62
60
# and add in all file versions
65
class RepoFetcher(object):
66
"""Pull revisions and texts from one repository to another.
69
if set, try to limit to the data this revision references.
64
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
65
f = Fetcher(to_branch, from_branch, revision, pb)
66
return f.count_copied, f.failed_revisions
70
class Fetcher(object):
71
"""Pull revisions and texts from one branch to another.
73
This doesn't update the destination's history; that can be done
74
separately if desired.
77
If set, pull only up to this revision_id.
81
last_revision -- if last_revision
82
is given it will be that, otherwise the last revision of
72
85
count_copied -- number of revisions copied
74
This should not be used directly, it's essential a object to encapsulate
75
the logic in InterRepository.fetch().
87
count_weaves -- number of file weaves copied
78
def __init__(self, to_repository, from_repository, last_revision=None, pb=None,
80
"""Create a repo fetcher.
82
:param find_ghosts: If True search the entire history for ghosts.
89
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
90
if to_branch == from_branch:
91
raise Exception("can't fetch from a branch to itself")
92
self.to_branch = to_branch
93
self.to_weaves = to_branch.weave_store
94
self.to_control = to_branch.control_weaves
95
self.from_branch = from_branch
96
self.from_weaves = from_branch.weave_store
97
self.from_control = from_branch.control_weaves
85
98
self.failed_revisions = []
86
99
self.count_copied = 0
87
if to_repository.has_same_location(from_repository):
88
# repository.fetch should be taking care of this case.
89
raise errors.BzrError('RepoFetcher run '
90
'between two objects at the same location: '
91
'%r and %r' % (to_repository, from_repository))
92
self.to_repository = to_repository
93
self.from_repository = from_repository
94
# must not mutate self._last_revision as its potentially a shared instance
95
self._last_revision = last_revision
96
self.find_ghosts = find_ghosts
101
self.count_weaves = 0
102
self.copied_file_ids = set()
98
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
99
self.nested_pb = self.pb
104
self.pb = bzrlib.ui.ui_factory.progress_bar()
102
self.nested_pb = None
103
self.from_repository.lock_read()
105
self.to_repository.lock_write()
107
self.to_repository.start_write_group()
111
self.to_repository.abort_write_group()
114
self.to_repository.commit_write_group()
116
if self.nested_pb is not None:
117
self.nested_pb.finished()
118
self.to_repository.unlock()
120
self.from_repository.unlock()
123
"""Primary worker function.
125
This initialises all the needed variables, and then fetches the
126
requested revisions, finally clearing the progress bar.
128
self.to_weaves = self.to_repository.weave_store
129
self.from_weaves = self.from_repository.weave_store
131
self.file_ids_names = {}
132
pp = ProgressPhase('Transferring', 4, self.pb)
135
search = self._revids_to_fetch()
138
if getattr(self, '_fetch_everything_for_search', None) is not None:
139
self._fetch_everything_for_search(search, pp)
141
# backward compatibility
142
self._fetch_everything_for_revisions(search.get_keys, pp)
107
self.from_branch.lock_read()
109
self._fetch_revisions(last_revision)
111
self.from_branch.unlock()
146
def _fetch_everything_for_search(self, search, pp):
147
"""Fetch all data for the given set of revisions."""
148
# The first phase is "file". We pass the progress bar for it directly
149
# into item_keys_introduced_by, which has more information about how
150
# that phase is progressing than we do. Progress updates for the other
151
# phases are taken care of in this function.
152
# XXX: there should be a clear owner of the progress reporting. Perhaps
153
# item_keys_introduced_by should have a richer API than it does at the
154
# moment, so that it can feed the progress information back to this
157
pb = bzrlib.ui.ui_factory.nested_progress_bar()
159
revs = search.get_keys()
160
data_to_fetch = self.from_repository.item_keys_introduced_by(revs, pb)
161
for knit_kind, file_id, revisions in data_to_fetch:
162
if knit_kind != phase:
164
# Make a new progress bar for this phase
167
pb = bzrlib.ui.ui_factory.nested_progress_bar()
168
if knit_kind == "file":
169
self._fetch_weave_text(file_id, revisions)
170
elif knit_kind == "inventory":
