1
# Copyright (C) 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""An adapter between a Git Repository and a Bazaar Branch"""
32
from bzrlib.transport import get_transport
34
from bzrlib.plugins.git import (
44
class GitRepository(repository.Repository):
45
"""An adapter to git repositories for bzr."""
49
def __init__(self, gitdir, lockfiles):
51
self.control_files = lockfiles
52
gitdirectory = gitdir.transport.local_abspath('.')
53
self.base = gitdirectory
54
self._git = model.GitModel(gitdirectory)
55
self._revision_cache = {}
57
self._blob_info_cache = {}
58
cache_dir = cache.create_cache_dir()
59
cachedir_transport = get_transport(cache_dir)
60
cache_file = os.path.join(cache_dir, 'cache-%s' % ids.NAMESPACE)
61
if not cachedbs.has_key(cache_file):
62
cachedbs[cache_file] = cache.sqlite3.connect(cache_file)
63
self.cachedb = cachedbs[cache_file]
66
def _init_cachedb(self):
67
self.cachedb.executescript("""
68
create table if not exists inventory (
70
create unique index if not exists inventory_revid
72
create table if not exists entry_revision (
78
create unique index if not exists entry_revision_revid_path
79
on entry_revision (inventory, path);
84
def _ancestor_revisions(self, revision_ids):
85
if revision_ids is not None:
86
git_revisions = [gitrevid_from_bzr(r) for r in revision_ids]
89
for lines in self._git.ancestor_lines(git_revisions):
90
yield self._parse_rev(lines)
91
# print "fetched ancestors:", git_revisions
96
def supports_rich_root(self):
99
def get_revision_graph(self, revision_id=None):
101
if revision_id is not None:
102
param = [ids.convert_revision_id_bzr_to_git(revision_id)]
105
git_graph = self._git.get_revision_graph(param)
106
# print "fetched revision graph:", param
107
for node, parents in git_graph.iteritems():
108
bzr_node = ids.convert_revision_id_git_to_bzr(node)
109
bzr_parents = [ids.convert_revision_id_git_to_bzr(n)
111
result[bzr_node] = bzr_parents
114
def get_revision_graph_with_ghosts(self, revision_ids=None):
115
graph = deprecated_graph.Graph()
116
if revision_ids is not None:
117
revision_ids = [ids.convert_revision_id_bzr_to_git(r)
118
for r in revision_ids]
119
git_graph = self._git.get_revision_graph(revision_ids)
120
# print "fetched revision graph (ghosts):", revision_ids
121
for node, parents in git_graph.iteritems():
122
bzr_node = ids.convert_revision_id_git_to_bzr(node)
123
bzr_parents = [ids.convert_revision_id_git_to_bzr(n)
126
graph.add_node(bzr_node, bzr_parents)
129
def get_ancestry(self, revision_id):
130
param = [ids.convert_revision_id_bzr_to_git(revision_id)]
131
git_ancestry = self._git.get_ancestry(param)
132
# print "fetched ancestry:", param
134
ids.convert_revision_id_git_to_bzr(git_id)
135
for git_id in git_ancestry]
137
def get_signature_text(self, revision_id):
138
raise errors.NoSuchRevision(self, revision_id)
141
def get_revision(self, revision_id):
142
if revision_id in self._revision_cache:
143
return self._revision_cache[revision_id]
144
git_commit_id = ids.convert_revision_id_bzr_to_git(revision_id)
145
raw = self._git.rev_list([git_commit_id], max_count=1, header=True)
146
# print "fetched revision:", git_commit_id
147
revision = self._parse_rev(raw)
148
self._revision_cache[revision_id] = revision
151
def has_revision(self, revision_id):
153
self.get_revision(revision_id)
154
except NoSuchRevision:
159
def get_revisions(self, revisions):
160
return [self.get_revision(r) for r in revisions]
163
def _parse_rev(klass, raw):
164
"""Parse a single git revision.
166
* The first line is the git commit id.
167
* Following lines conform to the 'name value' structure, until the
169
* All lines after the first blank line and until the NULL line have 4
170
leading spaces and constitute the commit message.
172
:param raw: sequence of newline-terminated strings, its last item is a
173
single NULL character.
174
:return: a `bzrlib.revision.Revision` object.
179
committer_was_set = False
180
revision_id = ids.convert_revision_id_git_to_bzr(raw[0][:-1])
181
rev = revision.Revision(revision_id)
182
rev.inventory_sha1 = ""
183
assert raw[-1] == '\x00', (
184
"Last item of raw was not a single NULL character.")
