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"""
33
from bzrlib.transport import get_transport
35
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 = git.repo.Repo(gitdir.root_transport.local_abspath("."))
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]
64
self.signatures = None
66
self._format = GitFormat()
68
def _init_cachedb(self):
69
self.cachedb.executescript("""
70
create table if not exists inventory (
72
create unique index if not exists inventory_revid
74
create table if not exists entry_revision (
80
create unique index if not exists entry_revision_revid_path
81
on entry_revision (inventory, path);
85
def _ancestor_revisions(self, revision_ids):
86
if revision_ids is not None:
87
git_revisions = [gitrevid_from_bzr(r) for r in revision_ids]
90
for lines in self._git.ancestor_lines(git_revisions):
91
yield self._parse_rev(lines)
92
# print "fetched ancestors:", git_revisions
97
def supports_rich_root(self):
100
def get_revision_graph(self, revision_id=None):
102
if revision_id is not None:
103
param = [ids.convert_revision_id_bzr_to_git(revision_id)]
106
git_graph = self._git.get_revision_graph(param)
107
# print "fetched revision graph:", param
108
for node, parents in git_graph.iteritems():
109
bzr_node = ids.convert_revision_id_git_to_bzr(node)
110
bzr_parents = [ids.convert_revision_id_git_to_bzr(n)
112
result[bzr_node] = bzr_parents
115
def get_revision_graph_with_ghosts(self, revision_ids=None):
116
graph = deprecated_graph.Graph()
117
if revision_ids is not None:
118
revision_ids = [ids.convert_revision_id_bzr_to_git(r)
119
for r in revision_ids]
120
git_graph = self._git.get_revision_graph(revision_ids)
121
# print "fetched revision graph (ghosts):", revision_ids
122
for node, parents in git_graph.iteritems():
123
bzr_node = ids.convert_revision_id_git_to_bzr(node)
124
bzr_parents = [ids.convert_revision_id_git_to_bzr(n)
127
graph.add_node(bzr_node, bzr_parents)
130
def get_ancestry(self, revision_id):
131
param = [ids.convert_revision_id_bzr_to_git(revision_id)]
132
git_ancestry = self._git.get_ancestry(param)
133
# print "fetched ancestry:", param
135
ids.convert_revision_id_git_to_bzr(git_id)
136
for git_id in git_ancestry]
138
def get_signature_text(self, revision_id):
139
raise errors.NoSuchRevision(self, revision_id)
142
def get_revision(self, revision_id):
143
if revision_id in self._revision_cache:
144
return self._revision_cache[revision_id]
145
git_commit_id = ids.convert_revision_id_bzr_to_git(revision_id)
146
raw = self._git.rev_list([git_commit_id], max_count=1, header=True)
147
# print "fetched revision:", git_commit_id
148
revision = self._parse_rev(raw)
149
self._revision_cache[revision_id] = revision
152
def has_revision(self, revision_id):
154
self.get_revision(revision_id)
155
except NoSuchRevision:
160
def get_revisions(self, revisions):
161
return [self.get_revision(r) for r in revisions]
164
def _parse_rev(klass, raw):
165
"""Parse a single git revision.
167
* The first line is the git commit id.
168
* Following lines conform to the 'name value' structure, until the
170
* All lines after the first blank line and until the NULL line have 4
171
leading spaces and constitute the commit message.
173
:param raw: sequence of newline-terminated strings, its last item is a
174
single NULL character.
175
:return: a `bzrlib.revision.Revision` object.
180
committer_was_set = False
181
revision_id = ids.convert_revision_id_git_to_bzr(raw[0][:-1])
182
rev = revision.Revision(revision_id)
183
rev.inventory_sha1 = ""
184
assert raw[-1] == '\x00', (
185
"Last item of raw was not a single NULL character.")
