« back to all changes in this revision
Viewing changes to bzrlib/fetch.py
- browse files at revision 1219
- compare with another revision
- download diff
- download tarball
- view history from revision 1219
- Committer: Martin Pool
- Date: 2005-09-13 01:37:23 UTC
- Revision ID: mbp@sourcefrog.net-20050913013723-7e0026b48cbf08ff
- BROKEN: start refactoring fetch code to work well with weaves
- files removed:
- bzrlib/selftest/test_ancestry.py
- files renamed:
- bzrlib/selftest/test_weave.py => tools/testweave.py
- files modified:
- .bzrignore
added
removed
bzrlib/fetch.py
14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
16
16
17
import sys
17
18
import os
18
from cStringIO import StringIO
19
19
20
20
import bzrlib.errors
21
from bzrlib.trace import mutter, note, warning
22
from bzrlib.branch import Branch, INVENTORY_FILEID, ANCESTRY_FILEID
21
from bzrlib.trace import mutter, note
22
from bzrlib.branch import Branch
23
23
from bzrlib.progress import ProgressBar
24
from bzrlib.xml5 import serializer_v5
25
from bzrlib.osutils import sha_string, split_lines
26
from bzrlib.errors import NoSuchRevision
27
28
"""Copying of history from one branch to another.
29
30
The basic plan is that every branch knows the history of everything
31
that has merged into it. As the first step of a merge, pull, or
32
branch operation we copy history from the source into the destination
33
branch.
34
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.
41
"""
42
43
# TODO: Avoid repeatedly opening weaves so many times.
44
45
# XXX: This doesn't handle ghost (not present in branch) revisions at
46
# all yet. I'm not sure they really should be supported.
47
48
# NOTE: This doesn't copy revisions which may be present but not
49
# merged into the last revision. I'm not sure we want to do that.
50
51
# - get a list of revisions that need to be pulled in
52
# - for each one, pull in that revision file
53
# and get the inventory, and store the inventory with right
54
# parents.
55
# - and get the ancestry, and store that with right parents too
56
# - and keep a note of all file ids and version seen
57
# - then go through all files; for each one get the weave,
58
# and add in all file versions
59
60
61
62
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
24
25
26
def greedy_fetch(to_branch, from_branch, revision, pb):
63
27
f = Fetcher(to_branch, from_branch, revision, pb)
64
28
return f.count_copied, f.failed_revisions
65
29
66
30
67
68
31
class Fetcher(object):
69
"""Pull revisions and texts from one branch to another.
70
71
This doesn't update the destination's history; that can be done
72
separately if desired.
73
74
revision_limit
75
If set, pull only up to this revision_id.
76
77
After running:
78
79
last_revision -- if last_revision
80
is given it will be that, otherwise the last revision of
81
from_branch
82
83
count_copied -- number of revisions copied
84
85
count_texts -- number of file texts copied
86
"""
87
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
32
"""Pull history from one branch to another."""
33
def __init__(self, to_branch, from_branch, revision_limit=None, pb=None):
88
34
self.to_branch = to_branch
89
self.to_weaves = to_branch.weave_store
90
35
self.from_branch = from_branch
91
self.from_weaves = from_branch.weave_store
92
self.failed_revisions = []
93
self.count_copied = 0
94
self.count_total = 0
95
self.count_texts = 0
36
self.revision_limit = revision_limit
96
37
if pb is None:
97
38
self.pb = bzrlib.ui.ui_factory.progress_bar()
98
39
else:
99
40
self.pb = pb
100
self.last_revision = self._find_last_revision(last_revision)
101
mutter('fetch up to rev {%s}', self.last_revision)
102
revs_to_fetch = self._compare_ancestries()
103
self._copy_revisions(revs_to_fetch)
104
self.new_ancestry = revs_to_fetch
105
106
107
108
def _find_last_revision(self, last_revision):
109
"""Find the limiting source revision.
110
111
Every ancestor of that revision will be merged across.
112
113
Returns the revision_id, or returns None if there's no history
114
in the source branch."""
115
self.pb.update('get source history')
116
from_history = self.from_branch.revision_history()
117
self.pb.update('get destination history')
118
if last_revision:
119
if last_revision not in from_history:
120
raise NoSuchRevision(self.from_branch, last_revision)
121
else:
122
return last_revision
123
elif from_history:
124
return from_history[-1]
125
else:
126
return None # no history in the source branch
127
128
129
def _compare_ancestries(self):
130
"""Get a list of revisions that must be copied.
