60
59
# and add in all file versions
63
@deprecated_function(zero_eight)
64
63
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
65
"""Legacy API, please see branch.fetch(from_branch, last_revision, pb)."""
66
64
f = Fetcher(to_branch, from_branch, revision, pb)
67
65
return f.count_copied, f.failed_revisions
72
class RepoFetcher(object):
73
"""Pull revisions and texts from one repository to another.
76
if set, try to limit to the data this revision references.
69
class Fetcher(object):
70
"""Pull revisions and texts from one branch to another.
72
This doesn't update the destination's history; that can be done
73
separately if desired.
76
If set, pull only up to this revision_id.
80
last_revision -- if last_revision
81
is given it will be that, otherwise the last revision of
79
84
count_copied -- number of revisions copied
81
This should not be used directory, its essential a object to encapsulate
82
the logic in InterRepository.fetch().
86
count_weaves -- number of file weaves copied
84
def __init__(self, to_repository, from_repository, last_revision=None, pb=None):
88
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
89
if to_branch == from_branch:
90
raise Exception("can't fetch from a branch to itself")
91
self.to_branch = to_branch
92
self.to_weaves = to_branch.weave_store
93
self.to_control = to_branch.control_weaves
94
self.from_branch = from_branch
95
self.from_weaves = from_branch.weave_store
96
self.from_control = from_branch.control_weaves
86
97
self.failed_revisions = []
87
98
self.count_copied = 0
88
if to_repository.control_files._transport.base == from_repository.control_files._transport.base:
89
# check that last_revision is in 'from' and then return a no-operation.
90
if last_revision not in (None, NULL_REVISION):
91
from_repository.get_revision(last_revision)
93
self.to_repository = to_repository
94
self.from_repository = from_repository
95
# must not mutate self._last_revision as its potentially a shared instance
96
self._last_revision = last_revision
100
self.count_weaves = 0
101
self.copied_file_ids = set()
98
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
99
self.nested_pb = self.pb
103
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()
109
if self.nested_pb is not None:
110
self.nested_pb.finished()
111
self.to_repository.unlock()
113
self.from_repository.unlock()
116
"""Primary worker function.
118
This initialises all the needed variables, and then fetches the
119
requested revisions, finally clearing the progress bar.
121
self.to_weaves = self.to_repository.weave_store
122
self.to_control = self.to_repository.control_weaves
123
self.from_weaves = self.from_repository.weave_store
124
self.from_control = self.from_repository.control_weaves
126
self.file_ids_names = {}
127
pp = ProgressPhase('fetch phase', 4, self.pb)
129
revs = self._revids_to_fetch()
133
self._fetch_weave_texts(revs)
135
self._fetch_inventory_weave(revs)
137
self._fetch_revision_texts(revs)
138
self.count_copied += len(revs)
106
self.from_branch.lock_read()
108
self._fetch_revisions(last_revision)
110
self.from_branch.unlock()
142
def _revids_to_fetch(self):
143
mutter('fetch up to rev {%s}', self._last_revision)
144
if self._last_revision is NULL_REVISION:
145
# explicit limit of no revisions needed
147
if (self._last_revision != None and
148
self.to_repository.has_revision(self._last_revision)):
113
def _fetch_revisions(self, last_revision):
114
self.last_revision = self._find_last_revision(last_revision)
115
mutter('fetch up to rev {%s}', self.last_revision)
116
if (self.last_revision is not None and
117
self.to_branch.has_revision(self.last_revision)):
120
revs_to_fetch = self._compare_ancestries()
122
raise InstallFailed([self.last_revision])
123
self._copy_revisions(revs_to_fetch)
124
self.new_ancestry = revs_to_fetch
126
def _find_last_revision(self, last_revision):
127
"""Find the limiting source revision.
129
Every ancestor of that revision will be merged across.
131
Returns the revision_id, or returns None if there's no history
132
in the source branch."""
135
self.pb.update('get source history')
136
from_history = self.from_branch.revision_history()
137
self.pb.update('get destination history')
139
return from_history[-1]
141
return None # no history in the source branch
152
return self.to_repository.missing_revision_ids(self.from_repository,
154
except errors.NoSuchRevision:
155
raise InstallFailed([self._last_revision])
157
def _fetch_weave_texts(self, revs):
158
texts_pb = bzrlib.ui.ui_factory.nested_progress_bar()
160
file_ids = self.from_repository.fileid_involved_by_set(revs)
162
num_file_ids = len(file_ids)
163
for file_id in file_ids:
164
texts_pb.update("fetch texts", count, num_file_ids)
166
to_weave = self.to_weaves.get_weave_or_empty(file_id,
167
self.to_repository.get_transaction())
168
from_weave = self.from_weaves.get_weave(file_id,
169
self.from_repository.get_transaction())
170
# we fetch all the texts, because texts do
171
# not reference anything, and its cheap enough
172
to_weave.join(from_weave)
176
def _fetch_inventory_weave(self, revs):
177
pb = bzrlib.ui.ui_factory.nested_progress_bar()
179
pb.update("fetch inventory", 0, 2)
180
to_weave = self.to_control.get_weave('inventory',
181
self.to_repository.get_transaction())
183
child_pb = bzrlib.ui.ui_factory.nested_progress_bar()
185
# just merge, this is optimisable and its means we dont
186
# copy unreferenced data such as not-needed inventories.
187
pb.update("fetch inventory", 1, 3)
188
from_weave = self.from_repository.get_inventory_weave()
189
pb.update("fetch inventory", 2, 3)
190
# we fetch only the referenced inventories because we do not
191
# know for unselected inventories whether all their required
192
# texts are present in the other repository - it could be
194
to_weave.join(from_weave, pb=child_pb, msg='merge inventory',
202
class GenericRepoFetcher(RepoFetcher):
203
"""This is a generic repo to repo fetcher.
