31
20
that has merged into it. As the first step of a merge, pull, or
32
21
branch operation we copy history from the source into the destination
35
The copying is done in a slightly complicated order. We don't want to
36
add a revision to the store until everything it refers to is also
37
stored, so that if a revision is present we can totally recreate it.
38
However, we can't know what files are included in a revision until we
39
read its inventory. Therefore, we first pull the XML and hold it in
40
memory until we've updated all of the files referenced.
43
# TODO: Avoid repeatedly opening weaves so many times.
46
def greedy_fetch(to_branch, from_branch, revision, pb):
47
f = Fetcher(to_branch, from_branch, revision, pb)
48
return f.count_copied, f.failed_revisions
51
class Fetcher(object):
52
"""Pull history from one branch to another.
55
If set, pull only up to this revision_id.
57
def __init__(self, to_branch, from_branch, revision_limit=None, pb=None):
58
self.to_branch = to_branch
59
self.from_branch = from_branch
60
self.revision_limit = revision_limit
61
self.failed_revisions = []
64
self.pb = bzrlib.ui.ui_factory.progress_bar()
67
self._load_histories()
68
revs_to_fetch = self._compare_ancestries()
69
self._copy_revisions(revs_to_fetch)
70
# - get a list of revisions that need to be pulled in
71
# - for each one, pull in that revision file
72
# and get the inventory, and store the inventory with right
74
# - and get the ancestry, and store that with right parents too
75
# - and keep a note of all file ids and version seen
76
# - then go through all files; for each one get the weave,
77
# and add in all file versions
80
def _load_histories(self):
81
"""Load histories of both branches, up to the limit."""
82
self.from_history = self.from_branch.revision_history()
83
self.to_history = self.to_branch.revision_history()
84
if self.revision_limit:
85
assert isinstance(revision_limit, basestring)
87
rev_index = self.from_history.index(revision_limit)
90
if rev_index is not None:
91
self.from_history = self.from_history[:rev_index + 1]
93
self.from_history = [revision]
96
def _compare_ancestries(self):
97
"""Get a list of revisions that must be copied.
99
That is, every revision that's in the ancestry of the source
100
branch and not in the destination branch."""
101
if self.from_history:
102
self.from_ancestry = self.from_branch.get_ancestry(self.from_history[-1])
104
self.from_ancestry = []
106
self.to_history = self.to_branch.get_ancestry(self.to_history[-1])
109
ss = set(self.to_history)
111
for rev_id in self.from_ancestry:
113
to_fetch.append(rev_id)
114
mutter('need to get revision {%s}', rev_id)
115
mutter('need to get %d revisions in total', len(to_fetch))
120
def _copy_revisions(self, revs_to_fetch):
121
for rev_id in revs_to_fetch:
122
self._copy_one_revision(rev_id)
125
def _copy_one_revision(self, rev_id):
126
"""Copy revision and everything referenced by it."""
127
mutter('copying revision {%s}', rev_id)
128
rev_xml = self.from_branch.get_revision_xml(rev_id)
129
inv_xml = self.from_branch.get_inventory_xml(rev_id)
130
rev = serializer_v5.read_revision_from_string(rev_xml)
131
inv = serializer_v5.read_inventory_from_string(inv_xml)
132
assert rev.revision_id == rev_id
133
assert rev.inventory_sha1 == sha_string(inv_xml)
134
mutter(' commiter %s, %d parents',
137
self._copy_new_texts(rev_id, inv)
138
self.to_branch.weave_store.add_text(INVENTORY_FILEID, rev_id,
139
split_lines(inv_xml), rev.parents)
140
self.to_branch.revision_store.add(StringIO(rev_xml), rev_id)
143
def _copy_new_texts(self, rev_id, inv):
144
"""Copy any new texts occuring in this revision."""
145
# TODO: Rather than writing out weaves every time, hold them
146
# in memory until everything's done? But this way is nicer
147
# if it's interrupted.
148
for path, ie in inv.iter_entries():
149
if ie.kind != 'file':
151
if ie.text_version != rev_id:
153
mutter('%s {%s} is changed in this revision',
155
self._copy_one_text(rev_id, ie.file_id)
158
def _copy_one_text(self, rev_id, file_id):
159
"""Copy one file text."""
