None:
158
# The fetch spec is already a concrete search result.
159
return self._fetch_spec
160
elif self._last_revision == NULL_REVISION:
161
# fetch_spec is None + last_revision is null => empty fetch.
162
# explicit limit of no revisions needed
163
return _mod_graph.EmptySearchResult()
164
elif self._last_revision is not None:
165
return _mod_graph.NotInOtherForRevs(self.to_repository,
166
self.from_repository, [self._last_revision],
167
find_ghosts=self.find_ghosts).execute()
168
else: # self._last_revision is None:
169
return _mod_graph.EverythingNotInOther(self.to_repository,
170
self.from_repository,
171
find_ghosts=self.find_ghosts).execute()
174
class Inter1and2Helper(object):
175
"""Helper for operations that convert data from model 1 and 2
177
This is for use by fetchers and converters.
180
# This is a class variable so that the test suite can override it.
181
known_graph_threshold = 100
183
def __init__(self, source):
186
:param source: The repository data comes from
190
def iter_rev_trees(self, revs):
191
"""Iterate through RevisionTrees efficiently.
193
Additionally, the inventory's revision_id is set if unset.
195
Trees are retrieved in batches of 100, and then yielded in the order
198
:param revs: A list of revision ids
200
# In case that revs is not a list.
203
for tree in self.source.revision_trees(revs[:100]):
204
if tree.inventory.revision_id is None:
205
tree.inventory.revision_id = tree.get_revision_id()
209
def _find_root_ids(self, revs, parent_map, graph):
211
for tree in self.iter_rev_trees(revs):
212
revision_id = tree.inventory.root.revision
213
root_id = tree.get_root_id()
214
revision_root[revision_id] = root_id
215
# Find out which parents we don't already know root ids for
217
for revision_parents in parent_map.itervalues():
218
parents.update(revision_parents)
219
parents.difference_update(revision_root.keys() + [NULL_REVISION])
220
# Limit to revisions present in the versionedfile
221
parents = graph.get_parent_map(parents).keys()
222
for tree in self.iter_rev_trees(parents):
223
root_id = tree.get_root_id()
224
revision_root[tree.get_revision_id()] = root_id
227
def generate_root_texts(self, revs):
228
"""Generate VersionedFiles for all root ids.
230
:param revs: the revisions to include
232
graph = self.source.get_graph()
233
parent_map = graph.get_parent_map(revs)
234
rev_order = tsort.topo_sort(parent_map)
235
rev_id_to_root_id = self._find_root_ids(revs, parent_map, graph)
236
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
238
# Guaranteed stable, this groups all the file id operations together
239
# retaining topological order within the revisions of a file id.
240
# File id splits and joins would invalidate this, but they don't exist
241
# yet, and are unlikely to in non-rich-root environments anyway.
242
root_id_order.sort(key=operator.itemgetter(0))
243
# Create a record stream containing the roots to create.
244
if len(revs) > self.known_graph_threshold:
245
graph = self.source.get_known_graph_ancestry(revs)
246
new_roots_stream = _new_root_data_stream(
247
root_id_order, rev_id_to_root_id, parent_map, self.source, graph)
248
return [('texts', new_roots_stream)]
251
def _new_root_data_stream(
252
root_keys_to_create, rev_id_to_root_id_map, parent_map, repo, graph=None):
253
"""Generate a texts substream of synthesised root entries.
255
Used in fetches that do rich-root upgrades.
257
:param root_keys_to_create: iterable of (root_id, rev_id) pairs describing
258
the root entries to create.
259
:param rev_id_to_root_id_map: dict of known rev_id -> root_id mappings for
260
calculating the parents. If a parent rev_id is not found here then it
261
will be recalculated.
262
:param parent_map: a parent map for all the revisions in
264
:param graph: a graph to use instead of repo.get_graph().
266
for root_key in root_keys_to_create:
267
root_id, rev_id = root_key
268
parent_keys = _parent_keys_for_root_version(
269
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph)
270
yield versionedfile.FulltextContentFactory(
271
root_key, parent_keys, None, '')
274
def _parent_keys_for_root_version(
275
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph=None):
276
"""Get the parent keys for a given root id.
278
A helper function for _new_root_data_stream.
280
# Include direct parents of the revision, but only if they used the same
281
# root_id and are heads.
282
rev_parents = parent_map[rev_id]
284
for parent_id in rev_parents:
285
if parent_id == NULL_REVISION:
287
if parent_id not in rev_id_to_root_id_map:
288
# We probably didn't read this revision, go spend the extra effort
291
tree = repo.revision_tree(parent_id)
292
except errors.NoSuchRevision:
293
# Ghost, fill out rev_id_to_root_id in case we encounter this
295
# But set parent_root_id to None since we don't really know
296
parent_root_id = None
298
parent_root_id = tree.get_root_id()