205
This makes minimal assumptions about repo layout and contents.
206
It triggers a reconciliation after fetching to ensure integrity.
209
def _fetch_revision_texts(self, revs):
210
"""Fetch revision object texts"""
211
rev_pb = bzrlib.ui.ui_factory.nested_progress_bar()
213
to_txn = self.to_transaction = self.to_repository.get_transaction()
216
to_store = self.to_repository._revision_store
218
pb = bzrlib.ui.ui_factory.nested_progress_bar()
220
pb.update('copying revisions', count, total)
222
sig_text = self.from_repository.get_signature_text(rev)
223
to_store.add_revision_signature_text(rev, sig_text, to_txn)
224
except errors.NoSuchRevision:
227
to_store.add_revision(self.from_repository.get_revision(rev),
232
# fixup inventory if needed:
233
# this is expensive because we have no inverse index to current ghosts.
234
# but on local disk its a few seconds and sftp push is already insane.
236
# FIXME: repository should inform if this is needed.
237
self.to_repository.reconcile()
242
class KnitRepoFetcher(RepoFetcher):
243
"""This is a knit format repository specific fetcher.
245
This differs from the GenericRepoFetcher by not doing a
246
reconciliation after copying, and using knit joining to
250
def _fetch_revision_texts(self, revs):
251
# may need to be a InterRevisionStore call here.
252
from_transaction = self.from_repository.get_transaction()
253
to_transaction = self.to_repository.get_transaction()
254
to_sf = self.to_repository._revision_store.get_signature_file(
256
from_sf = self.from_repository._revision_store.get_signature_file(
258
to_sf.join(from_sf, version_ids=revs, ignore_missing=True)
259
to_rf = self.to_repository._revision_store.get_revision_file(
261
from_rf = self.from_repository._revision_store.get_revision_file(
263
to_rf.join(from_rf, version_ids=revs)
266
class Fetcher(object):
267
"""Backwards compatability glue for branch.fetch()."""
269
@deprecated_method(zero_eight)
270
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
271
"""Please see branch.fetch()."""
272
to_branch.fetch(from_branch, last_revision, pb)
144
def _compare_ancestries(self):
145
"""Get a list of revisions that must be copied.
147
That is, every revision that's in the ancestry of the source
148
branch and not in the destination branch."""
149
self.pb.update('get source ancestry')
150
self.from_ancestry = self.from_branch.get_ancestry(self.last_revision)
152
dest_last_rev = self.to_branch.last_revision()
153
self.pb.update('get destination ancestry')
155
dest_ancestry = self.to_branch.get_ancestry(dest_last_rev)
158
ss = set(dest_ancestry)
160
for rev_id in self.from_ancestry:
162
to_fetch.append(rev_id)
163
mutter('need to get revision {%s}', rev_id)
164
mutter('need to get %d revisions in total', len(to_fetch))
165
self.count_total = len(to_fetch)
168
def _copy_revisions(self, revs_to_fetch):
170
for rev_id in revs_to_fetch:
174
if self.to_branch.has_revision(rev_id):
176
self.pb.update('fetch revision', i, self.count_total)
177
self._copy_one_revision(rev_id)
178
self.count_copied += 1
181
def _copy_one_revision(self, rev_id):
182
"""Copy revision and everything referenced by it."""
183
mutter('copying revision {%s}', rev_id)
184
rev_xml = self.from_branch.get_revision_xml(rev_id)
185
inv_xml = self.from_branch.get_inventory_xml(rev_id)
186
rev = serializer_v5.read_revision_from_string(rev_xml)
187
inv = serializer_v5.read_inventory_from_string(inv_xml)
188
assert rev.revision_id == rev_id
189
assert rev.inventory_sha1 == sha_string(inv_xml)
190
mutter(' commiter %s, %d parents',
193
self._copy_new_texts(rev_id, inv)
194
parents = rev.parent_ids
195
for parent in parents:
196
if not self.to_branch.has_revision(parent):
197
parents.pop(parents.index(parent))
198
self._copy_inventory(rev_id, inv_xml, parents)
199
self.to_branch.revision_store.add(StringIO(rev_xml), rev_id)
200
mutter('copied revision %s', rev_id)
203
def _copy_inventory(self, rev_id, inv_xml, parent_ids):
204
self.to_control.add_text('inventory', rev_id,
205
split_lines(inv_xml), parent_ids,
206
self.to_branch.get_transaction())
208
def _copy_new_texts(self, rev_id, inv):
209
"""Copy any new texts occuring in this revision."""
210
# TODO: Rather than writing out weaves every time, hold them
211
# in memory until everything's done? But this way is nicer
212
# if it's interrupted.
213
for path, ie in inv.iter_entries():
214
if ie.revision != rev_id:
216
mutter('%s {%s} is changed in this revision',
218
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())
229
to_weave = self.to_weaves.get_weave_or_empty(file_id,
230
self.to_branch.get_transaction())
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
