126
126
ie.text_size = None
127
127
ie.text_sha1 = None
129
ie.text_size = sum(imap(len, blob.chunked))
130
ie.text_sha1 = osutils.sha_strings(blob.chunked)
129
ie.text_size = len(blob.data)
130
ie.text_sha1 = osutils.sha_string(blob.data)
131
131
# Check what revision we should store
133
133
for pinv in parent_invs:
138
if (pie.text_sha1 == ie.text_sha1 and
139
pie.executable == ie.executable and
140
pie.symlink_target == ie.symlink_target):
134
if pinv.revision_id == base_inv.revision_id:
143
if pie.text_sha1 == ie.text_sha1 and pie.executable == ie.executable and pie.symlink_target == ie.symlink_target:
141
144
# found a revision in one of the parents to use
142
145
ie.revision = pie.revision
144
parent_key = (file_id, pie.revision)
145
if not parent_key in parent_keys:
146
parent_keys.append(parent_key)
147
parent_keys.append((file_id, pie.revision))
147
148
if ie.revision is None:
148
149
# Need to store a new revision
149
150
ie.revision = revision_id
151
assert file_id is not None
150
152
assert ie.revision is not None
151
153
if ie.kind == 'symlink':
154
chunks = blob.chunked
155
texts.insert_record_stream([
156
ChunkedContentFactory((file_id, ie.revision),
157
tuple(parent_keys), ie.text_sha1, chunks)])
157
texts.insert_record_stream([FulltextContentFactory((file_id, ie.revision), tuple(parent_keys), ie.text_sha1, data)])
158
shamap = [(hexsha, "blob", (ie.file_id, ie.revision))]
159
if base_hexsha is not None:
160
old_path = path # Renames are not supported yet
161
if stat.S_ISDIR(base_mode):
162
invdelta.extend(remove_disappeared_children(base_inv, old_path,
163
lookup_object(base_hexsha), [], lookup_object))
162
if base_ie is not None:
163
old_path = base_inv.id2path(file_id)
164
if base_ie.kind == "directory":
165
invdelta.extend(remove_disappeared_children(old_path, base_ie.children, []))
166
168
invdelta.append((old_path, path, file_id, ie))
167
if base_hexsha != hexsha:
168
store_updater.add_object(blob, ie)
169
return (invdelta, shamap)
172
172
class SubmodulesRequireSubtrees(BzrError):
177
def import_git_submodule(texts, mapping, path, name, (base_hexsha, hexsha),
178
base_inv, parent_id, revision_id, parent_invs, lookup_object,
179
(base_mode, mode), store_updater, lookup_file_id):
180
if base_hexsha == hexsha and base_mode == mode:
182
file_id = lookup_file_id(path)
183
ie = TreeReference(file_id, name.decode("utf-8"), parent_id)
177
def import_git_submodule(texts, mapping, path, hexsha, base_inv, base_ie,
178
parent_id, revision_id, parent_invs, shagitmap, lookup_object):
179
file_id = mapping.generate_file_id(path)
180
ie = TreeReference(file_id, urlutils.basename(path.decode("utf-8")),
184
182
ie.revision = revision_id
185
if base_hexsha is None:
187
if (base_ie.kind == ie.kind and
188
base_ie.reference_revision == ie.reference_revision):
189
ie.revision = base_ie.revision
189
190
ie.reference_revision = mapping.revision_id_foreign_to_bzr(hexsha)
190
texts.insert_record_stream([
191
ChunkedContentFactory((file_id, ie.revision), (), None, [])])
191
texts.insert_record_stream([FulltextContentFactory((file_id, ie.revision), (), None, "")])
192
192
invdelta = [(oldpath, path, file_id, ie)]
196
def remove_disappeared_children(base_inv, path, base_tree, existing_children,
193
return invdelta, {}, {}
196
def remove_disappeared_children(path, base_children, existing_children):
199
for name, mode, hexsha in base_tree.iteritems():
200
if name in existing_children:
202
c_path = posixpath.join(path, name.decode("utf-8"))
203
ret.append((c_path, None, base_inv.path2id(c_path), None))
204
if stat.S_ISDIR(mode):
205
ret.extend(remove_disappeared_children(
206
base_inv, c_path, lookup_object(hexsha), [], lookup_object))
198
deletable = [(osutils.pathjoin(path, k), v) for k,v in base_children.iteritems() if k not in existing_children]
200
(path, ie) = deletable.pop()
201
ret.append((path, None, ie.file_id, None))
202
if ie.kind == "directory":
203
for name, child_ie in ie.children.iteritems():
204
deletable.append((osutils.pathjoin(path, name), child_ie))
210
def import_git_tree(texts, mapping, path, name, (base_hexsha, hexsha),
211
base_inv, parent_id, revision_id, parent_invs,
212
lookup_object, (base_mode, mode), store_updater,
213
lookup_file_id, allow_submodules=False):
208
def import_git_tree(texts, mapping, path, hexsha, base_inv, base_ie, parent_id,
209
revision_id, parent_invs, shagitmap, lookup_object, allow_submodules=False):
214
210
"""Import a git tree object into a bzr repository.
