20
34
that has merged into it. As the first step of a merge, pull, or
21
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.
27
from ..lazy_import import lazy_import
28
lazy_import(globals(), """
32
from breezy.bzr import (
41
from ..i18n import gettext
42
from ..revision import NULL_REVISION
43
from ..trace import mutter
46
class RepoFetcher(object):
47
"""Pull revisions and texts from one repository to another.
49
This should not be used directly, it's essential a object to encapsulate
50
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
53
def __init__(self, to_repository, from_repository, last_revision=None,
54
find_ghosts=True, fetch_spec=None):
55
"""Create a repo fetcher.
57
:param last_revision: If set, try to limit to the data this revision
59
:param fetch_spec: A SearchResult specifying which revisions to fetch.
60
If set, this overrides last_revision.
61
:param find_ghosts: If True search the entire history for ghosts.
63
# repository.fetch has the responsibility for short-circuiting
64
# attempts to copy between a repository and itself.
65
self.to_repository = to_repository
66
self.from_repository = from_repository
67
self.sink = to_repository._get_sink()
68
# must not mutate self._last_revision as its potentially a shared instance
69
self._last_revision = last_revision
70
self._fetch_spec = fetch_spec
71
self.find_ghosts = find_ghosts
72
with self.from_repository.lock_read():
73
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
74
str(self.from_repository), str(self.from_repository._format),
75
str(self.to_repository), str(self.to_repository._format))
79
"""Primary worker function.
81
This initialises all the needed variables, and then fetches the
82
requested revisions, finally clearing the progress bar.
84
# Roughly this is what we're aiming for fetch to become:
86
# missing = self.sink.insert_stream(self.source.get_stream(search))
88
# 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
90
101
self.count_total = 0
102
self.count_weaves = 0
103
self.copied_file_ids = set()
91
104
self.file_ids_names = {}
92
with ui.ui_factory.nested_progress_bar() as pb:
93
pb.show_pct = pb.show_count = False
94
pb.update(gettext("Finding revisions"), 0, 2)
95
search_result = self._revids_to_fetch()
96
mutter('fetching: %s', str(search_result))
97
if search_result.is_empty():
99
pb.update(gettext("Fetching revisions"), 1, 2)
100
self._fetch_everything_for_search(search_result)
102
def _fetch_everything_for_search(self, search):
103
"""Fetch all data for the given set of revisions."""
104
# The first phase is "file". We pass the progress bar for it directly
105
# into item_keys_introduced_by, which has more information about how
106
# that phase is progressing than we do. Progress updates for the other
107
# phases are taken care of in this function.
108
# XXX: there should be a clear owner of the progress reporting. Perhaps
109
# item_keys_introduced_by should have a richer API than it does at the
110
# moment, so that it can feed the progress information back to this
112
if (self.from_repository._format.rich_root_data and
113
not self.to_repository._format.rich_root_data):
114
raise errors.IncompatibleRepositories(
115
self.from_repository, self.to_repository,
116
"different rich-root support")
117
with ui.ui_factory.nested_progress_bar() as pb:
118
pb.update("Get stream source")
119
source = self.from_repository._get_source(
120
self.to_repository._format)
121
stream = source.get_stream(search)
122
from_format = self.from_repository._format
123
pb.update("Inserting stream")
124
resume_tokens, missing_keys = self.sink.insert_stream(
125
stream, from_format, [])
127
pb.update("Missing keys")
128
stream = source.get_stream_for_missing_keys(missing_keys)
129
pb.update("Inserting missing keys")
130
resume_tokens, missing_keys = self.sink.insert_stream(
131
stream, from_format, resume_tokens)
133
raise AssertionError(
134
"second push failed to complete a fetch %r." % (
137
raise AssertionError(
138
"second push failed to commit the fetch %r." % (
140
pb.update("Finishing stream")
143
def _revids_to_fetch(self):
144
"""Determines the exact revisions needed from self.from_repository to
145
install self._last_revision in self.to_repository.
147
:returns: A SearchResult of some sort. (Possibly a
148
PendingAncestryResult, EmptySearchResult, etc.)
150
if self._fetch_spec is not None:
151
# The fetch spec is already a concrete search result.
152
return self._fetch_spec
153
elif self._last_revision == NULL_REVISION:
154
# fetch_spec is None + last_revision is null => empty fetch.
155
# explicit limit of no revisions needed
156
return vf_search.EmptySearchResult()
157
elif self._last_revision is not None:
158
return vf_search.NotInOtherForRevs(self.to_repository,
159
self.from_repository, [
160
self._last_revision],
161
find_ghosts=self.find_ghosts).execute()
162
else: # self._last_revision is None:
163
return vf_search.EverythingNotInOther(self.to_repository,
164
self.from_repository,
165
find_ghosts=self.find_ghosts).execute()
168
class Inter1and2Helper(object):
169
"""Helper for operations that convert data from model 1 and 2
171
This is for use by fetchers and converters.
