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
self._git = self._make_model(gitdir.transport)
53
self._revision_cache = {}
55
self._blob_info_cache = {}
56
cache_dir = cache.create_cache_dir()
57
cachedir_transport = get_transport(cache_dir)
58
cache_file = os.path.join(cache_dir, 'cache-%s' % ids.NAMESPACE)
59
if not cachedbs.has_key(cache_file):
60
cachedbs[cache_file] = cache.sqlite3.connect(cache_file)
61
self.cachedb = cachedbs[cache_file]
63
self._format = GitFormat()
65
def _init_cachedb(self):
66
self.cachedb.executescript("""
67
create table if not exists inventory (
69
create unique index if not exists inventory_revid
71
create table if not exists entry_revision (
77
create unique index if not exists entry_revision_revid_path
78
on entry_revision (inventory, path);
84
def _make_model(klass, transport):
85
gitdirectory = transport.local_abspath('.')
86
return model.GitModel(gitdirectory)
89
def _ancestor_revisions(self, revision_ids):
90
if revision_ids is not None:
91
git_revisions = [gitrevid_from_bzr(r) for r in revision_ids]
94
for lines in self._git.ancestor_lines(git_revisions):
95
yield self._parse_rev(lines)
96
# print "fetched ancestors:", git_revisions
101
def supports_rich_root(self):
104
def get_revision_graph(self, revision_id=None):
106
if revision_id is not None:
107
param = [ids.convert_revision_id_bzr_to_git(revision_id)]
110
git_graph = self._git.get_revision_graph(param)
111
# print "fetched revision graph:", param
112
for node, parents in git_graph.iteritems():
113
bzr_node = ids.convert_revision_id_git_to_bzr(node)
114
bzr_parents = [ids.convert_revision_id_git_to_bzr(n)
116
result[bzr_node] = bzr_parents
119
def get_revision_graph_with_ghosts(self, revision_ids=None):
120
graph = deprecated_graph.Graph()
121
if revision_ids is not None:
122
revision_ids = [ids.convert_revision_id_bzr_to_git(r)
123
for r in revision_ids]
124
git_graph = self._git.get_revision_graph(revision_ids)
125
# print "fetched revision graph (ghosts):", revision_ids
126
for node, parents in git_graph.iteritems():
127
bzr_node = ids.convert_revision_id_git_to_bzr(node)
128
bzr_parents = [ids.convert_revision_id_git_to_bzr(n)
131
graph.add_node(bzr_node, bzr_parents)
134
def get_ancestry(self, revision_id):
135
param = [ids.convert_revision_id_bzr_to_git(revision_id)]
136
git_ancestry = self._git.get_ancestry(param)
137
# print "fetched ancestry:", param
139
ids.convert_revision_id_git_to_bzr(git_id)
140
for git_id in git_ancestry]
142
def get_signature_text(self, revision_id):
143
raise errors.NoSuchRevision(self, revision_id)
146
def get_revision(self, revision_id):
147
if revision_id in self._revision_cache:
148
return self._revision_cache[revision_id]
149
git_commit_id = ids.convert_revision_id_bzr_to_git(revision_id)
150
raw = self._git.rev_list([git_commit_id], max_count=1, header=True)
151
print "fetched revision:", git_commit_id
152
revision = self._parse_rev(raw)
153
self._revision_cache[revision_id] = revision
156
def has_revision(self, revision_id):
158
self.get_revision(revision_id)
159
except NoSuchRevision:
164
def get_revisions(self, revisions):
165
return [self.get_revision(r) for r in revisions]
168
def _parse_rev(klass, raw):
169
"""Parse a single git revision.
171
* The first line is the git commit id.
172
* Following lines conform to the 'name value' structure, until the
174
* All lines after the first blank line and until the NULL line have 4
175
leading spaces and constitute the commit message.
177
:param raw: sequence of newline-terminated strings, its last item is a
178
single NULL character.
179
:return: a `bzrlib.revision.Revision` object.
184
committer_was_set = False
185
revision_id = ids.convert_revision_id_git_to_bzr(raw[0][:-1])
186
rev = revision.Revision(revision_id)
187
rev.inventory_sha1 = ""
188
assert raw[-1] == '\x00', (
189
"Last item of raw was not a single NULL character.")
