17
17
"""An adapter between a Git Repository and a Bazaar Branch"""
23
23
from bzrlib import (
34
from bzrlib.foreign import (
37
from bzrlib.trace import mutter
38
33
from bzrlib.transport import get_transport
40
from bzrlib.plugins.git.foreign import (
35
from bzrlib.plugins.git import (
43
from bzrlib.plugins.git.mapping import default_mapping, mapping_registry
44
from bzrlib.plugins.git.versionedfiles import GitTexts
49
class GitTags(object):
51
def __init__(self, tags):
55
return iter(self._tags)
58
class GitRepository(ForeignRepository):
44
class GitRepository(repository.Repository):
59
45
"""An adapter to git repositories for bzr."""
63
49
def __init__(self, gitdir, lockfiles):
64
ForeignRepository.__init__(self, GitFormat(), gitdir, lockfiles)
65
from bzrlib.plugins.git import fetch
66
repository.InterRepository.register_optimiser(fetch.InterGitRepository)
67
repository.InterRepository.register_optimiser(fetch.InterGitNonGitRepository)
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
69
94
def is_shared(self):
72
97
def supports_rich_root(self):
75
def _warn_if_deprecated(self):
76
# This class isn't deprecated
79
def get_mapping(self):
80
return default_mapping
82
def make_working_trees(self):
86
class LocalGitRepository(GitRepository):
88
def __init__(self, gitdir, lockfiles):
89
# FIXME: This also caches negatives. Need to be more careful
90
# about this once we start writing to git
91
self._parents_provider = graph.CachingParentsProvider(self)
92
GitRepository.__init__(self, gitdir, lockfiles)
93
self.base = gitdir.root_transport.base
94
self._git = gitdir._git
96
self.signatures = versionedfiles.VirtualSignatureTexts(self)
97
self.revisions = versionedfiles.VirtualRevisionTexts(self)
98
self.inventories = versionedfiles.VirtualInventoryTexts(self)
99
self.texts = GitTexts(self)
100
self.tags = GitTags(self._git.get_tags())
102
def all_revision_ids(self):
103
ret = set([revision.NULL_REVISION])
104
if self._git.heads() == []:
106
bzr_heads = [self.get_mapping().revision_id_foreign_to_bzr(h) for h in self._git.heads()]
108
graph = self.get_graph()
109
for rev, parents in graph.iter_ancestry(bzr_heads):
113
#def get_revision_delta(self, revision_id):
114
# parent_revid = self.get_revision(revision_id).parent_ids[0]
115
# diff = self._git.diff(ids.convert_revision_id_bzr_to_git(parent_revid),
116
# ids.convert_revision_id_bzr_to_git(revision_id))
118
def _make_parents_provider(self):
119
"""See Repository._make_parents_provider()."""
120
return self._parents_provider
122
def get_parent_map(self, revids):
124
mutter("get_parent_map(%r)", revids)
125
for revision_id in revids:
126
assert isinstance(revision_id, str)
127
if revision_id == revision.NULL_REVISION:
128
parent_map[revision_id] = ()
130
hexsha, mapping = self.lookup_git_revid(revision_id)
131
commit = self._git.commit(hexsha)
135
parent_map[revision_id] = [mapping.revision_id_foreign_to_bzr(p) for p in commit.parents]
138
def get_ancestry(self, revision_id, topo_sorted=True):
139
"""See Repository.get_ancestry().
141
if revision_id is None:
142
return self._all_revision_ids()
143
assert isinstance(revision_id, str)
145
graph = self.get_graph()
146
for rev, parents in graph.iter_ancestry([revision_id]):
147
if rev == revision.NULL_REVISION:
153
def import_revision_gist(self, source, revid):
156
def dfetch(self, source, stop_revision):
157
if stop_revision is None:
158
raise NotImplementedError
163
graph = source.get_graph()
164
for revid, parents in graph.iter_ancestry([stop_revision]):
165
if not self.has_revision(revid):
167
pb = ui.ui_factory.nested_progress_bar()
169
for i, revid in enumerate(reversed(todo)):
170
pb.update("pushing revisions", i, len(todo))
171
revidmap[revid] = self.import_revision_gist(source, revid)
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]
178
138
def get_signature_text(self, revision_id):
179
139
raise errors.NoSuchRevision(self, revision_id)
181
def lookup_revision_id(self, revid):
182
"""Lookup a revision id.
184
:param revid: Bazaar revision id.
185
:return: Tuple with git revisionid and mapping.