160
from_weave = self.from_branch.weave_store.get_weave(file_id)
161
from_idx = from_weave.lookup(rev_id)
162
from_parents = map(from_weave.idx_to_name, from_weave.parents(from_idx))
163
text_lines = from_weave.get(from_idx)
164
to_weave = self.to_branch.weave_store.get_weave_or_empty(file_id)
165
if rev_id in to_weave._name_map:
166
warning('version {%s} already present in weave of file {%s}',
169
to_parents = map(to_weave.lookup, from_parents)
170
to_weave.add(rev_id, to_parents, text_lines)
171
self.to_branch.weave_store.put_weave(file_id, to_weave)
174
def has_revision(branch, revision_id):
176
branch.get_revision_xml_file(revision_id)
178
except bzrlib.errors.NoSuchRevision:
182
def old_greedy_fetch(to_branch, from_branch, revision=None, pb=None):
183
"""Copy all history from one branch to another.
186
If set, copy only up to this point in the source branch.
188
@returns: number copied, missing ids
190
from_history = from_branch.revision_history()
191
required_revisions = set(from_history)
193
if revision is not None:
194
required_revisions.add(revision)
196
rev_index = from_history.index(revision)
199
if rev_index is not None:
200
from_history = from_history[:rev_index + 1]
202
from_history = [revision]
203
to_history = to_branch.revision_history()
205
for rev_id in from_history:
206
if not has_revision(to_branch, rev_id):
207
missing.append(rev_id)
209
# recurse down through the revision graph, looking for things that
212
while len(missing) > 0:
213
installed, failed = to_branch.install_revisions(from_branch,
214
revision_ids=missing,
217
required_failed = failed.intersection(required_revisions)
218
if len(required_failed) > 0:
219
raise bzrlib.errors.InstallFailed(required_failed)
220
for rev_id in failed:
221
note("Failed to install %s" % rev_id)
222
all_failed.update(failed)
224
for rev_id in missing:
226
revision = from_branch.get_revision(rev_id)
227
except bzrlib.errors.NoSuchRevision:
228
if revision in from_history:
232
for parent in [p.revision_id for p in revision.parents]:
233
if not has_revision(to_branch, parent):
234
new_missing.add(parent)
235
missing = new_missing
236
return count, all_failed
239
def old_install_revisions(branch, other, revision_ids, pb):
240
"""Copy revisions from other branch into branch.
242
This is a lower-level function used by a pull or a merge. It
243
incorporates some history from one branch into another, but
244
does not update the revision history or operate on the working
248
Sequence of revisions to copy.
251
Progress bar for copying.
254
if hasattr(other.revision_store, "prefetch"):
255
other.revision_store.prefetch(revision_ids)
256
if hasattr(other.inventory_store, "prefetch"):
257
other.inventory_store.prefetch(revision_ids)
260
pb = bzrlib.ui.ui_factory.progress_bar()
267
for i, rev_id in enumerate(revision_ids):
268
pb.update('fetching revision', i+1, len(revision_ids))
270
rev = other.get_revision(rev_id)
271
except bzrlib.errors.NoSuchRevision:
25
from __future__ import absolute_import
29
from bzrlib.lazy_import import lazy_import
30
lazy_import(globals(), """
41
from bzrlib.i18n import gettext
42
from bzrlib.revision import NULL_REVISION
43
from bzrlib.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().
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
self.from_repository.lock_read()
73
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
74
self.from_repository, self.from_repository._format,
75
self.to_repository, self.to_repository._format)
79
self.from_repository.unlock()
82
"""Primary worker function.
84
This initialises all the needed variables, and then fetches the
85
requested revisions, finally clearing the progress bar.
87
# Roughly this is what we're aiming for fetch to become:
89
# missing = self.sink.insert_stream(self.source.get_stream(search))
91
# missing = self.sink.insert_stream(self.source.get_items(missing))
94
self.file_ids_names = {}
95
pb = ui.ui_factory.nested_progress_bar()
96
pb.show_pct = pb.show_count = False
98
pb.update(gettext("Finding revisions"), 0, 2)
99
search_result = self._revids_to_fetch()
100
mutter('fetching: %s', search_result)
101
if search_result.is_empty():
103
pb.update(gettext("Fetching revisions"), 1, 2)
104
self._fetch_everything_for_search(search_result)
108
def _fetch_everything_for_search(self, search):
109
"""Fetch all data for the given set of revisions."""