190
for line in raw[1:-1]:
192
assert line[:4] == ' ', (
193
"Unexpected line format in commit message: %r" % line)
194
message_lines.append(line[4:])
199
name, value = line[:-1].split(' ', 1)
201
rev.parent_ids.append(
202
ids.convert_revision_id_git_to_bzr(value))
205
author, timestamp, timezone = value.rsplit(' ', 2)
206
rev.properties['author'] = author
207
rev.properties['git-author-timestamp'] = timestamp
208
rev.properties['git-author-timezone'] = timezone
209
if not committer_was_set:
210
rev.committer = author
211
rev.timestamp = float(timestamp)
212
rev.timezone = klass._parse_tz(timezone)
214
if name == 'committer':
215
committer_was_set = True
216
committer, timestamp, timezone = value.rsplit(' ', 2)
217
rev.committer = committer
218
rev.timestamp = float(timestamp)
219
rev.timezone = klass._parse_tz(timezone)
222
rev.properties['git-tree-id'] = value
225
rev.message = ''.join(message_lines)
227
# XXX: That should not be needed, but current revision serializers do
228
# not know how how to handle text that is illegal in xml. Note: when
229
# this is fixed, we will need to rev up the revision namespace when
230
# removing the escaping code. -- David Allouche 2007-12-30
231
rev.message = escape_for_xml(rev.message)
232
rev.committer = escape_for_xml(rev.committer)
233
rev.properties['author'] = escape_for_xml(rev.properties['author'])
238
def _parse_tz(klass, tz):
239
"""Parse a timezone specification in the [+|-]HHMM format.
241
:return: the timezone offset in seconds.
244
sign = {'+': +1, '-': -1}[tz[0]]
246
minutes = int(tz[3:])
247
return sign * 60 * (60 * hours + minutes)
249
def revision_trees(self, revids):
251
yield self.revision_tree(revid)
253
def revision_tree(self, revision_id):
254
return GitRevisionTree(self, revision_id)
256
def _fetch_blob(self, git_id):
257
lines = self._git.cat_file('blob', git_id)
258
print "fetched blob:", git_id
259
if self._building_inventory is not None:
260
self._building_inventory.git_file_data[git_id] = lines
263
def _get_blob(self, git_id):
265
return self._blob_cache[git_id]
267
return self._fetch_blob(git_id)
269
def _get_blob_caching(self, git_id):
271
return self._blob_cache[git_id]
273
lines = self._fetch_blob(git_id)
274
self._blob_cache[git_id] = lines
277
def _get_blob_info(self, git_id):
279
return self._blob_info_cache[git_id]
281
lines = self._get_blob(git_id)
282
size = sum(len(line) for line in lines)
283
sha1 = osutils.sha_strings(lines)
284
self._blob_info_cache[git_id] = (size, sha1)
287
def get_inventory(self, revision_id):
288
if revision_id is None:
289
revision_id = revision.NULL_REVISION
290
if revision_id == revision.NULL_REVISION:
291
return inventory.Inventory(
292
revision_id=revision_id, root_id=None)
294
# First pass at building the inventory. We need this one to get the
295
# git ids, so we do not have to cache the entire tree text. Ideally,
296
# this should be all we need to do.
297
git_commit = ids.convert_revision_id_bzr_to_git(revision_id)
298
git_inventory = self._git.get_inventory(git_commit)
299
# print "fetched inventory:", git_commit
300
inv = self._parse_inventory(revision_id, git_inventory)
302
# Second pass at building the inventory. There we retrieve additional
303
# data that bzrlib requires: text sizes, sha1s, symlink targets and
304
# revisions that introduced inventory entries
305
self._building_inventory = inv
306
self._building_inventory.git_file_data = {}
307
for file_id in sorted(inv.git_ids.iterkeys()):
308
git_id = inv.git_ids[file_id]
310
self._set_entry_text_info(inv, entry, git_id)
311
for file_id in sorted(inv.git_ids.iterkeys()):
312
git_id = inv.git_ids[file_id]
314
path = inv.id2path(file_id)
315
self._set_entry_revision(entry, revision_id, path, git_id)
317
# At this point the entry_revision table is fully populated for this
318
# revision. So record that we have inventory data for this revision.
319
self.cachedb.execute(
320
"insert or ignore into inventory (revid) values (?)",
322
self.cachedb.commit()
323
self._building_inventory = None
327
def _parse_inventory(klass, revid, git_inv):
328
# For now, git inventory do not have root ids. It is not clear that we
329
# can reliably support root ids. -- David Allouche 2007-12-28
330
inv = inventory.Inventory(revision_id=revid)
332
for perms, git_kind, git_id, path in git_inv:
335
if git_kind == 'blob':
338
executable = bool(int(perms[-3:], 8) & 0111)
339
elif perms[1] == '2':
342
raise AssertionError(
343
"Unknown blob kind, perms=%r." % (perms,))
344
elif git_kind == 'tree':
347
raise AssertionError(
348
"Unknown git entry kind: %r" % (git_kind,))