185
for line in raw[1:-1]:
187
assert line[:4] == ' ', (
188
"Unexpected line format in commit message: %r" % line)
189
message_lines.append(line[4:])
194
name, value = line[:-1].split(' ', 1)
196
rev.parent_ids.append(
197
ids.convert_revision_id_git_to_bzr(value))
200
author, timestamp, timezone = value.rsplit(' ', 2)
201
rev.properties['author'] = author
202
rev.properties['git-author-timestamp'] = timestamp
203
rev.properties['git-author-timezone'] = timezone
204
if not committer_was_set:
205
rev.committer = author
206
rev.timestamp = float(timestamp)
207
rev.timezone = klass._parse_tz(timezone)
209
if name == 'committer':
210
committer_was_set = True
211
committer, timestamp, timezone = value.rsplit(' ', 2)
212
rev.committer = committer
213
rev.timestamp = float(timestamp)
214
rev.timezone = klass._parse_tz(timezone)
217
rev.properties['git-tree-id'] = value
220
rev.message = ''.join(message_lines)
222
# XXX: That should not be needed, but current revision serializers do
223
# not know how how to handle text that is illegal in xml. Note: when
224
# this is fixed, we will need to rev up the revision namespace when
225
# removing the escaping code. -- David Allouche 2007-12-30
226
rev.message = escape_for_xml(rev.message)
227
rev.committer = escape_for_xml(rev.committer)
228
rev.properties['author'] = escape_for_xml(rev.properties['author'])
233
def _parse_tz(klass, tz):
234
"""Parse a timezone specification in the [+|-]HHMM format.
236
:return: the timezone offset in seconds.
239
sign = {'+': +1, '-': -1}[tz[0]]
241
minutes = int(tz[3:])
242
return sign * 60 * (60 * hours + minutes)
244
def revision_trees(self, revids):
246
yield self.revision_tree(revid)
248
def revision_tree(self, revision_id):
249
return GitRevisionTree(self, revision_id)
251
def _get_blob(self, git_id):
253
return self._blob_cache[git_id]
255
blob = self._git.cat_file('blob', git_id)
256
# print "fetched blob:", git_id
257
self._blob_cache[git_id] = blob
260
def _get_blob_info(self, git_id):
262
return self._blob_info_cache[git_id]
264
lines = self._get_blob(git_id)
265
size = sum(len(line) for line in lines)
266
sha1 = osutils.sha_strings(lines)
267
self._blob_info_cache[git_id] = (size, sha1)
270
def get_inventory(self, revision_id):
271
if revision_id is None:
272
revision_id = revision.NULL_REVISION
273
if revision_id == revision.NULL_REVISION:
274
return inventory.Inventory(
275
revision_id=revision_id, root_id=None)
277
# First pass at building the inventory. We need this one to get the
278
# git ids, so we do not have to cache the entire tree text. Ideally,
279
# this should be all we need to do.
280
git_commit = ids.convert_revision_id_bzr_to_git(revision_id)
281
git_inventory = self._git.get_inventory(git_commit)
282
# print "fetched inventory:", git_commit
283
inv = self._parse_inventory(revision_id, git_inventory)
285
# Second pass at building the inventory. There we retrieve additional
286
# data that bzrlib requires: text sizes, sha1s, symlink targets and
287
# revisions that introduced inventory entries
288
inv.git_file_data = {}
289
for file_id in sorted(inv.git_ids.iterkeys()):
290
git_id = inv.git_ids[file_id]
292
self._set_entry_text_info(inv, entry, git_id)
293
for file_id in sorted(inv.git_ids.iterkeys()):
294
git_id = inv.git_ids[file_id]
296
path = inv.id2path(file_id)
297
self._set_entry_revision(entry, revision_id, path, git_id)
299
# At this point the entry_revision table is fully populated for this
300
# revision. So record that we have inventory data for this revision.
301
self.cachedb.execute(
302
"insert or ignore into inventory (revid) values (?)",
304
self.cachedb.commit()
308
def _parse_inventory(klass, revid, git_inv):
309
# For now, git inventory do not have root ids. It is not clear that we
310
# can reliably support root ids. -- David Allouche 2007-12-28
311
inv = inventory.Inventory(revision_id=revid)
313
for perms, git_kind, git_id, path in git_inv:
316
if git_kind == 'blob':
319
executable = bool(int(perms[-3:], 8) & 0111)
320
elif perms[1] == '2':
323
raise AssertionError(
324
"Unknown blob kind, perms=%r." % (perms,))
325
elif git_kind == 'tree':
328
raise AssertionError(
329
"Unknown git entry kind: %r" % (git_kind,))