216
212
:param texts: VersionedFiles object to add to
219
215
:param base_inv: Base inventory against which to return inventory delta
220
216
:return: Inventory delta for this subtree
222
if base_hexsha == hexsha and base_mode == mode:
223
# If nothing has changed since the base revision, we're done
226
file_id = lookup_file_id(path)
219
file_id = mapping.generate_file_id(path)
227
220
# We just have to hope this is indeed utf-8:
228
ie = InventoryDirectory(file_id, name.decode("utf-8"), parent_id)
229
tree = lookup_object(hexsha)
230
if base_hexsha is None:
232
old_path = None # Newly appeared here
234
base_tree = lookup_object(base_hexsha)
235
old_path = path # Renames aren't supported yet
236
if base_tree is None or type(base_tree) is not Tree:
221
ie = InventoryDirectory(file_id, urlutils.basename(path.decode("utf-8")),
224
# Newly appeared here
237
225
ie.revision = revision_id
238
invdelta.append((old_path, path, ie.file_id, ie))
239
texts.insert_record_stream([
240
ChunkedContentFactory((ie.file_id, ie.revision), (), None, [])])
226
texts.insert_record_stream([FulltextContentFactory((file_id, ie.revision), (), None, "")])
227
invdelta.append((None, path, file_id, ie))
229
# See if this has changed at all
231
base_sha = shagitmap.lookup_tree(file_id, base_inv.revision_id)
235
if base_sha == hexsha:
236
# If nothing has changed since the base revision, we're done
238
if base_ie.kind != "directory":
239
ie.revision = revision_id
240
texts.insert_record_stream([FulltextContentFactory((ie.file_id, ie.revision), (), None, "")])
241
invdelta.append((base_inv.id2path(ie.file_id), path, ie.file_id, ie))
242
if base_ie is not None and base_ie.kind == "directory":
243
base_children = base_ie.children
241
246
# Remember for next time
242
247
existing_children = set()
244
for child_mode, name, child_hexsha in tree.entries():
245
existing_children.add(name)
246
child_path = posixpath.join(path, name)
247
if type(base_tree) is Tree:
249
child_base_mode, child_base_hexsha = base_tree[name]
251
child_base_hexsha = None
254
child_base_hexsha = None
256
if stat.S_ISDIR(child_mode):
257
subinvdelta, grandchildmodes = import_git_tree(texts, mapping,
258
child_path, name, (child_base_hexsha, child_hexsha), base_inv,
259
file_id, revision_id, parent_invs, lookup_object,
260
(child_base_mode, child_mode), store_updater, lookup_file_id,
261
allow_submodules=allow_submodules)
262
elif S_ISGITLINK(child_mode): # submodule
250
tree = lookup_object(hexsha)
251
for mode, name, child_hexsha in tree.entries():
252
basename = name.decode("utf-8")
253
existing_children.add(basename)
254
child_path = osutils.pathjoin(path, name)
255
if stat.S_ISDIR(mode):
256
subinvdelta, grandchildmodes, subshamap = import_git_tree(
257
texts, mapping, child_path, child_hexsha, base_inv,
258
base_children.get(basename), file_id, revision_id,
259
parent_invs, shagitmap, lookup_object,
260
allow_submodules=allow_submodules)
261
invdelta.extend(subinvdelta)
262
child_modes.update(grandchildmodes)
263
shamap.extend(subshamap)
264
elif S_ISGITLINK(mode): # submodule
263
265
if not allow_submodules:
264
266
raise SubmodulesRequireSubtrees()
265
subinvdelta, grandchildmodes = import_git_submodule(texts, mapping,
266
child_path, name, (child_base_hexsha, child_hexsha), base_inv,
267
file_id, revision_id, parent_invs, lookup_object,
268
(child_base_mode, child_mode), store_updater, lookup_file_id)
267
subinvdelta, grandchildmodes, subshamap = import_git_submodule(
268
texts, mapping, child_path, child_hexsha, base_inv, base_children.get(basename),
269
file_id, revision_id, parent_invs, shagitmap, lookup_object)
270
invdelta.extend(subinvdelta)
271
child_modes.update(grandchildmodes)
272
shamap.extend(subshamap)
270
subinvdelta = import_git_blob(texts, mapping, child_path, name,
271
(child_base_hexsha, child_hexsha), base_inv, file_id,
272
revision_id, parent_invs, lookup_object,
273
(child_base_mode, child_mode), store_updater, lookup_file_id)
275
child_modes.update(grandchildmodes)
276
invdelta.extend(subinvdelta)
277
if child_mode not in (stat.S_IFDIR, DEFAULT_FILE_MODE,
274
subinvdelta, subshamap = import_git_blob(texts, mapping,
275
child_path, child_hexsha, base_inv, base_children.get(basename), file_id,
276
revision_id, parent_invs, shagitmap, lookup_object,
277
mode_is_executable(mode), stat.S_ISLNK(mode))
278
invdelta.extend(subinvdelta)