349
# XXX: Maybe the file id should be prefixed by file kind, so when
350
# the kind of path changes, the id changes too.
351
# -- David Allouche 2007-12-28.
352
file_id = escape_file_id(path.encode('utf-8'))
353
entry = inv.add_path(path, kind, file_id=file_id)
354
entry.executable = executable
355
inv.git_ids[file_id] = git_id
356
inv.root.revision = revid
359
def _set_entry_text_info(self, inv, entry, git_id):
360
if entry.kind == 'directory':
362
size, sha1 = self._get_blob_info(git_id)
363
entry.text_size = size
364
entry.text_sha1 = sha1
365
if entry.kind == 'symlink':
366
lines = self._get_blob_caching(git_id)
367
entry.symlink_target = ''.join(lines)
369
def _get_file_revision(self, revision_id, path):
370
lines = self._git.rev_list(
371
[ids.convert_revision_id_bzr_to_git(revision_id)],
372
max_count=1, topo_order=True, paths=[path])
374
result = ids.convert_revision_id_git_to_bzr(line[:-1])
375
print "fetched file revision", line[:-1], path
378
def _get_entry_revision_from_db(self, revid, path, git_id, executable):
379
result = self.cachedb.execute(
380
"select revision from entry_revision where"
381
" inventory=? and path=? and gitid=? and executable=?",
382
(revid, path, git_id, executable)).fetchone()
388
def _set_entry_revision_in_db(self, revid, path, git_id, executable, revision):
389
self.cachedb.execute(
390
"insert into entry_revision"
391
" (inventory, path, gitid, executable, revision)"
392
" values (?, ?, ?, ?, ?)",
393
(revid, path, git_id, executable, revision))
395
def _all_inventories_in_db(self, revids):
397
result = self.cachedb.execute(
398
"select count(*) from inventory where revid = ?",
404
def _set_entry_revision(self, entry, revid, path, git_id):
405
# If a revision is in the cache, we assume it contains entries for the
406
# whole inventory. So if all parent revisions are in the cache, but no
407
# parent entry is present, then the entry revision is the current
408
# revision. That amortizes the number of _get_file_revision calls for
409
# large pulls to a "small number".
410
entry_rev = self._get_entry_revision_from_db(
411
revid, path, git_id, entry.executable)
412
if entry_rev is not None:
413
entry.revision = entry_rev
416
revision = self.get_revision(revid)
417
for parent_id in revision.parent_ids:
418
entry_rev = self._get_entry_revision_from_db(
419
parent_id, path, git_id, entry.executable)
420
if entry_rev is not None:
423
if self._all_inventories_in_db(revision.parent_ids):
426
entry_rev = self._get_file_revision(revid, path)
427
self._set_entry_revision_in_db(
428
revid, path, git_id, entry.executable, entry_rev)
429
#self.cachedb.commit()
430
entry.revision = entry_rev
433
def escape_file_id(file_id):
434
return file_id.replace('_', '__').replace(' ', '_s')
437
def escape_for_xml(message):
438
"""Replace xml-incompatible control characters."""
439
# Copied from _escape_commit_message from bzr-svn.
440
# -- David Allouche 2007-12-29.
444
# FIXME: RBC 20060419 this should be done by the revision
445
# serialiser not by commit. Then we can also add an unescaper
446
# in the deserializer and start roundtripping revision messages
447
# precisely. See repository_implementations/test_repository.py
449
# Python strings can include characters that can't be
450
# represented in well-formed XML; escape characters that
451
# aren't listed in the XML specification
452
# (http://www.w3.org/TR/REC-xml/#NT-Char).
453
message, _ = re.subn(
454
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
455
lambda match: match.group(0).encode('unicode_escape'),
460
class GitRevisionTree(revisiontree.RevisionTree):
462
def __init__(self, repository, revision_id):
463
if revision_id is None:
464
revision_id = revision.NULL_REVISION
465
self._inventory = repository.get_inventory(revision_id)
466
self._repository = repository
467
self._revision_id = revision_id
469
def get_file_lines(self, file_id):
470
entry = self._inventory[file_id]
471
if entry.kind == 'directory': return []
472
git_id = self._inventory.git_ids[file_id]
473
if git_id in self._inventory.git_file_data:
474
return self._inventory.git_file_data[git_id]
475
return self._repository._get_blob(git_id)
478
class GitFormat(object):
480
supports_tree_reference = False
added
removed
git_repository.py