110
# The first phase is "file". We pass the progress bar for it directly
111
# into item_keys_introduced_by, which has more information about how
112
# that phase is progressing than we do. Progress updates for the other
113
# phases are taken care of in this function.
114
# XXX: there should be a clear owner of the progress reporting. Perhaps
115
# item_keys_introduced_by should have a richer API than it does at the
116
# moment, so that it can feed the progress information back to this
118
if (self.from_repository._format.rich_root_data and
119
not self.to_repository._format.rich_root_data):
120
raise errors.IncompatibleRepositories(
121
self.from_repository, self.to_repository,
122
"different rich-root support")
123
pb = ui.ui_factory.nested_progress_bar()
125
pb.update("Get stream source")
126
source = self.from_repository._get_source(
127
self.to_repository._format)
128
stream = source.get_stream(search)
129
from_format = self.from_repository._format
130
pb.update("Inserting stream")
131
resume_tokens, missing_keys = self.sink.insert_stream(
132
stream, from_format, [])
134
pb.update("Missing keys")
135
stream = source.get_stream_for_missing_keys(missing_keys)
136
pb.update("Inserting missing keys")
137
resume_tokens, missing_keys = self.sink.insert_stream(
138
stream, from_format, resume_tokens)
140
raise AssertionError(
141
"second push failed to complete a fetch %r." % (
144
raise AssertionError(
145
"second push failed to commit the fetch %r." % (
147
pb.update("Finishing stream")
152
def _revids_to_fetch(self):
153
"""Determines the exact revisions needed from self.from_repository to
154
install self._last_revision in self.to_repository.
156
:returns: A SearchResult of some sort. (Possibly a
157
PendingAncestryResult, EmptySearchResult, etc.)
159
if self._fetch_spec is not None:
160
# The fetch spec is already a concrete search result.
161
return self._fetch_spec
162
elif self._last_revision == NULL_REVISION:
163
# fetch_spec is None + last_revision is null => empty fetch.
164
# explicit limit of no revisions needed
165
return vf_search.EmptySearchResult()
166
elif self._last_revision is not None:
167
return vf_search.NotInOtherForRevs(self.to_repository,
168
self.from_repository, [self._last_revision],
169
find_ghosts=self.find_ghosts).execute()
170
else: # self._last_revision is None:
171
return vf_search.EverythingNotInOther(self.to_repository,
172
self.from_repository,
173
find_ghosts=self.find_ghosts).execute()
176
class Inter1and2Helper(object):
177
"""Helper for operations that convert data from model 1 and 2
179
This is for use by fetchers and converters.
182
# This is a class variable so that the test suite can override it.
183
known_graph_threshold = 100
185
def __init__(self, source):
188
:param source: The repository data comes from
192
def iter_rev_trees(self, revs):
193
"""Iterate through RevisionTrees efficiently.
195
Additionally, the inventory's revision_id is set if unset.
197
Trees are retrieved in batches of 100, and then yielded in the order
200
:param revs: A list of revision ids
202
# In case that revs is not a list.
205
for tree in self.source.revision_trees(revs[:100]):
206
if tree.root_inventory.revision_id is None:
207
tree.root_inventory.revision_id = tree.get_revision_id()
211
def _find_root_ids(self, revs, parent_map, graph):
213
for tree in self.iter_rev_trees(revs):
214
root_id = tree.get_root_id()
215
revision_id = tree.get_file_revision(root_id, u"")
216
revision_root[revision_id] = root_id
217
# Find out which parents we don't already know root ids for
219
for revision_parents in parent_map.itervalues():
220
parents.update(revision_parents)
221
parents.difference_update(revision_root.keys() + [NULL_REVISION])
222
# Limit to revisions present in the versionedfile
223
parents = graph.get_parent_map(parents).keys()
224
for tree in self.iter_rev_trees(parents):
225
root_id = tree.get_root_id()
226
revision_root[tree.get_revision_id()] = root_id
229
def generate_root_texts(self, revs):
230
"""Generate VersionedFiles for all root ids.
