61
63
# 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
66
class RepoFetcher(object):
67
"""Pull revisions and texts from one repository to another.
70
if set, try to limit to the data this revision references.
86
73
count_copied -- number of revisions copied
88
count_weaves -- number of file weaves copied
75
This should not be used directly, it's essential a object to encapsulate
76
the logic in InterRepository.fetch().
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
78
def __init__(self, to_repository, from_repository, last_revision=None, pb=None):
99
80
self.failed_revisions = []
100
81
self.count_copied = 0
82
if to_repository.has_same_location(from_repository):
83
# repository.fetch should be taking care of this case.
84
raise errors.BzrError('RepoFetcher run '
85
'between two objects at the same location: '
86
'%r and %r' % (to_repository, from_repository))
87
self.to_repository = to_repository
88
self.from_repository = from_repository
89
# must not mutate self._last_revision as its potentially a shared instance
90
self._last_revision = last_revision
92
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
93
self.nested_pb = self.pb
97
self.from_repository.lock_read()
99
self.to_repository.lock_write()
101
self.to_repository.start_write_group()
105
self.to_repository.abort_write_group()
108
self.to_repository.commit_write_group()
110
if self.nested_pb is not None:
111
self.nested_pb.finished()
112
self.to_repository.unlock()
114
self.from_repository.unlock()
117
"""Primary worker function.
119
This initialises all the needed variables, and then fetches the
120
requested revisions, finally clearing the progress bar.
122
self.to_weaves = self.to_repository.weave_store
123
self.from_weaves = self.from_repository.weave_store
101
124
self.count_total = 0
102
self.count_weaves = 0
103
self.copied_file_ids = set()
104
125
self.file_ids_names = {}
106
self.pb = bzrlib.ui.ui_factory.progress_bar()
109
self.from_branch.lock_read()
126
pp = ProgressPhase('Fetch phase', 4, self.pb)
111
self._fetch_revisions(last_revision)
129
revs = self._revids_to_fetch()
132
self._fetch_everything_for_revisions(revs, pp)
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)):
136
def _fetch_everything_for_revisions(self, revs, pp):
137
"""Fetch all data for the given set of revisions."""
138
# The first phase is "file". We pass the progress bar for it directly
139
# into item_keys_introduced_by, which has more information about how
140
# that phase is progressing than we do. Progress updates for the other
141
# phases are taken care of in this function.
142
# XXX: there should be a clear owner of the progress reporting. Perhaps
143
# item_keys_introduced_by should have a richer API than it does at the
144
# moment, so that it can feed the progress information back to this
147
pb = bzrlib.ui.ui_factory.nested_progress_bar()
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
149
data_to_fetch = self.from_repository.item_keys_introduced_by(revs, pb)
150
for knit_kind, file_id, revisions in data_to_fetch:
151
if knit_kind != phase:
153
# Make a new progress bar for this phase
156
pb = bzrlib.ui.ui_factory.nested_progress_bar()
157
if knit_kind == "file":
158
self._fetch_weave_text(file_id, revisions)
159
elif knit_kind == "inventory":
161
# Once we've processed all the files, then we generate the root
162
# texts (if necessary), then we process the inventory. It's a
163
# bit distasteful to have knit_kind == "inventory" mean this,
164
# perhaps it should happen on the first non-"file" knit, in case
165
# it's not always inventory?
166
self._generate_root_texts(revs)
167
self._fetch_inventory_weave(revs, pb)
168
elif knit_kind == "signatures":
169
# Nothing to do here; this will be taken care of when
170
# _fetch_revision_texts happens.
172
elif knit_kind == "revisions":
173
self._fetch_revision_texts(revs, pb)
175
raise AssertionError("Unknown knit kind %r" % knit_kind)
179
self.count_copied += len(revs)
181
def _revids_to_fetch(self):
182
"""Determines the exact revisions needed from self.from_repository to
183
install self._last_revision in self.to_repository.
185
If no revisions need to be fetched, then this just returns None.
187
mutter('fetch up to rev {%s}', self._last_revision)
188
if self._last_revision is NULL_REVISION:
189
# explicit limit of no revisions needed
191
if (self._last_revision is not None and
192
self.to_repository.has_revision(self._last_revision)):
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]:
196
# XXX: this gets the full graph on both sides, and will make sure
197
# that ghosts are filled whether or not you care about them.
198
return self.to_repository.missing_revision_ids(self.from_repository,
200
except errors.NoSuchRevision:
201
raise InstallFailed([self._last_revision])
203
def _fetch_weave_text(self, file_id, required_versions):
225
204
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:
205
self.to_repository.get_transaction())
231
206
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
207
self.from_repository.get_transaction())
208
# we fetch all the texts, because texts do
209
# not reference anything, and its cheap enough
210
to_weave.join(from_weave, version_ids=required_versions)
211
# we don't need *all* of this data anymore, but we dont know
212
# what we do. This cache clearing will result in a new read
213
# of the knit data when we do the checkout, but probably we
214
# want to emit the needed data on the fly rather than at the
216
# the from weave should know not to cache data being joined,
217
# but its ok to ask it to clear.
218
from_weave.clear_cache()
219
to_weave.clear_cache()
221
def _fetch_inventory_weave(self, revs, pb):
222
pb.update("fetch inventory", 0, 2)
223
to_weave = self.to_repository.get_inventory_weave()
224
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
226
# just merge, this is optimisable and its means we don't
227
# copy unreferenced data such as not-needed inventories.
228
pb.update("fetch inventory", 1, 3)
229
from_weave = self.from_repository.get_inventory_weave()
230
pb.update("fetch inventory", 2, 3)
231
# we fetch only the referenced inventories because we do not
232
# know for unselected inventories whether all their required
233
# texts are present in the other repository - it could be
235
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
237
from_weave.clear_cache()
241
def _generate_root_texts(self, revs):
242
"""This will be called by __fetch between fetching weave texts and
243
fetching the inventory weave.
