|
1 # bzr.py - bzr support for the convert extension |
|
2 # |
|
3 # Copyright 2008, 2009 Marek Kubica <marek@xivilization.net> and others |
|
4 # |
|
5 # This software may be used and distributed according to the terms of the |
|
6 # GNU General Public License version 2 or any later version. |
|
7 |
|
8 # This module is for handling 'bzr', that was formerly known as Bazaar-NG; |
|
9 # it cannot access 'bar' repositories, but they were never used very much |
|
10 |
|
11 import os |
|
12 from mercurial import demandimport |
|
13 # these do not work with demandimport, blacklist |
|
14 demandimport.ignore.extend([ |
|
15 'bzrlib.transactions', |
|
16 'bzrlib.urlutils', |
|
17 'ElementPath', |
|
18 ]) |
|
19 |
|
20 from mercurial.i18n import _ |
|
21 from mercurial import util |
|
22 from common import NoRepo, commit, converter_source |
|
23 |
|
24 try: |
|
25 # bazaar imports |
|
26 from bzrlib import branch, revision, errors |
|
27 from bzrlib.revisionspec import RevisionSpec |
|
28 except ImportError: |
|
29 pass |
|
30 |
|
31 supportedkinds = ('file', 'symlink') |
|
32 |
|
33 class bzr_source(converter_source): |
|
34 """Reads Bazaar repositories by using the Bazaar Python libraries""" |
|
35 |
|
36 def __init__(self, ui, path, rev=None): |
|
37 super(bzr_source, self).__init__(ui, path, rev=rev) |
|
38 |
|
39 if not os.path.exists(os.path.join(path, '.bzr')): |
|
40 raise NoRepo(_('%s does not look like a Bazaar repository') |
|
41 % path) |
|
42 |
|
43 try: |
|
44 # access bzrlib stuff |
|
45 branch |
|
46 except NameError: |
|
47 raise NoRepo(_('Bazaar modules could not be loaded')) |
|
48 |
|
49 path = os.path.abspath(path) |
|
50 self._checkrepotype(path) |
|
51 self.branch = branch.Branch.open(path) |
|
52 self.sourcerepo = self.branch.repository |
|
53 self._parentids = {} |
|
54 |
|
55 def _checkrepotype(self, path): |
|
56 # Lightweight checkouts detection is informational but probably |
|
57 # fragile at API level. It should not terminate the conversion. |
|
58 try: |
|
59 from bzrlib import bzrdir |
|
60 dir = bzrdir.BzrDir.open_containing(path)[0] |
|
61 try: |
|
62 tree = dir.open_workingtree(recommend_upgrade=False) |
|
63 branch = tree.branch |
|
64 except (errors.NoWorkingTree, errors.NotLocalUrl): |
|
65 tree = None |
|
66 branch = dir.open_branch() |
|
67 if (tree is not None and tree.bzrdir.root_transport.base != |
|
68 branch.bzrdir.root_transport.base): |
|
69 self.ui.warn(_('warning: lightweight checkouts may cause ' |
|
70 'conversion failures, try with a regular ' |
|
71 'branch instead.\n')) |
|
72 except: |
|
73 self.ui.note(_('bzr source type could not be determined\n')) |
|
74 |
|
75 def before(self): |
|
76 """Before the conversion begins, acquire a read lock |
|
77 for all the operations that might need it. Fortunately |
|
78 read locks don't block other reads or writes to the |
|
79 repository, so this shouldn't have any impact on the usage of |
|
80 the source repository. |
|
81 |
|
82 The alternative would be locking on every operation that |
|
83 needs locks (there are currently two: getting the file and |
|
84 getting the parent map) and releasing immediately after, |
|
85 but this approach can take even 40% longer.""" |
|
86 self.sourcerepo.lock_read() |
|
87 |
|
88 def after(self): |
|
89 self.sourcerepo.unlock() |
|
90 |
|
91 def getheads(self): |
|
92 if not self.rev: |
|
93 return [self.branch.last_revision()] |
|
94 try: |
|
95 r = RevisionSpec.from_string(self.rev) |
|
96 info = r.in_history(self.branch) |
|
97 except errors.BzrError: |
|
98 raise util.Abort(_('%s is not a valid revision in current branch') |
|
99 % self.rev) |
|
100 return [info.rev_id] |
|
101 |
|
102 def getfile(self, name, rev): |
|
103 revtree = self.sourcerepo.revision_tree(rev) |
|
104 fileid = revtree.path2id(name.decode(self.encoding or 'utf-8')) |
|
105 kind = None |
|
106 if fileid is not None: |
|
107 kind = revtree.kind(fileid) |
|
108 if kind not in supportedkinds: |
|
109 # the file is not available anymore - was deleted |
|
110 raise IOError(_('%s is not available in %s anymore') % |
|
111 (name, rev)) |
|
112 mode = self._modecache[(name, rev)] |
|
113 if kind == 'symlink': |
|
114 target = revtree.get_symlink_target(fileid) |
|
115 if target is None: |
|
116 raise util.Abort(_('%s.%s symlink has no target') |
|
117 % (name, rev)) |
|
118 return target, mode |
|
119 else: |
|
120 sio = revtree.get_file(fileid) |
|
121 return sio.read(), mode |
|
122 |
|
123 def getchanges(self, version): |
|
124 # set up caches: modecache and revtree |
|
125 self._modecache = {} |
|
126 self._revtree = self.sourcerepo.revision_tree(version) |
|
127 # get the parentids from the cache |
|
128 parentids = self._parentids.pop(version) |
|
129 # only diff against first parent id |
|