186
for line in raw[1:-1]:
188
assert line[:4] == ' ', (
189
"Unexpected line format in commit message: %r" % line)
190
message_lines.append(line[4:])
195
name, value = line[:-1].split(' ', 1)
197
rev.parent_ids.append(
198
ids.convert_revision_id_git_to_bzr(value))
201
author, timestamp, timezone = value.rsplit(' ', 2)
202
rev.properties['author'] = author
203
rev.properties['git-author-timestamp'] = timestamp
204
rev.properties['git-author-timezone'] = timezone
205
if not committer_was_set:
206
rev.committer = author
207
rev.timestamp = float(timestamp)
208
rev.timezone = klass._parse_tz(timezone)
210
if name == 'committer':
211
committer_was_set = True
212
committer, timestamp, timezone = value.rsplit(' ', 2)
213
rev.committer = committer
214
rev.timestamp = float(timestamp)
215
rev.timezone = klass._parse_tz(timezone)
218
rev.properties['git-tree-id'] = value
221
rev.message = ''.join(message_lines)
223
# XXX: That should not be needed, but current revision serializers do
224
# not know how how to handle text that is illegal in xml. Note: when
225
# this is fixed, we will need to rev up the revision namespace when
226
# removing the escaping code. -- David Allouche 2007-12-30
227
rev.message = escape_for_xml(rev.message)
228
rev.committer = escape_for_xml(rev.committer)
229
rev.properties['author'] = escape_for_xml(rev.properties['author'])
234
def _parse_tz(klass, tz):
235
"""Parse a timezone specification in the [+|-]HHMM format.
237
:return: the timezone offset in seconds.
240
sign = {'+': +1, '-': -1}[tz[0]]
242
minutes = int(tz[3:])
243
return sign * 60 * (60 * hours + minutes)
245
def revision_trees(self, revids):
247
yield self.revision_tree(revid)
249
def revision_tree(self, revision_id):
250
return GitRevisionTree(self, revision_id)
252
def _fetch_blob(self, git_id):
253
lines = self._git.cat_file('blob', git_id)
254
# print "fetched blob:", git_id
255
if self._building_inventory is not None:
256
self._building_inventory.git_file_data[git_id] = lines
259
def _get_blob(self, git_id):
261
return self._blob_cache[git_id]
263
return self._fetch_blob(git_id)
265
def _get_blob_caching(self, git_id):
267
return self._blob_cache[git_id]
269
lines = self._fetch_blob(git_id)
270
self._blob_cache[git_id] = lines
273
def _get_blob_info(self, git_id):
275
return self._blob_info_cache[git_id]
277
lines = self._get_blob(git_id)
278
size = sum(len(line) for line in lines)
279
sha1 = osutils.sha_strings(lines)
280
self._blob_info_cache[git_id] = (size, sha1)
283
def get_inventory(self, revision_id):
284
if revision_id is None:
285
revision_id = revision.NULL_REVISION
286
if revision_id == revision.NULL_REVISION:
287
return inventory.Inventory(
288
revision_id=revision_id, root_id=None)
290
# First pass at building the inventory. We need this one to get the
291
# git ids, so we do not have to cache the entire tree text. Ideally,
292
# this should be all we need to do.
293
git_commit = ids.convert_revision_id_bzr_to_git(revision_id)
294
git_inventory = self._git.get_inventory(git_commit)
295
# print "fetched inventory:", git_commit
296
inv = self._parse_inventory(revision_id, git_inventory)
298
# Second pass at building the inventory. There we retrieve additional
299
# data that bzrlib requires: text sizes, sha1s, symlink targets and
300
# revisions that introduced inventory entries
301
self._building_inventory = inv
302
self._building_inventory.git_file_data = {}
303
for file_id in sorted(inv.git_ids.iterkeys()):
304
git_id = inv.git_ids[file_id]
306
self._set_entry_text_info(inv, entry, git_id)
307
for file_id in sorted(inv.git_ids.iterkeys()):
308
git_id = inv.git_ids[file_id]
310
path = inv.id2path(file_id)
311
self._set_entry_revision(entry, revision_id, path, git_id)
313
# At this point the entry_revision table is fully populated for this
314
# revision. So record that we have inventory data for this revision.
315
self.cachedb.execute(
316
"insert or ignore into inventory (revid) values (?)",
318
self.cachedb.commit()
319
self._building_inventory = None
323
def _parse_inventory(klass, revid, git_inv):
324
# For now, git inventory do not have root ids. It is not clear that we
325
# can reliably support root ids. -- David Allouche 2007-12-28
326
inv = inventory.Inventory(revision_id=revid)
328
for perms, git_kind, git_id, path in git_inv:
331
if git_kind == 'blob':
334
executable = bool(int(perms[-3:], 8) & 0111)
335
elif perms[1] == '2':
338
raise AssertionError(
339
"Unknown blob kind, perms=%r." % (perms,))
340
elif git_kind == 'tree':
343
raise AssertionError(
344
"Unknown git entry kind: %r" % (git_kind,))