245
Subclasses should override this if they need to generate root texts
246
after fetching weave texts.
251
class GenericRepoFetcher(RepoFetcher):
252
"""This is a generic repo to repo fetcher.
254
This makes minimal assumptions about repo layout and contents.
255
It triggers a reconciliation after fetching to ensure integrity.
258
def _fetch_revision_texts(self, revs, pb):
259
"""Fetch revision object texts"""
260
to_txn = self.to_transaction = self.to_repository.get_transaction()
263
to_store = self.to_repository._revision_store
265
pb.update('copying revisions', count, total)
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)
267
sig_text = self.from_repository.get_signature_text(rev)
268
to_store.add_revision_signature_text(rev, sig_text, to_txn)
269
except errors.NoSuchRevision:
272
to_store.add_revision(self.from_repository.get_revision(rev),
275
# fixup inventory if needed:
276
# this is expensive because we have no inverse index to current ghosts.
277
# but on local disk its a few seconds and sftp push is already insane.
279
# FIXME: repository should inform if this is needed.
280
self.to_repository.reconcile()
283
class KnitRepoFetcher(RepoFetcher):
284
"""This is a knit format repository specific fetcher.
286
This differs from the GenericRepoFetcher by not doing a
287
reconciliation after copying, and using knit joining to
291
def _fetch_revision_texts(self, revs, pb):
292
# may need to be a InterRevisionStore call here.
293
from_transaction = self.from_repository.get_transaction()
294
to_transaction = self.to_repository.get_transaction()
295
to_sf = self.to_repository._revision_store.get_signature_file(
297
from_sf = self.from_repository._revision_store.get_signature_file(
299
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
300
to_rf = self.to_repository._revision_store.get_revision_file(
302
from_rf = self.from_repository._revision_store.get_revision_file(
304
to_rf.join(from_rf, version_ids=revs)
307
class Inter1and2Helper(object):
308
"""Helper for operations that convert data from model 1 and 2
310
This is for use by fetchers and converters.
313
def __init__(self, source, target):
316
:param source: The repository data comes from
317
:param target: The repository data goes to
322
def iter_rev_trees(self, revs):
323
"""Iterate through RevisionTrees efficiently.
325
Additionally, the inventory's revision_id is set if unset.
327
Trees are retrieved in batches of 100, and then yielded in the order
330
:param revs: A list of revision ids
333
for tree in self.source.revision_trees(revs[:100]):
334
if tree.inventory.revision_id is None:
335
tree.inventory.revision_id = tree.get_revision_id()
339
def generate_root_texts(self, revs):
340
"""Generate VersionedFiles for all root ids.
342
:param revs: the revisions to include
344
inventory_weave = self.source.get_inventory_weave()
347
to_store = self.target.weave_store
348
for tree in self.iter_rev_trees(revs):
349
revision_id = tree.inventory.root.revision
350
root_id = tree.inventory.root.file_id
351
parents = inventory_weave.get_parents(revision_id)
352
if root_id not in versionedfile:
353
versionedfile[root_id] = to_store.get_weave_or_empty(root_id,
354
self.target.get_transaction())
355
_, _, parent_texts[root_id] = versionedfile[root_id].add_lines(
356
revision_id, parents, [], parent_texts)
358
def regenerate_inventory(self, revs):
359
"""Generate a new inventory versionedfile in target, convertin data.
361
The inventory is retrieved from the source, (deserializing it), and
362
stored in the target (reserializing it in a different format).
363
:param revs: The revisions to include
365
inventory_weave = self.source.get_inventory_weave()
366
for tree in self.iter_rev_trees(revs):
367
parents = inventory_weave.get_parents(tree.get_revision_id())
368
self.target.add_inventory(tree.get_revision_id(), tree.inventory,
372
class Model1toKnit2Fetcher(GenericRepoFetcher):
373
"""Fetch from a Model1 repository into a Knit2 repository
375
def __init__(self, to_repository, from_repository, last_revision=None,
377
self.helper = Inter1and2Helper(from_repository, to_repository)
378
GenericRepoFetcher.__init__(self, to_repository, from_repository,
381
def _generate_root_texts(self, revs):
382
self.helper.generate_root_texts(revs)
384
def _fetch_inventory_weave(self, revs, pb):
385
self.helper.regenerate_inventory(revs)
388
class Knit1to2Fetcher(KnitRepoFetcher):
389
"""Fetch from a Knit1 repository into a Knit2 repository"""
391
def __init__(self, to_repository, from_repository, last_revision=None,
393
self.helper = Inter1and2Helper(from_repository, to_repository)
394
KnitRepoFetcher.__init__(self, to_repository, from_repository,
397
def _generate_root_texts(self, revs):
398
self.helper.generate_root_texts(revs)
400
def _fetch_inventory_weave(self, revs, pb):
401
self.helper.regenerate_inventory(revs)
404
class RemoteToOtherFetcher(GenericRepoFetcher):
406
def _fetch_everything_for_revisions(self, revs, pp):
407
data_stream = self.from_repository.get_data_stream(revs)
408
self.to_repository.insert_data_stream(data_stream)
added
removed
bzrlib/fetch.py