187
# Yes, this doesn't really work, but good enough as a stub
188
return osutils.sha(rev_id).hexdigest(), self.get_mapping()
190
def has_signature_for_revision_id(self, revision_id):
193
def lookup_git_revid(self, bzr_revid):
195
return mapping_registry.revision_id_bzr_to_foreign(bzr_revid)
196
except errors.InvalidRevisionId:
197
raise errors.NoSuchRevision(self, bzr_revid)
199
142
def get_revision(self, revision_id):
200
git_commit_id, mapping = self.lookup_git_revid(revision_id)
202
commit = self._git.commit(git_commit_id)
204
raise errors.NoSuchRevision(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)
205
147
# print "fetched revision:", git_commit_id
206
revision = mapping.import_commit(commit)
207
assert revision is not None
148
revision = self._parse_rev(raw)
149
self._revision_cache[revision_id] = revision
210
152
def has_revision(self, revision_id):
212
154
self.get_revision(revision_id)
213
except errors.NoSuchRevision:
155
except NoSuchRevision:
218
def get_revisions(self, revids):
219
return [self.get_revision(r) for r in revids]
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)
221
245
def revision_trees(self, revids):
222
246
for revid in revids:
223
247
yield self.revision_tree(revid)
225
249
def revision_tree(self, revision_id):
226
revision_id = revision.ensure_null(revision_id)
228
if revision_id == revision.NULL_REVISION:
229
inv = inventory.Inventory(root_id=None)
230
inv.revision_id = revision_id
231
return revisiontree.RevisionTree(self, inv, revision_id)
233
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)
235
283
def get_inventory(self, revision_id):
236
assert revision_id != None
237
return self.revision_tree(revision_id).inventory
239
def set_make_working_trees(self, trees):
242
def fetch_objects(self, determine_wants, graph_walker, progress=None):
243
return self._git.fetch_objects(determine_wants, graph_walker, progress)
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'),
246
456
class GitRevisionTree(revisiontree.RevisionTree):
248
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)
249
462
self._repository = repository
250
self.revision_id = revision_id
251
assert isinstance(revision_id, str)
252
git_id, self.mapping = repository.lookup_git_revid(revision_id)
254
commit = repository._git.commit(git_id)
256
raise errors.NoSuchRevision(repository, revision_id)
257
self.tree = commit.tree
258
self._inventory = inventory.Inventory(revision_id=revision_id)
259
self._inventory.root.revision = revision_id
260
self._build_inventory(self.tree, self._inventory.root, "")
262
def get_revision_id(self):
263
return self.revision_id
265
def get_file_text(self, file_id):
463
self._revision_id = revision_id
465
def get_file_lines(self, file_id):
266
466
entry = self._inventory[file_id]
267
if entry.kind == 'directory': return ""
268
return self._repository._git.get_blob(entry.text_id).data
270
def _build_inventory(self, tree_id, ie, path):
271
assert isinstance(path, str)
272
tree = self._repository._git.tree(tree_id)
273
for mode, name, hexsha in tree.entries():
274
basename = name.decode("utf-8")
278
child_path = urlutils.join(path, name)
279
file_id = self.mapping.generate_file_id(child_path)
280
entry_kind = (mode & 0700000) / 0100000
282
child_ie = inventory.InventoryDirectory(file_id, basename, ie.file_id)
283
elif entry_kind == 1:
284
file_kind = (mode & 070000) / 010000
285
b = self._repository._git.get_blob(hexsha)
287
child_ie = inventory.InventoryFile(file_id, basename, ie.file_id)
288
child_ie.text_sha1 = osutils.sha_string(b.data)
290
child_ie = inventory.InventoryLink(file_id, basename, ie.file_id)
291
child_ie.text_sha1 = osutils.sha_string("")
293
raise AssertionError(
294
"Unknown file kind, perms=%o." % (mode,))
295
child_ie.text_id = b.id
296
child_ie.text_size = len(b.data)
298
raise AssertionError(
299
"Unknown blob kind, perms=%r." % (mode,))
300
fs_mode = mode & 0777
301
child_ie.executable = bool(fs_mode & 0111)
302
child_ie.revision = self.revision_id
303
self._inventory.add(child_ie)
305
self._build_inventory(hexsha, child_ie, child_path)
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)
308
474
class GitFormat(object):
310
476
supports_tree_reference = False
311
rich_root_data = True
313
def get_format_description(self):
314
return "Git Repository"
316
def initialize(self, url, shared=False, _internal=False):
317
raise bzr_errors.UninitializableFormat(self)
319
def check_conversion_target(self, target_repo_format):
320
return target_repo_format.rich_root_data