345
# XXX: Maybe the file id should be prefixed by file kind, so when
346
# the kind of path changes, the id changes too.
347
# -- David Allouche 2007-12-28.
348
file_id = escape_file_id(path.encode('utf-8'))
349
entry = inv.add_path(path, kind, file_id=file_id)
350
entry.executable = executable
351
inv.git_ids[file_id] = git_id
352
inv.root.revision = revid
355
def _set_entry_text_info(self, inv, entry, git_id):
356
if entry.kind == 'directory':
358
size, sha1 = self._get_blob_info(git_id)
359
entry.text_size = size
360
entry.text_sha1 = sha1
361
if entry.kind == 'symlink':
362
lines = self._get_blob_caching(git_id)
363
entry.symlink_target = ''.join(lines)
365
def _get_file_revision(self, revision_id, path):
366
lines = self._git.rev_list(
367
[ids.convert_revision_id_bzr_to_git(revision_id)],
368
max_count=1, topo_order=True, paths=[path])
370
result = ids.convert_revision_id_git_to_bzr(line[:-1])
371
# print "fetched file revision", line[:-1], path
374
def _get_entry_revision_from_db(self, revid, path, git_id, executable):
375
result = self.cachedb.execute(
376
"select revision from entry_revision where"
377
" inventory=? and path=? and gitid=? and executable=?",
378
(revid, path, git_id, executable)).fetchone()
384
def _set_entry_revision_in_db(self, revid, path, git_id, executable, revision):
385
self.cachedb.execute(
386
"insert into entry_revision"
387
" (inventory, path, gitid, executable, revision)"
388
" values (?, ?, ?, ?, ?)",
389
(revid, path, git_id, executable, revision))
391
def _all_inventories_in_db(self, revids):
393
result = self.cachedb.execute(
394
"select count(*) from inventory where revid = ?",
400
def _set_entry_revision(self, entry, revid, path, git_id):
401
# If a revision is in the cache, we assume it contains entries for the
402
# whole inventory. So if all parent revisions are in the cache, but no
403
# parent entry is present, then the entry revision is the current
404
# revision. That amortizes the number of _get_file_revision calls for
405
# large pulls to a "small number".
406
entry_rev = self._get_entry_revision_from_db(
407
revid, path, git_id, entry.executable)
408
if entry_rev is not None:
409
entry.revision = entry_rev
412
revision = self.get_revision(revid)
413
for parent_id in revision.parent_ids:
414
entry_rev = self._get_entry_revision_from_db(
415
parent_id, path, git_id, entry.executable)
416
if entry_rev is not None:
419
if self._all_inventories_in_db(revision.parent_ids):
422
entry_rev = self._get_file_revision(revid, path)
423
self._set_entry_revision_in_db(
424
revid, path, git_id, entry.executable, entry_rev)
425
#self.cachedb.commit()
426
entry.revision = entry_rev
429
def escape_file_id(file_id):
430
return file_id.replace('_', '__').replace(' ', '_s')
433
def escape_for_xml(message):
434
"""Replace xml-incompatible control characters."""
435
# Copied from _escape_commit_message from bzr-svn.
436
# -- David Allouche 2007-12-29.
440
# FIXME: RBC 20060419 this should be done by the revision
441
# serialiser not by commit. Then we can also add an unescaper
442
# in the deserializer and start roundtripping revision messages
443
# precisely. See repository_implementations/test_repository.py
445
# Python strings can include characters that can't be
446
# represented in well-formed XML; escape characters that
447
# aren't listed in the XML specification
448
# (http://www.w3.org/TR/REC-xml/#NT-Char).
449
message, _ = re.subn(
450
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
451
lambda match: match.group(0).encode('unicode_escape'),
456
class GitRevisionTree(revisiontree.RevisionTree):
458
def __init__(self, repository, revision_id):
459
if revision_id is None:
460
revision_id = revision.NULL_REVISION
461
self._inventory = repository.get_inventory(revision_id)
462
self._repository = repository
463
self._revision_id = revision_id
465
def get_file_lines(self, file_id):
466
entry = self._inventory[file_id]
467
if entry.kind == 'directory': return []
468
git_id = self._inventory.git_ids[file_id]
469
if git_id in self._inventory.git_file_data:
470
return self._inventory.git_file_data[git_id]
471
return self._repository._get_blob(git_id)
474
class GitFormat(object):
476
supports_tree_reference = False
added
removed
git_repository.py