174
# This is a class variable so that the test suite can override it.
175
known_graph_threshold = 100
177
def __init__(self, source):
180
:param source: The repository data comes from
184
def iter_rev_trees(self, revs):
185
"""Iterate through RevisionTrees efficiently.
187
Additionally, the inventory's revision_id is set if unset.
189
Trees are retrieved in batches of 100, and then yielded in the order
192
:param revs: A list of revision ids
194
# In case that revs is not a list.
197
for tree in self.source.revision_trees(revs[:100]):
198
if tree.root_inventory.revision_id is None:
199
tree.root_inventory.revision_id = tree.get_revision_id()
203
def _find_root_ids(self, revs, parent_map, graph):
205
for tree in self.iter_rev_trees(revs):
206
root_id = tree.path2id('')
207
revision_id = tree.get_file_revision(u'')
208
revision_root[revision_id] = root_id
209
# Find out which parents we don't already know root ids for
210
parents = set(parent_map.values())
211
parents.difference_update(revision_root)
212
parents.discard(NULL_REVISION)
213
# Limit to revisions present in the versionedfile
214
parents = graph.get_parent_map(parents)
215
for tree in self.iter_rev_trees(parents):
216
root_id = tree.path2id('')
217
revision_root[tree.get_revision_id()] = root_id
220
def generate_root_texts(self, revs):
221
"""Generate VersionedFiles for all root ids.
223
:param revs: the revisions to include
225
graph = self.source.get_graph()
226
parent_map = graph.get_parent_map(revs)
227
rev_order = tsort.topo_sort(parent_map)
228
rev_id_to_root_id = self._find_root_ids(revs, parent_map, graph)
229
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
231
# Guaranteed stable, this groups all the file id operations together
232
# retaining topological order within the revisions of a file id.
233
# File id splits and joins would invalidate this, but they don't exist
234
# yet, and are unlikely to in non-rich-root environments anyway.
235
root_id_order.sort(key=operator.itemgetter(0))
236
# Create a record stream containing the roots to create.
237
if len(revs) > self.known_graph_threshold:
238
graph = self.source.get_known_graph_ancestry(revs)
239
new_roots_stream = _new_root_data_stream(
240
root_id_order, rev_id_to_root_id, parent_map, self.source, graph)
241
return [('texts', new_roots_stream)]
244
def _new_root_data_stream(
245
root_keys_to_create, rev_id_to_root_id_map, parent_map, repo, graph=None):
246
"""Generate a texts substream of synthesised root entries.
248
Used in fetches that do rich-root upgrades.
250
:param root_keys_to_create: iterable of (root_id, rev_id) pairs describing
251
the root entries to create.
252
:param rev_id_to_root_id_map: dict of known rev_id -> root_id mappings for
253
calculating the parents. If a parent rev_id is not found here then it
254
will be recalculated.
255
:param parent_map: a parent map for all the revisions in
257
:param graph: a graph to use instead of repo.get_graph().
259
for root_key in root_keys_to_create:
260
root_id, rev_id = root_key
261
parent_keys = _parent_keys_for_root_version(
262
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph)
263
yield versionedfile.ChunkedContentFactory(
264
root_key, parent_keys, None, [])
267
def _parent_keys_for_root_version(
268
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph=None):
269
"""Get the parent keys for a given root id.
271
A helper function for _new_root_data_stream.
273
# Include direct parents of the revision, but only if they used the same
274
# root_id and are heads.
275
rev_parents = parent_map[rev_id]
277
for parent_id in rev_parents:
278
if parent_id == NULL_REVISION:
280
if parent_id not in rev_id_to_root_id_map:
281
# We probably didn't read this revision, go spend the extra effort
284
tree = repo.revision_tree(parent_id)
285
except errors.NoSuchRevision:
286
# Ghost, fill out rev_id_to_root_id in case we encounter this
288
# But set parent_root_id to None since we don't really know
289
parent_root_id = None
291
parent_root_id = tree.path2id('')
292
rev_id_to_root_id_map[parent_id] = None
294
# rev_id_to_root_id_map[parent_id] = parent_root_id
295
# memory consumption maybe?
297
parent_root_id = rev_id_to_root_id_map[parent_id]
298
if root_id == parent_root_id:
299
# With stacking we _might_ want to refer to a non-local revision,
300
# but this code path only applies when we have the full content
301
# available, so ghosts really are ghosts, not just the edge of
303
parent_ids.append(parent_id)