232
:param revs: the revisions to include
234
graph = self.source.get_graph()
235
parent_map = graph.get_parent_map(revs)
236
rev_order = tsort.topo_sort(parent_map)
237
rev_id_to_root_id = self._find_root_ids(revs, parent_map, graph)
238
root_id_order = [(rev_id_to_root_id[rev_id], rev_id) for rev_id in
240
# Guaranteed stable, this groups all the file id operations together
241
# retaining topological order within the revisions of a file id.
242
# File id splits and joins would invalidate this, but they don't exist
243
# yet, and are unlikely to in non-rich-root environments anyway.
244
root_id_order.sort(key=operator.itemgetter(0))
245
# Create a record stream containing the roots to create.
246
if len(revs) > self.known_graph_threshold:
247
graph = self.source.get_known_graph_ancestry(revs)
248
new_roots_stream = _new_root_data_stream(
249
root_id_order, rev_id_to_root_id, parent_map, self.source, graph)
250
return [('texts', new_roots_stream)]
253
def _new_root_data_stream(
254
root_keys_to_create, rev_id_to_root_id_map, parent_map, repo, graph=None):
255
"""Generate a texts substream of synthesised root entries.
257
Used in fetches that do rich-root upgrades.
259
:param root_keys_to_create: iterable of (root_id, rev_id) pairs describing
260
the root entries to create.
261
:param rev_id_to_root_id_map: dict of known rev_id -> root_id mappings for
262
calculating the parents. If a parent rev_id is not found here then it
263
will be recalculated.
264
:param parent_map: a parent map for all the revisions in
266
:param graph: a graph to use instead of repo.get_graph().
268
for root_key in root_keys_to_create:
269
root_id, rev_id = root_key
270
parent_keys = _parent_keys_for_root_version(
271
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph)
272
yield versionedfile.FulltextContentFactory(
273
root_key, parent_keys, None, '')
276
def _parent_keys_for_root_version(
277
root_id, rev_id, rev_id_to_root_id_map, parent_map, repo, graph=None):
278
"""Get the parent keys for a given root id.
280
A helper function for _new_root_data_stream.
282
# Include direct parents of the revision, but only if they used the same
283
# root_id and are heads.
284
rev_parents = parent_map[rev_id]
286
for parent_id in rev_parents:
287
if parent_id == NULL_REVISION:
275
revisions.append(rev)
276
inv = other.get_inventory(rev_id)
277
for key, entry in inv.iter_entries():
278
if entry.text_id is None:
280
if entry.text_id not in branch.text_store:
281
needed_texts.add(entry.text_id)
285
count, cp_fail = branch.text_store.copy_multi(other.text_store,
287
count, cp_fail = branch.inventory_store.copy_multi(other.inventory_store,
289
count, cp_fail = branch.revision_store.copy_multi(other.revision_store,
292
assert len(cp_fail) == 0
293
return count, failures
289
if parent_id not in rev_id_to_root_id_map:
290
# We probably didn't read this revision, go spend the extra effort
293
tree = repo.revision_tree(parent_id)
294
except errors.NoSuchRevision:
295
# Ghost, fill out rev_id_to_root_id in case we encounter this
297
# But set parent_root_id to None since we don't really know
298
parent_root_id = None
300
parent_root_id = tree.get_root_id()
301
rev_id_to_root_id_map[parent_id] = None
303
# rev_id_to_root_id_map[parent_id] = parent_root_id
304
# memory consumption maybe?
306
parent_root_id = rev_id_to_root_id_map[parent_id]
307
if root_id == parent_root_id:
308
# With stacking we _might_ want to refer to a non-local revision,
309
# but this code path only applies when we have the full content
310
# available, so ghosts really are ghosts, not just the edge of
312
parent_ids.append(parent_id)
314
# root_id may be in the parent anyway.
316
tree = repo.revision_tree(parent_id)
317
except errors.NoSuchRevision:
318
# ghost, can't refer to it.
322
parent_ids.append(tree.get_file_revision(root_id))
323
except errors.NoSuchId:
326
# Drop non-head parents
328
graph = repo.get_graph()
329
heads = graph.heads(parent_ids)
331
for parent_id in parent_ids:
332
if parent_id in heads and parent_id not in selected_ids:
333
selected_ids.append(parent_id)
334
parent_keys = [(root_id, parent_id) for parent_id in selected_ids]
338
class TargetRepoKinds(object):
339
"""An enum-like set of constants.