330
# XXX: Maybe the file id should be prefixed by file kind, so when
331
# the kind of path changes, the id changes too.
332
# -- David Allouche 2007-12-28.
333
file_id = escape_file_id(path.encode('utf-8'))
334
entry = inv.add_path(path, kind, file_id=file_id)
335
entry.executable = executable
336
inv.git_ids[file_id] = git_id
337
inv.root.revision = revid
340
def _set_entry_text_info(self, inv, entry, git_id):
341
if entry.kind == 'directory':
343
size, sha1 = self._get_blob_info(git_id)
344
entry.text_size = size
345
entry.text_sha1 = sha1
346
lines = self._get_blob(git_id)
347
if entry.kind == 'symlink':
348
entry.symlink_target = ''.join(lines)
349
inv.git_file_data[entry.file_id] = lines
351
def _get_file_revision(self, revision_id, path):
352
lines = self._git.rev_list(
353
[ids.convert_revision_id_bzr_to_git(revision_id)],
354
max_count=1, topo_order=True, paths=[path])
356
result = ids.convert_revision_id_git_to_bzr(line[:-1])
357
print "fetched file revision", line[:-1], path
360
def _get_entry_revision_from_db(self, revid, path, git_id, executable):
361
result = self.cachedb.execute(
362
"select revision from entry_revision where"
363
" inventory=? and path=? and gitid=? and executable=?",
364
(revid, path, git_id, executable)).fetchone()
370
def _set_entry_revision_in_db(self, revid, path, git_id, executable, revision):
371
self.cachedb.execute(
372
"insert into entry_revision"
373
" (inventory, path, gitid, executable, revision)"
374
" values (?, ?, ?, ?, ?)",
375
(revid, path, git_id, executable, revision))
377
def _all_inventories_in_db(self, revids):
379
result = self.cachedb.execute(
380
"select count(*) from inventory where revid = ?",
386
def _set_entry_revision(self, entry, revid, path, git_id):
387
# If a revision is in the cache, we assume it contains entries for the
388
# whole inventory. So if all parent revisions are in the cache, but no
389
# parent entry is present, then the entry revision is the current
390
# revision. That amortizes the number of _get_file_revision calls for
391
# large pulls to a "small number".
392
entry_rev = self._get_entry_revision_from_db(
393
revid, path, git_id, entry.executable)
394
if entry_rev is not None:
395
entry.revision = entry_rev
398
revision = self.get_revision(revid)
399
for parent_id in revision.parent_ids:
400
entry_rev = self._get_entry_revision_from_db(
401
parent_id, path, git_id, entry.executable)
402
if entry_rev is not None:
405
if self._all_inventories_in_db(revision.parent_ids):
408
entry_rev = self._get_file_revision(revid, path)
409
self._set_entry_revision_in_db(
410
revid, path, git_id, entry.executable, entry_rev)
411
#self.cachedb.commit()
412
entry.revision = entry_rev
415
def escape_file_id(file_id):
416
return file_id.replace('_', '__').replace(' ', '_s')
419
def escape_for_xml(message):
420
"""Replace xml-incompatible control characters."""
421
# Copied from _escape_commit_message from bzr-svn.
422
# -- David Allouche 2007-12-29.
426
# FIXME: RBC 20060419 this should be done by the revision
427
# serialiser not by commit. Then we can also add an unescaper
428
# in the deserializer and start roundtripping revision messages
429
# precisely. See repository_implementations/test_repository.py
431
# Python strings can include characters that can't be
432
# represented in well-formed XML; escape characters that
433
# aren't listed in the XML specification
434
# (http://www.w3.org/TR/REC-xml/#NT-Char).
435
message, _ = re.subn(
436
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
437
lambda match: match.group(0).encode('unicode_escape'),
442
class GitRevisionTree(revisiontree.RevisionTree):
444
def __init__(self, repository, revision_id):
445
if revision_id is None:
446
revision_id = revision.NULL_REVISION
447
self._inventory = repository.get_inventory(revision_id)
448
self._repository = repository
449
self._revision_id = revision_id
451
def get_file_lines(self, file_id):
452
entry = self._inventory[file_id]
453
if entry.kind == 'directory': return []
454
return self._inventory.git_file_data[file_id]
456
obj_id = self._inventory.git_ids[file_id]
457
assert obj_id is not None, (
458
"git_id must not be None: %r" % (self._inventory[file_id],))
459
return self._repository._git.cat_file('blob', obj_id)
added
removed
git_repository.py