172
# Once we've processed all the files, then we generate the root
173
# texts (if necessary), then we process the inventory. It's a
174
# bit distasteful to have knit_kind == "inventory" mean this,
175
# perhaps it should happen on the first non-"file" knit, in case
176
# it's not always inventory?
177
self._generate_root_texts(revs)
178
self._fetch_inventory_weave(revs, pb)
179
elif knit_kind == "signatures":
180
# Nothing to do here; this will be taken care of when
181
# _fetch_revision_texts happens.
183
elif knit_kind == "revisions":
184
self._fetch_revision_texts(revs, pb)
186
raise AssertionError("Unknown knit kind %r" % knit_kind)
190
self.count_copied += len(revs)
192
def _revids_to_fetch(self):
193
"""Determines the exact revisions needed from self.from_repository to
194
install self._last_revision in self.to_repository.
196
If no revisions need to be fetched, then this just returns None.
198
mutter('fetch up to rev {%s}', self._last_revision)
199
if self._last_revision is NULL_REVISION:
200
# explicit limit of no revisions needed
202
if (self._last_revision is not None and
203
self.to_repository.has_revision(self._last_revision)):
206
return self.to_repository.search_missing_revision_ids(
207
self.from_repository, self._last_revision,
208
find_ghosts=self.find_ghosts)
209
except errors.NoSuchRevision:
210
raise InstallFailed([self._last_revision])
212
def _fetch_weave_text(self, file_id, required_versions):
114
def _fetch_revisions(self, last_revision):
115
self.last_revision = self._find_last_revision(last_revision)
116
mutter('fetch up to rev {%s}', self.last_revision)
117
if (self.last_revision is not None and
118
self.to_branch.has_revision(self.last_revision)):
121
revs_to_fetch = self._compare_ancestries()
123
raise InstallFailed([self.last_revision])
124
self._copy_revisions(revs_to_fetch)
125
self.new_ancestry = revs_to_fetch
127
def _find_last_revision(self, last_revision):
128
"""Find the limiting source revision.
130
Every ancestor of that revision will be merged across.
132
Returns the revision_id, or returns None if there's no history
133
in the source branch."""
136
self.pb.update('get source history')
137
from_history = self.from_branch.revision_history()
138
self.pb.update('get destination history')
140
return from_history[-1]
142
return None # no history in the source branch
145
def _compare_ancestries(self):
146
"""Get a list of revisions that must be copied.
148
That is, every revision that's in the ancestry of the source
149
branch and not in the destination branch."""
150
self.pb.update('get source ancestry')
151
self.from_ancestry = self.from_branch.get_ancestry(self.last_revision)
153
dest_last_rev = self.to_branch.last_revision()
154
self.pb.update('get destination ancestry')
156
dest_ancestry = self.to_branch.get_ancestry(dest_last_rev)
159
ss = set(dest_ancestry)
161
for rev_id in self.from_ancestry:
163
to_fetch.append(rev_id)
164
mutter('need to get revision {%s}', rev_id)
165
mutter('need to get %d revisions in total', len(to_fetch))
166
self.count_total = len(to_fetch)
169
def _copy_revisions(self, revs_to_fetch):
171
for rev_id in revs_to_fetch:
175
if self.to_branch.has_revision(rev_id):
177
self.pb.update('fetch revision', i, self.count_total)
178
self._copy_one_revision(rev_id)
179
self.count_copied += 1
182
def _copy_one_revision(self, rev_id):
183
"""Copy revision and everything referenced by it."""
184
mutter('copying revision {%s}', rev_id)
185
rev_xml = self.from_branch.get_revision_xml(rev_id)
186
inv_xml = self.from_branch.get_inventory_xml(rev_id)
187
rev = serializer_v5.read_revision_from_string(rev_xml)
188
inv = serializer_v5.read_inventory_from_string(inv_xml)
189
assert rev.revision_id == rev_id
190
assert rev.inventory_sha1 == sha_string(inv_xml)
191
mutter(' commiter %s, %d parents',
194
self._copy_new_texts(rev_id, inv)
195
parents = rev.parent_ids
196
new_parents = copy(parents)
197
for parent in parents:
198
if not self.to_branch.has_revision(parent):
199
new_parents.pop(new_parents.index(parent))
200
self._copy_inventory(rev_id, inv_xml, new_parents)
201
self.to_branch.revision_store.add(StringIO(rev_xml), rev_id)
202
mutter('copied revision %s', rev_id)
204
def _copy_inventory(self, rev_id, inv_xml, parent_ids):
205
self.to_control.add_text('inventory', rev_id,
206
split_lines(inv_xml), parent_ids,
207
self.to_branch.get_transaction())
209
def _copy_new_texts(self, rev_id, inv):
210
"""Copy any new texts occuring in this revision."""