299
rev_id_to_root_id_map[parent_id] = None
301
# rev_id_to_root_id_map[parent_id] = parent_root_id
302
# memory consumption maybe?
304
parent_root_id = rev_id_to_root_id_map[parent_id]
305
if root_id == parent_root_id:
306
# With stacking we _might_ want to refer to a non-local revision,
307
# but this code path only applies when we have the full content
308
# available, so ghosts really are ghosts, not just the edge of
310
parent_ids.append(parent_id)
312
# root_id may be in the parent anyway.
314
tree = repo.revision_tree(parent_id)
315
except errors.NoSuchRevision:
316
# ghost, can't refer to it.
320
parent_ids.append(tree.get_file_revision(root_id))
321
except errors.NoSuchId:
324
# Drop non-head parents
326
graph = repo.get_graph()
327
heads = graph.heads(parent_ids)
329
for parent_id in parent_ids:
330
if parent_id in heads and parent_id not in selected_ids:
331
selected_ids.append(parent_id)
332
parent_keys = [(root_id, parent_id) for parent_id in selected_ids]
336
class TargetRepoKinds(object):
337
"""An enum-like set of constants.
339
They are the possible values of FetchSpecFactory.target_repo_kinds.
342
PREEXISTING = 'preexisting'
347
class FetchSpecFactory(object):
348
"""A helper for building the best fetch spec for a sprout call.
350
Factors that go into determining the sort of fetch to perform:
351
* did the caller specify any revision IDs?
352
* did the caller specify a source branch (need to fetch its
353
heads_to_fetch(), usually the tip + tags)
354
* is there an existing target repo (don't need to refetch revs it
356
* target is stacked? (similar to pre-existing target repo: even if
357
the target itself is new don't want to refetch existing revs)
359
:ivar source_branch: the source branch if one specified, else None.
360
:ivar source_branch_stop_revision_id: fetch up to this revision of
361
source_branch, rather than its tip.
362
:ivar source_repo: the source repository if one found, else None.
363
:ivar target_repo: the target repository acquired by sprout.
364
:ivar target_repo_kind: one of the TargetRepoKinds constants.
368
self._explicit_rev_ids = set()
369
self.source_branch = None
370
self.source_branch_stop_revision_id = None
371
self.source_repo = None
372
self.target_repo = None
373
self.target_repo_kind = None
376
def add_revision_ids(self, revision_ids):
377
"""Add revision_ids to the set of revision_ids to be fetched."""
378
self._explicit_rev_ids.update(revision_ids)
380
def make_fetch_spec(self):
381
"""Build a SearchResult or PendingAncestryResult or etc."""
382
if self.target_repo_kind is None or self.source_repo is None:
383
raise AssertionError(
384
'Incomplete FetchSpecFactory: %r' % (self.__dict__,))
385
if len(self._explicit_rev_ids) == 0 and self.source_branch is None:
386
if self.limit is not None:
387
raise NotImplementedError(
388
"limit is only supported with a source branch set")
389
# Caller hasn't specified any revisions or source branch
390
if self.target_repo_kind == TargetRepoKinds.EMPTY:
391
return _mod_graph.EverythingResult(self.source_repo)
393
# We want everything not already in the target (or target's
395
return _mod_graph.EverythingNotInOther(
396
self.target_repo, self.source_repo).execute()
397
heads_to_fetch = set(self._explicit_rev_ids)
398
if self.source_branch is not None:
399
must_fetch, if_present_fetch = self.source_branch.heads_to_fetch()
400
if self.source_branch_stop_revision_id is not None:
401
# Replace the tip rev from must_fetch with the stop revision
402
# XXX: this might be wrong if the tip rev is also in the
403
# must_fetch set for other reasons (e.g. it's the tip of
404
# multiple loom threads?), but then it's pretty unclear what it
405
# should mean to specify a stop_revision in that case anyway.
406
must_fetch.discard(self.source_branch.last_revision())
407
must_fetch.add(self.source_branch_stop_revision_id)
408
heads_to_fetch.update(must_fetch)
410
if_present_fetch = set()
411
if self.target_repo_kind == TargetRepoKinds.EMPTY:
412
# PendingAncestryResult does not raise errors if a requested head
413
# is absent. Ideally it would support the
414
# required_ids/if_present_ids distinction, but in practice
415
# heads_to_fetch will almost certainly be present so this doesn't
417
all_heads = heads_to_fetch.union(if_present_fetch)
418
ret = _mod_graph.PendingAncestryResult(all_heads, self.source_repo)
419
if self.limit is not None:
420
graph = self.source_repo.get_graph()
421
topo_order = list(graph.iter_topo_order(ret.get_keys()))
422
result_set = topo_order[:self.limit]
423
ret = self.source_repo.revision_ids_to_search_result(result_set)
426
return _mod_graph.NotInOtherForRevs(self.target_repo, self.source_repo,
427
required_ids=heads_to_fetch, if_present_ids=if_present_fetch,
428
limit=self.limit).execute()
106
self.pb = bzrlib.ui.ui_factory.progress_bar()
109
self.from_branch.lock_read()
111
self._fetch_revisions(last_revision)
113
self.from_branch.unlock()
116
def _fetch_revisions(self, last_revision):
117
self.last_revision = self._find_last_revision(last_revision)
118
mutter('fetch up to rev {%s}', self.last_revision)
119
if (self.last_revision is not None and
120
self.to_branch.has_revision(self.last_revision)):
123
revs_to_fetch = self._compare_ancestries()
125
raise InstallFailed([self.last_revision])
126
self._copy_revisions(revs_to_fetch)
127
self.new_ancestry = revs_to_fetch
129
def _find_last_revision(self, last_revision):
130
"""Find the limiting source revision.