341
They are the possible values of FetchSpecFactory.target_repo_kinds.
344
PREEXISTING = 'preexisting'
349
class FetchSpecFactory(object):
350
"""A helper for building the best fetch spec for a sprout call.
352
Factors that go into determining the sort of fetch to perform:
353
* did the caller specify any revision IDs?
354
* did the caller specify a source branch (need to fetch its
355
heads_to_fetch(), usually the tip + tags)
356
* is there an existing target repo (don't need to refetch revs it
358
* target is stacked? (similar to pre-existing target repo: even if
359
the target itself is new don't want to refetch existing revs)
361
:ivar source_branch: the source branch if one specified, else None.
362
:ivar source_branch_stop_revision_id: fetch up to this revision of
363
source_branch, rather than its tip.
364
:ivar source_repo: the source repository if one found, else None.
365
:ivar target_repo: the target repository acquired by sprout.
366
:ivar target_repo_kind: one of the TargetRepoKinds constants.
370
self._explicit_rev_ids = set()
371
self.source_branch = None
372
self.source_branch_stop_revision_id = None
373
self.source_repo = None
374
self.target_repo = None
375
self.target_repo_kind = None
378
def add_revision_ids(self, revision_ids):
379
"""Add revision_ids to the set of revision_ids to be fetched."""
380
self._explicit_rev_ids.update(revision_ids)
382
def make_fetch_spec(self):
383
"""Build a SearchResult or PendingAncestryResult or etc."""
384
if self.target_repo_kind is None or self.source_repo is None:
385
raise AssertionError(
386
'Incomplete FetchSpecFactory: %r' % (self.__dict__,))
387
if len(self._explicit_rev_ids) == 0 and self.source_branch is None:
388
if self.limit is not None:
389
raise NotImplementedError(
390
"limit is only supported with a source branch set")
391
# Caller hasn't specified any revisions or source branch
392
if self.target_repo_kind == TargetRepoKinds.EMPTY:
393
return vf_search.EverythingResult(self.source_repo)
395
# We want everything not already in the target (or target's
397
return vf_search.EverythingNotInOther(
398
self.target_repo, self.source_repo).execute()
399
heads_to_fetch = set(self._explicit_rev_ids)
400
if self.source_branch is not None:
401
must_fetch, if_present_fetch = self.source_branch.heads_to_fetch()
402
if self.source_branch_stop_revision_id is not None:
403
# Replace the tip rev from must_fetch with the stop revision
404
# XXX: this might be wrong if the tip rev is also in the
405
# must_fetch set for other reasons (e.g. it's the tip of
406
# multiple loom threads?), but then it's pretty unclear what it
407
# should mean to specify a stop_revision in that case anyway.
408
must_fetch.discard(self.source_branch.last_revision())
409
must_fetch.add(self.source_branch_stop_revision_id)
410
heads_to_fetch.update(must_fetch)
412
if_present_fetch = set()
413
if self.target_repo_kind == TargetRepoKinds.EMPTY:
414
# PendingAncestryResult does not raise errors if a requested head
415
# is absent. Ideally it would support the
416
# required_ids/if_present_ids distinction, but in practice
417
# heads_to_fetch will almost certainly be present so this doesn't
419
all_heads = heads_to_fetch.union(if_present_fetch)
420
ret = vf_search.PendingAncestryResult(all_heads, self.source_repo)
421
if self.limit is not None:
422
graph = self.source_repo.get_graph()
423
topo_order = list(graph.iter_topo_order(ret.get_keys()))
424
result_set = topo_order[:self.limit]
425
ret = self.source_repo.revision_ids_to_search_result(result_set)
428
return vf_search.NotInOtherForRevs(self.target_repo, self.source_repo,
429
required_ids=heads_to_fetch, if_present_ids=if_present_fetch,
430
limit=self.limit).execute()
added
removed
bzrlib/fetch.py