211
# TODO: Rather than writing out weaves every time, hold them
212
# in memory until everything's done? But this way is nicer
213
# if it's interrupted.
214
for path, ie in inv.iter_entries():
215
if ie.revision != rev_id:
217
mutter('%s {%s} is changed in this revision',
219
self._copy_one_weave(rev_id, ie.file_id)
221
def _copy_one_weave(self, rev_id, file_id):
222
"""Copy one file weave."""
223
mutter('copy file {%s} modified in {%s}', file_id, rev_id)
224
if file_id in self.copied_file_ids:
225
mutter('file {%s} already copied', file_id)
227
from_weave = self.from_weaves.get_weave(file_id,
228
self.from_branch.get_transaction())
213
229
to_weave = self.to_weaves.get_weave_or_empty(file_id,
214
self.to_repository.get_transaction())
215
from_weave = self.from_weaves.get_weave(file_id,
216
self.from_repository.get_transaction())
217
# we fetch all the texts, because texts do
218
# not reference anything, and its cheap enough
219
to_weave.join(from_weave, version_ids=required_versions)
220
# we don't need *all* of this data anymore, but we dont know
221
# what we do. This cache clearing will result in a new read
222
# of the knit data when we do the checkout, but probably we
223
# want to emit the needed data on the fly rather than at the
225
# the from weave should know not to cache data being joined,
226
# but its ok to ask it to clear.
227
from_weave.clear_cache()
228
to_weave.clear_cache()
230
def _fetch_inventory_weave(self, revs, pb):
231
pb.update("fetch inventory", 0, 2)
232
to_weave = self.to_repository.get_inventory_weave()
233
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
230
self.to_branch.get_transaction())
235
# just merge, this is optimisable and its means we don't
236
# copy unreferenced data such as not-needed inventories.
237
pb.update("fetch inventory", 1, 3)
238
from_weave = self.from_repository.get_inventory_weave()
239
pb.update("fetch inventory", 2, 3)
240
# we fetch only the referenced inventories because we do not
241
# know for unselected inventories whether all their required
242
# texts are present in the other repository - it could be
244
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
246
from_weave.clear_cache()
250
def _generate_root_texts(self, revs):
251
"""This will be called by __fetch between fetching weave texts and
252
fetching the inventory weave.
254
Subclasses should override this if they need to generate root texts
255
after fetching weave texts.
260
class GenericRepoFetcher(RepoFetcher):
261
"""This is a generic repo to repo fetcher.
263
This makes minimal assumptions about repo layout and contents.
264
It triggers a reconciliation after fetching to ensure integrity.
267
def _fetch_revision_texts(self, revs, pb):
268
"""Fetch revision object texts"""
269
to_txn = self.to_transaction = self.to_repository.get_transaction()
272
to_store = self.to_repository._revision_store
274
pb.update('copying revisions', count, total)
276
sig_text = self.from_repository.get_signature_text(rev)
277
to_store.add_revision_signature_text(rev, sig_text, to_txn)
278
except errors.NoSuchRevision:
281
to_store.add_revision(self.from_repository.get_revision(rev),
284
# fixup inventory if needed:
285
# this is expensive because we have no inverse index to current ghosts.
286
# but on local disk its a few seconds and sftp push is already insane.
288
# FIXME: repository should inform if this is needed.
289
self.to_repository.reconcile()
292
class KnitRepoFetcher(RepoFetcher):
293
"""This is a knit format repository specific fetcher.