132
Every ancestor of that revision will be merged across.
134
Returns the revision_id, or returns None if there's no history
135
in the source branch."""
138
self.pb.update('get source history')
139
from_history = self.from_branch.revision_history()
140
self.pb.update('get destination history')
142
return from_history[-1]
144
return None # no history in the source branch
147
def _compare_ancestries(self):
148
"""Get a list of revisions that must be copied.
150
That is, every revision that's in the ancestry of the source
151
branch and not in the destination branch."""
152
self.pb.update('get source ancestry')
153
self.from_ancestry = self.from_branch.get_ancestry(self.last_revision)
155
dest_last_rev = self.to_branch.last_revision()
156
self.pb.update('get destination ancestry')
158
dest_ancestry = self.to_branch.get_ancestry(dest_last_rev)
161
ss = set(dest_ancestry)
163
for rev_id in self.from_ancestry:
165
to_fetch.append(rev_id)
166
mutter('need to get revision {%s}', rev_id)
167
mutter('need to get %d revisions in total', len(to_fetch))
168
self.count_total = len(to_fetch)
171
def _copy_revisions(self, revs_to_fetch):
173
for rev_id in revs_to_fetch:
177
if self.to_branch.has_revision(rev_id):
179
self.pb.update('fetch revision', i, self.count_total)
180
self._copy_one_revision(rev_id)
181
self.count_copied += 1
184
def _copy_one_revision(self, rev_id):
185
"""Copy revision and everything referenced by it."""
186
mutter('copying revision {%s}', rev_id)
187
rev_xml = self.from_branch.get_revision_xml(rev_id)
188
inv_xml = self.from_branch.get_inventory_xml(rev_id)
189
rev = serializer_v5.read_revision_from_string(rev_xml)
190
inv = serializer_v5.read_inventory_from_string(inv_xml)
191
assert rev.revision_id == rev_id
192
assert rev.inventory_sha1 == sha_string(inv_xml)
193
mutter(' commiter %s, %d parents',
196
self._copy_new_texts(rev_id, inv)
197
parents = rev.parent_ids
198
new_parents = copy(parents)
199
for parent in parents:
200
if not self.to_branch.has_revision(parent):
201
new_parents.pop(new_parents.index(parent))
202
self._copy_inventory(rev_id, inv_xml, new_parents)
203
self.to_branch.revision_store.add(StringIO(rev_xml), rev_id)
204
mutter('copied revision %s', rev_id)
206
def _copy_inventory(self, rev_id, inv_xml, parent_ids):
207
self.to_control.add_text('inventory', rev_id,
208
split_lines(inv_xml), parent_ids,
209
self.to_branch.get_transaction())
211
def _copy_new_texts(self, rev_id, inv):
212
"""Copy any new texts occuring in this revision."""
213
# TODO: Rather than writing out weaves every time, hold them
214
# in memory until everything's done? But this way is nicer
215
# if it's interrupted.
216
for path, ie in inv.iter_entries():
217
self._copy_one_weave(rev_id, ie.file_id, ie.revision)
219
def _copy_one_weave(self, rev_id, file_id, text_revision):
220
"""Copy one file weave, esuring the result contains text_revision."""
221
# check if the revision is already there
222
if file_id in self.file_ids_names.keys( ) and \
223
text_revision in self.file_ids_names[file_id]:
225
to_weave = self.to_weaves.get_weave_or_empty(file_id,
226
self.to_branch.get_transaction())
227
if not file_id in self.file_ids_names.keys( ):
228
self.file_ids_names[file_id] = to_weave.names( )
229
if text_revision in to_weave:
231
from_weave = self.from_weaves.get_weave(file_id,
232
self.from_branch.get_transaction())
233
if text_revision not in from_weave:
234
raise MissingText(self.from_branch, text_revision, file_id)
235
mutter('copy file {%s} modified in {%s}', file_id, rev_id)
237
if to_weave.numversions() > 0:
238
# destination has contents, must merge
240
to_weave.join(from_weave)
241
except errors.WeaveParentMismatch:
242
to_weave.reweave(from_weave)
244
# destination is empty, just replace it
245
to_weave = from_weave.copy( )
246
self.to_weaves.put_weave(file_id, to_weave,
247
self.to_branch.get_transaction())
248
self.count_weaves += 1
249
self.copied_file_ids.add(file_id)
250
self.file_ids_names[file_id] = to_weave.names()
251
mutter('copied file {%s}', file_id)
added
removed
bzrlib/fetch.py