305
# root_id may be in the parent anyway.
307
tree = repo.revision_tree(parent_id)
308
except errors.NoSuchRevision:
309
# ghost, can't refer to it.
314
tree.get_file_revision(
315
tree.id2path(root_id, recurse='none')))
316
except errors.NoSuchId:
319
# Drop non-head parents
321
graph = repo.get_graph()
322
heads = graph.heads(parent_ids)
324
for parent_id in parent_ids:
325
if parent_id in heads and parent_id not in selected_ids:
326
selected_ids.append(parent_id)
327
parent_keys = [(root_id, parent_id) for parent_id in selected_ids]
331
class TargetRepoKinds(object):
332
"""An enum-like set of constants.
334
They are the possible values of FetchSpecFactory.target_repo_kinds.
337
PREEXISTING = 'preexisting'
342
class FetchSpecFactory(object):
343
"""A helper for building the best fetch spec for a sprout call.
345
Factors that go into determining the sort of fetch to perform:
346
* did the caller specify any revision IDs?
347
* did the caller specify a source branch (need to fetch its
348
heads_to_fetch(), usually the tip + tags)
349
* is there an existing target repo (don't need to refetch revs it
351
* target is stacked? (similar to pre-existing target repo: even if
352
the target itself is new don't want to refetch existing revs)
354
:ivar source_branch: the source branch if one specified, else None.
355
:ivar source_branch_stop_revision_id: fetch up to this revision of
356
source_branch, rather than its tip.
357
:ivar source_repo: the source repository if one found, else None.
358
:ivar target_repo: the target repository acquired by sprout.
359
:ivar target_repo_kind: one of the TargetRepoKinds constants.
363
self._explicit_rev_ids = set()
364
self.source_branch = None
365
self.source_branch_stop_revision_id = None
366
self.source_repo = None
367
self.target_repo = None
368
self.target_repo_kind = None
371
def add_revision_ids(self, revision_ids):
372
"""Add revision_ids to the set of revision_ids to be fetched."""
373
self._explicit_rev_ids.update(revision_ids)
375
def make_fetch_spec(self):
376
"""Build a SearchResult or PendingAncestryResult or etc."""
377
if self.target_repo_kind is None or self.source_repo is None:
378
raise AssertionError(
379
'Incomplete FetchSpecFactory: %r' % (self.__dict__,))
380
if len(self._explicit_rev_ids) == 0 and self.source_branch is None:
381
if self.limit is not None:
382
raise NotImplementedError(
383
"limit is only supported with a source branch set")
384
# Caller hasn't specified any revisions or source branch
385
if self.target_repo_kind == TargetRepoKinds.EMPTY:
386
return vf_search.EverythingResult(self.source_repo)
388
# We want everything not already in the target (or target's
390
return vf_search.EverythingNotInOther(
391
self.target_repo, self.source_repo).execute()
392
heads_to_fetch = set(self._explicit_rev_ids)
393
if self.source_branch is not None:
394
must_fetch, if_present_fetch = self.source_branch.heads_to_fetch()
395
if self.source_branch_stop_revision_id is not None:
396
# Replace the tip rev from must_fetch with the stop revision
397
# XXX: this might be wrong if the tip rev is also in the
398
# must_fetch set for other reasons (e.g. it's the tip of
399
# multiple loom threads?), but then it's pretty unclear what it
400
# should mean to specify a stop_revision in that case anyway.
401
must_fetch.discard(self.source_branch.last_revision())
402
must_fetch.add(self.source_branch_stop_revision_id)
403
heads_to_fetch.update(must_fetch)
405
if_present_fetch = set()
406
if self.target_repo_kind == TargetRepoKinds.EMPTY:
407
# PendingAncestryResult does not raise errors if a requested head
408
# is absent. Ideally it would support the
409
# required_ids/if_present_ids distinction, but in practice
410
# heads_to_fetch will almost certainly be present so this doesn't
412
all_heads = heads_to_fetch.union(if_present_fetch)
413
ret = vf_search.PendingAncestryResult(all_heads, self.source_repo)
414
if self.limit is not None:
415
graph = self.source_repo.get_graph()
416
topo_order = list(graph.iter_topo_order(ret.get_keys()))
417
result_set = topo_order[:self.limit]
418
ret = self.source_repo.revision_ids_to_search_result(
422
return vf_search.NotInOtherForRevs(self.target_repo, self.source_repo,
423
required_ids=heads_to_fetch, if_present_ids=if_present_fetch,
424
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('copy 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
