« back to all changes in this revision
Viewing changes to bzrlib/fetch.py
- browse files at revision 1285
- compare with another revision
- download diff
- download tarball
- view history from revision 1285
- Committer: Martin Pool
- Date: 2005-09-16 07:38:10 UTC
- Revision ID: mbp@sourcefrog.net-20050916073810-1f358be198c9ed91
- fix bug in committing files that are renamed but not modified
- add test for this
- add test for this
- files added:
- bzrlib/selftest/test_ancestry.py
- files renamed:
- tools/testweave.py => bzrlib/selftest/test_weave.py
- bzrlib/newinventory.py => contrib/newinventory.py
- files modified:
- .bzrignore
added
removed
bzrlib/fetch.py
13
13
# You should have received a copy of the GNU General Public License
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
17
import os
18
from cStringIO import StringIO
19
16
20
import bzrlib.errors
17
from bzrlib.trace import mutter, note
18
from bzrlib.branch import Branch
21
from bzrlib.trace import mutter, note, warning
22
from bzrlib.branch import Branch, INVENTORY_FILEID, ANCESTRY_FILEID
19
23
from bzrlib.progress import ProgressBar
20
import sys
21
import os
22
23
def has_revision(branch, revision_id):
24
try:
25
branch.get_revision_xml(revision_id)
26
return True
27
except bzrlib.errors.NoSuchRevision:
28
return False
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
29
60
30
61
31
62
def greedy_fetch(to_branch, from_branch, revision=None, pb=None):
32
"""Copy a revision and all available ancestors from one branch to another
33
If no revision is specified, uses the last revision in the source branch's
34
revision history.
63
f = Fetcher(to_branch, from_branch, revision, pb)
64
return f.count_copied, f.failed_revisions
65
66
67
68
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
35
86
"""
36
from_history = from_branch.revision_history()
37
required_revisions = set(from_history)
38
all_failed = set()
39
if revision is not None:
40
required_revisions.add(revision)
41
try:
42
rev_index = from_history.index(revision)
43
except ValueError:
44
rev_index = None
45
if rev_index is not None:
46
from_history = from_history[:rev_index + 1]
47
else:
48
from_history = [revision]
49
to_history = to_branch.revision_history()
50
missing = []
51
for rev_id in from_history:
52
if not has_revision(to_branch, rev_id):
53
missing.append(rev_id)
54
55
count = 0
56
while len(missing) > 0:
57
installed, failed = to_branch.install_revisions(from_branch,
58
revision_ids=missing,
59
pb=pb)
60
count += installed
61
required_failed = failed.intersection(required_revisions)
62
if len(required_failed) > 0:
63
raise bzrlib.errors.InstallFailed(required_failed)
64
for rev_id in failed:
65
note("Failed to install %s" % rev_id)
66
all_failed.update(failed)
67
new_missing = set()
68
for rev_id in missing:
69
try:
70
revision = from_branch.get_revision(rev_id)
71
except bzrlib.errors.NoSuchRevision:
72
if revision in from_history:
73
raise
74
else:
75
continue
76
for parent in [p.revision_id for p in revision.parents]:
77
if not has_revision(to_branch, parent):
78
new_missing.add(parent)
79
missing = new_missing
80
return count, all_failed
81
82
87
def __init__(self, to_branch, from_branch, last_revision=None, pb=None):
88
self.to_branch = to_branch
89
self.to_weaves = to_branch.weave_store
90
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
96
if pb is None:
97
self.pb = bzrlib.ui.ui_factory.progress_bar()
98
else:
99
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)
148
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
195
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
Loggerhead is a web-based interface for Breezy
Version: 2.0.1