131
132
That is, every revision that's in the ancestry of the source
133
branch and not in the destination branch."""
134
self.pb.update('get source ancestry')
135
self.from_ancestry = self.from_branch.get_ancestry(self.last_revision)
136
137
dest_last_rev = self.to_branch.last_revision()
138
self.pb.update('get destination ancestry')
139
if dest_last_rev:
140
dest_ancestry = self.to_branch.get_ancestry(dest_last_rev)
141
else:
142
dest_ancestry = []
143
ss = set(dest_ancestry)
144
to_fetch = []
145
for rev_id in self.from_ancestry:
146
if rev_id not in ss:
147
to_fetch.append(rev_id)
41
self._scan_histories()
42
self.failed_revisions = []
43
self.count_copied = 0
44
self._copy()
45
46
47
def _scan_histories(self):
48
self.from_history = from_branch.revision_history()
49
self.required_revisions = set(from_history)
50
self.to_history = to_branch.revision_history()
51
if self.revision_limit:
52
raise NotImplementedError('sorry, revision_limit not handled yet')
53
self.need_revisions = []
54
for rev_id in self.from_history:
55
if not has_revision(self.to_branch):
56
self.need_revisions.append(rev_id)
148
57
mutter('need to get revision {%s}', rev_id)
149
mutter('need to get %d revisions in total', len(to_fetch))
150
self.count_total = len(to_fetch)
151
return to_fetch
152
153
154
155
def _copy_revisions(self, revs_to_fetch):
156
i = 0
157
for rev_id in revs_to_fetch:
158
i += 1
159
if self.to_branch.has_revision(rev_id):
160
continue
161
self.pb.update('fetch revision', i, self.count_total)
162
self._copy_one_revision(rev_id)
163
self.count_copied += 1
164
165
166
def _copy_one_revision(self, rev_id):
167
"""Copy revision and everything referenced by it."""
168
mutter('copying revision {%s}', rev_id)
169
rev_xml = self.from_branch.get_revision_xml(rev_id)
170
inv_xml = self.from_branch.get_inventory_xml(rev_id)
171
rev = serializer_v5.read_revision_from_string(rev_xml)
172
inv = serializer_v5.read_inventory_from_string(inv_xml)
173
assert rev.revision_id == rev_id
174
assert rev.inventory_sha1 == sha_string(inv_xml)
175
mutter(' commiter %s, %d parents',
176
rev.committer,
177
len(rev.parents))
178
self._copy_new_texts(rev_id, inv)
179
parent_ids = [x.revision_id for x in rev.parents]
180
self._copy_inventory(rev_id, inv_xml, parent_ids)
181
self._copy_ancestry(rev_id, parent_ids)
182
self.to_branch.revision_store.add(StringIO(rev_xml), rev_id)
183
184
185
def _copy_inventory(self, rev_id, inv_xml, parent_ids):
186
self.to_weaves.add_text(INVENTORY_FILEID, rev_id,
187
split_lines(inv_xml), parent_ids)
188
189
190
def _copy_ancestry(self, rev_id, parent_ids):
191
ancestry_lines = self.from_weaves.get_lines(ANCESTRY_FILEID, rev_id)
192
self.to_weaves.add_text(ANCESTRY_FILEID, rev_id, ancestry_lines,
193
parent_ids)
194
58
59
60
def _copy(self):
61
while self.need_revisions:
62
rev_id = self.need_revisions.pop()
63
mutter('try to get revision {%s}', rev_id)
64
65
195
66
196
def _copy_new_texts(self, rev_id, inv):
197
"""Copy any new texts occuring in this revision."""
198
# TODO: Rather than writing out weaves every time, hold them
199
# in memory until everything's done? But this way is nicer
200
# if it's interrupted.
201
for path, ie in inv.iter_entries():
202
if ie.kind != 'file':
203
continue
204
if ie.text_version != rev_id:
205
continue
206
mutter('%s {%s} is changed in this revision',
207
path, ie.file_id)
208
self._copy_one_text(rev_id, ie.file_id)
209
210
211
def _copy_one_text(self, rev_id, file_id):
212
"""Copy one file text."""