295
This differs from the GenericRepoFetcher by not doing a
296
reconciliation after copying, and using knit joining to
300
def _fetch_revision_texts(self, revs, pb):
301
# may need to be a InterRevisionStore call here.
302
from_transaction = self.from_repository.get_transaction()
303
to_transaction = self.to_repository.get_transaction()
304
to_sf = self.to_repository._revision_store.get_signature_file(
306
from_sf = self.from_repository._revision_store.get_signature_file(
308
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
309
to_rf = self.to_repository._revision_store.get_revision_file(
311
from_rf = self.from_repository._revision_store.get_revision_file(
313
to_rf.join(from_rf, version_ids=revs)
316
class Inter1and2Helper(object):
317
"""Helper for operations that convert data from model 1 and 2
319
This is for use by fetchers and converters.
322
def __init__(self, source, target):
325
:param source: The repository data comes from
326
:param target: The repository data goes to
331
def iter_rev_trees(self, revs):
332
"""Iterate through RevisionTrees efficiently.
334
Additionally, the inventory's revision_id is set if unset.
336
Trees are retrieved in batches of 100, and then yielded in the order
339
:param revs: A list of revision ids
341
# In case that revs is not a list.
344
for tree in self.source.revision_trees(revs[:100]):
345
if tree.inventory.revision_id is None:
346
tree.inventory.revision_id = tree.get_revision_id()
350
def generate_root_texts(self, revs):
351
"""Generate VersionedFiles for all root ids.
353
:param revs: the revisions to include
355
inventory_weave = self.source.get_inventory_weave()
358
to_store = self.target.weave_store
359
for tree in self.iter_rev_trees(revs):
360
revision_id = tree.inventory.root.revision
361
root_id = tree.get_root_id()
362
parents = inventory_weave.get_parents(revision_id)
363
if root_id not in versionedfile:
364
versionedfile[root_id] = to_store.get_weave_or_empty(root_id,
365
self.target.get_transaction())
366
_, _, parent_texts[root_id] = versionedfile[root_id].add_lines(
367
revision_id, parents, [], parent_texts)
369
def regenerate_inventory(self, revs):
370
"""Generate a new inventory versionedfile in target, convertin data.
372
The inventory is retrieved from the source, (deserializing it), and
373
stored in the target (reserializing it in a different format).
374
:param revs: The revisions to include
376
for tree in self.iter_rev_trees(revs):
377
parents = tree.get_parent_ids()
378
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
382
class Model1toKnit2Fetcher(GenericRepoFetcher):
383
"""Fetch from a Model1 repository into a Knit2 repository
385
def __init__(self, to_repository, from_repository, last_revision=None,
386
pb=None, find_ghosts=True):
387
self.helper = Inter1and2Helper(from_repository, to_repository)
388
GenericRepoFetcher.__init__(self, to_repository, from_repository,
389
last_revision, pb, find_ghosts)
391
def _generate_root_texts(self, revs):
392
self.helper.generate_root_texts(revs)
394
def _fetch_inventory_weave(self, revs, pb):
395
self.helper.regenerate_inventory(revs)
398
class Knit1to2Fetcher(KnitRepoFetcher):
399
"""Fetch from a Knit1 repository into a Knit2 repository"""
401
def __init__(self, to_repository, from_repository, last_revision=None,
402
pb=None, find_ghosts=True):
403
self.helper = Inter1and2Helper(from_repository, to_repository)
404
KnitRepoFetcher.__init__(self, to_repository, from_repository,
405
last_revision, pb, find_ghosts)
407
def _generate_root_texts(self, revs):
408
self.helper.generate_root_texts(revs)
410
def _fetch_inventory_weave(self, revs, pb):
411
self.helper.regenerate_inventory(revs)
414
class RemoteToOtherFetcher(GenericRepoFetcher):
416
def _fetch_everything_for_search(self, search, pp):
417
data_stream = self.from_repository.get_data_stream_for_search(search)
418
self.to_repository.insert_data_stream(data_stream)
232
to_weave.join(from_weave)
233
except errors.WeaveParentMismatch:
234
to_weave.reweave(from_weave)
235
self.to_weaves.put_weave(file_id, to_weave,
236
self.to_branch.get_transaction())
237
self.count_weaves += 1
238
self.copied_file_ids.add(file_id)
239
mutter('copied file {%s}', file_id)
added
removed
bzrlib/fetch.py