130 prevtree = self.sourcerepo.revision_tree(parentids[0]) |
|
131 return self._gettreechanges(self._revtree, prevtree) |
|
132 |
|
133 def getcommit(self, version): |
|
134 rev = self.sourcerepo.get_revision(version) |
|
135 # populate parent id cache |
|
136 if not rev.parent_ids: |
|
137 parents = [] |
|
138 self._parentids[version] = (revision.NULL_REVISION,) |
|
139 else: |
|
140 parents = self._filterghosts(rev.parent_ids) |
|
141 self._parentids[version] = parents |
|
142 |
|
143 return commit(parents=parents, |
|
144 date='%d %d' % (rev.timestamp, -rev.timezone), |
|
145 author=self.recode(rev.committer), |
|
146 # bzr returns bytestrings or unicode, depending on the content |
|
147 desc=self.recode(rev.message), |
|
148 rev=version) |
|
149 |
|
150 def gettags(self): |
|
151 if not self.branch.supports_tags(): |
|
152 return {} |
|
153 tagdict = self.branch.tags.get_tag_dict() |
|
154 bytetags = {} |
|
155 for name, rev in tagdict.iteritems(): |
|
156 bytetags[self.recode(name)] = rev |
|
157 return bytetags |
|
158 |
|
159 def getchangedfiles(self, rev, i): |
|
160 self._modecache = {} |
|
161 curtree = self.sourcerepo.revision_tree(rev) |
|
162 if i is not None: |
|
163 parentid = self._parentids[rev][i] |
|
164 else: |
|
165 # no parent id, get the empty revision |
|
166 parentid = revision.NULL_REVISION |
|
167 |
|
168 prevtree = self.sourcerepo.revision_tree(parentid) |
|
169 changes = [e[0] for e in self._gettreechanges(curtree, prevtree)[0]] |
|
170 return changes |
|
171 |
|
172 def _gettreechanges(self, current, origin): |
|
173 revid = current._revision_id |
|
174 changes = [] |
|
175 renames = {} |
|
176 for (fileid, paths, changed_content, versioned, parent, name, |
|
177 kind, executable) in current.iter_changes(origin): |
|
178 |
|
179 if paths[0] == u'' or paths[1] == u'': |
|
180 # ignore changes to tree root |
|
181 continue |
|
182 |
|
183 # bazaar tracks directories, mercurial does not, so |
|
184 # we have to rename the directory contents |
|
185 if kind[1] == 'directory': |
|
186 if kind[0] not in (None, 'directory'): |
|
187 # Replacing 'something' with a directory, record it |
|
188 # so it can be removed. |
|
189 changes.append((self.recode(paths[0]), revid)) |
|
190 |
|
191 if None not in paths and paths[0] != paths[1]: |
|
192 # neither an add nor an delete - a move |
|
193 # rename all directory contents manually |
|
194 subdir = origin.inventory.path2id(paths[0]) |
|
195 # get all child-entries of the directory |
|
196 for name, entry in origin.inventory.iter_entries(subdir): |
|
197 # hg does not track directory renames |
|
198 if entry.kind == 'directory': |
|
199 continue |
|
200 frompath = self.recode(paths[0] + '/' + name) |
|
201 topath = self.recode(paths[1] + '/' + name) |
|
202 # register the files as changed |
|
203 changes.append((frompath, revid)) |
|
204 changes.append((topath, revid)) |
|
205 # add to mode cache |
|
206 mode = ((entry.executable and 'x') |
|
207 or (entry.kind == 'symlink' and 's') |
|
208 or '') |
|
209 self._modecache[(topath, revid)] = mode |
|
210 # register the change as move |
|
211 renames[topath] = frompath |
|
212 |
|
213 # no futher changes, go to the next change |
|
214 continue |
|
215 |
|
216 # we got unicode paths, need to convert them |
|
217 path, topath = [self.recode(part) for part in paths] |
|
218 |
|
219 if topath is None: |
|
220 # file deleted |
|
221 changes.append((path, revid)) |
|
222 continue |
|
223 |
|
224 # renamed |
|
225 if path and path != topath: |
|
226 renames[topath] = path |
|
227 changes.append((path, revid)) |
|
228 |
|
229 # populate the mode cache |
|
230 kind, executable = [e[1] for e in (kind, executable)] |
|
231 mode = ((executable and 'x') or (kind == 'symlink' and 'l') |
|
232 or '') |
|
233 self._modecache[(topath, revid)] = mode |
|
234 changes.append((topath, revid)) |
|
235 |
|
236 return changes, renames |
|
237 |
|
238 def _filterghosts(self, ids): |
|
239 """Filters out ghost revisions which hg does not support, see |
|
240 <http://bazaar-vcs.org/GhostRevision> |
|
241 """ |
|
242 parentmap = self.sourcerepo.get_parent_map(ids) |
|
243 parents = tuple([parent for parent in ids if parent in parentmap]) |
|
244 return parents |
|
245 |
|
246 def recode(self, s, encoding=None): |
|
247 """This version of recode tries to encode unicode to bytecode, |
|
248 and preferably using the UTF-8 codec. |
|
249 Other types than Unicode are silently returned, this is by |
|
250 intention, e.g. the None-type is not going to be encoded but instead |
|
251 just passed through |
|
252 """ |
|
253 if not encoding: |
|
254 encoding = self.encoding or 'utf-8' |
|
255 |
|
256 if isinstance(s, unicode): |
|
257 return s.encode(encoding) |
|
258 else: |
|
259 # leave it alone |
|
260 return s |