213
mutter('copy text version {%s} of file {%s}',
214
rev_id, file_id)
215
from_weave = self.from_weaves.get_weave(file_id)
216
from_idx = from_weave.lookup(rev_id)
217
from_parents = map(from_weave.idx_to_name, from_weave.parents(from_idx))
218
text_lines = from_weave.get(from_idx)
219
to_weave = self.to_weaves.get_weave_or_empty(file_id)
220
to_parents = map(to_weave.lookup, from_parents)
221
# it's ok to add even if the text is already there
222
to_weave.add(rev_id, to_parents, text_lines)
223
self.to_weaves.put_weave(file_id, to_weave)
224
self.count_texts += 1
225
226
227
fetch = Fetcher
67
68
69
def has_revision(branch, revision_id):
70
try:
71
branch.get_revision_xml(revision_id)
72
return True
73
except bzrlib.errors.NoSuchRevision:
74
return False
75
76
77
def old_greedy_fetch(to_branch, from_branch, revision=None, pb=None):
78
"""Copy all history from one branch to another.
79
80
revision
81
If set, copy only up to this point in the source branch.
82
83
@returns: number copied, missing ids
84
"""
85
from_history = from_branch.revision_history()
86
required_revisions = set(from_history)
87
all_failed = set()
88
if revision is not None:
89
required_revisions.add(revision)
90
try:
91
rev_index = from_history.index(revision)
92
except ValueError:
93
rev_index = None
94
if rev_index is not None:
95
from_history = from_history[:rev_index + 1]
96
else:
97
from_history = [revision]
98
to_history = to_branch.revision_history()
99
missing = []
100
for rev_id in from_history:
101
if not has_revision(to_branch, rev_id):
102
missing.append(rev_id)
103
104
# recurse down through the revision graph, looking for things that
105
# can't be found.
106
count = 0
107
while len(missing) > 0:
108
installed, failed = to_branch.install_revisions(from_branch,
109
revision_ids=missing,
110
pb=pb)
111
count += installed
112
required_failed = failed.intersection(required_revisions)
113
if len(required_failed) > 0:
114
raise bzrlib.errors.InstallFailed(required_failed)
115
for rev_id in failed:
116
note("Failed to install %s" % rev_id)
117
all_failed.update(failed)
118
new_missing = set()
119
for rev_id in missing:
120
try:
121
revision = from_branch.get_revision(rev_id)
122
except bzrlib.errors.NoSuchRevision:
123
if revision in from_history:
124
raise
125
else:
126
continue
127
for parent in [p.revision_id for p in revision.parents]:
128
if not has_revision(to_branch, parent):
129
new_missing.add(parent)
130
missing = new_missing
131
return count, all_failed
132
133
134
def old_install_revisions(branch, other, revision_ids, pb):
135
"""Copy revisions from other branch into branch.
136
137
This is a lower-level function used by a pull or a merge. It
138
incorporates some history from one branch into another, but
139
does not update the revision history or operate on the working
140
copy.
141
142
revision_ids
143
Sequence of revisions to copy.
144
145
pb
146
Progress bar for copying.
147
"""
148
if False:
149
if hasattr(other.revision_store, "prefetch"):
150
other.revision_store.prefetch(revision_ids)
151
if hasattr(other.inventory_store, "prefetch"):
152
other.inventory_store.prefetch(revision_ids)
153
154
if pb is None:
155
pb = bzrlib.ui.ui_factory.progress_bar()
156
157
revisions = []
158
needed_texts = set()
159
i = 0
160
161
failures = set()
162
for i, rev_id in enumerate(revision_ids):
163
pb.update('fetching revision', i+1, len(revision_ids))
164
try:
165
rev = other.get_revision(rev_id)
166
except bzrlib.errors.NoSuchRevision:
167
failures.add(rev_id)
168
continue
169
170
revisions.append(rev)
171
inv = other.get_inventory(rev_id)
172
for key, entry in inv.iter_entries():
173
if entry.text_id is None:
174
continue
175
if entry.text_id not in branch.text_store:
176
needed_texts.add(entry.text_id)
177
178
pb.clear()
179
180
count, cp_fail = branch.text_store.copy_multi(other.text_store,
181
needed_texts)
182
count, cp_fail = branch.inventory_store.copy_multi(other.inventory_store,
183
revision_ids)
184
count, cp_fail = branch.revision_store.copy_multi(other.revision_store,
185
revision_ids,
186
permit_failure=True)
187
assert len(cp_fail) == 0
188
return count, failures
189
190
Loggerhead is a web-based interface for Breezy
Version: 2.0.1