279
shamap.extend(subshamap)
280
if mode not in (stat.S_IFDIR, DEFAULT_FILE_MODE,
278
281
stat.S_IFLNK, DEFAULT_FILE_MODE|0111):
279
child_modes[child_path] = child_mode
282
child_modes[child_path] = mode
280
283
# Remove any children that have disappeared
281
if base_tree is not None and type(base_tree) is Tree:
282
invdelta.extend(remove_disappeared_children(base_inv, old_path,
283
base_tree, existing_children, lookup_object))
284
store_updater.add_object(tree, ie)
285
return invdelta, child_modes
288
def verify_commit_reconstruction(target_git_object_retriever, lookup_object,
289
o, rev, ret_tree, parent_trees, mapping, unusual_modes):
290
new_unusual_modes = mapping.export_unusual_file_modes(rev)
291
if new_unusual_modes != unusual_modes:
292
raise AssertionError("unusual modes don't match: %r != %r" % (
293
unusual_modes, new_unusual_modes))
294
# Verify that we can reconstruct the commit properly
295
rec_o = target_git_object_retriever._reconstruct_commit(rev, o.tree)
297
raise AssertionError("Reconstructed commit differs: %r != %r" % (
301
for path, obj, ie in _tree_to_objects(ret_tree, parent_trees,
302
target_git_object_retriever._cache.idmap, unusual_modes, mapping.BZR_DUMMY_FILE):
303
old_obj_id = tree_lookup_path(lookup_object, o.tree, path)[1]
305
if obj.id != old_obj_id:
306
diff.append((path, lookup_object(old_obj_id), obj))
307
for (path, old_obj, new_obj) in diff:
308
while (old_obj.type_name == "tree" and
309
new_obj.type_name == "tree" and
310
sorted(old_obj) == sorted(new_obj)):
312
if old_obj[name][0] != new_obj[name][0]:
313
raise AssertionError("Modes for %s differ: %o != %o" %
314
(path, old_obj[name][0], new_obj[name][0]))
315
if old_obj[name][1] != new_obj[name][1]:
316
# Found a differing child, delve deeper
317
path = posixpath.join(path, name)
318
old_obj = lookup_object(old_obj[name][1])
319
new_obj = new_objs[path]
321
raise AssertionError("objects differ for %s: %r != %r" % (path,
284
if base_ie is not None and base_ie.kind == "directory":
285
invdelta.extend(remove_disappeared_children(base_inv.id2path(file_id),
286
base_children, existing_children))
287
shamap.append((hexsha, "tree", (file_id, revision_id)))
288
return invdelta, child_modes, shamap
291
def approx_inv_size(inv):
292
# Very rough estimate, 1k per inventory entry
293
return len(inv) * 1024
325
296
def import_git_commit(repo, mapping, head, lookup_object,
326
target_git_object_retriever, trees_cache):
297
target_git_object_retriever, parent_invs_cache):
327
298
o = lookup_object(head)
328
299
rev = mapping.import_commit(o)
329
300
# We have to do this here, since we have to walk the tree and
330
301
# we need to make sure to import the blobs / trees with the right
331
302
# path; this may involve adding them more than once.
332
parent_trees = trees_cache.revision_trees(rev.parent_ids)
333
if parent_trees == []:
304
for parent_id in rev.parent_ids:
306
parent_invs.append(parent_invs_cache[parent_id])
308
parent_inv = repo.get_inventory(parent_id)
309
parent_invs.append(parent_inv)
310
parent_invs_cache[parent_id] = parent_inv
311
if parent_invs == []:
334
312
base_inv = Inventory(root_id=None)
338
base_inv = parent_trees[0].inventory
339
base_tree = lookup_object(o.parents[0]).tree
340
base_mode = stat.S_IFDIR
341
store_updater = target_git_object_retriever._get_updater(rev)
342
store_updater.add_object(o, None)
343
root_tree = lookup_object(o.tree)
345
file_id_map_sha = root_tree[mapping.BZR_FILE_IDS_FILE][1]
349
file_ids = mapping.import_fileid_map(lookup_object(file_id_map_sha))
350
lookup_file_id = GitFileIdMap(file_ids, mapping).lookup_file_id
351
inv_delta, unusual_modes = import_git_tree(repo.texts,
352
mapping, "", u"", (base_tree, o.tree), base_inv,
353
None, rev.revision_id, [p.inventory for p in parent_trees],
354
lookup_object, (base_mode, stat.S_IFDIR), store_updater,
315
base_inv = parent_invs[0]
316
base_ie = base_inv.root
317
inv_delta, unusual_modes, shamap = import_git_tree(repo.texts,
318
mapping, "", o.tree, base_inv, base_ie, None, rev.revision_id,
319
parent_invs, target_git_object_retriever._idmap, lookup_object,
356
320
allow_submodules=getattr(repo._format, "supports_tree_reference", False))
357
store_updater.finish()
321
target_git_object_retriever._idmap.add_entries(shamap)
358
322
if unusual_modes != {}:
359
323
for path, mode in unusual_modes.iteritems():
360
324
warn_unusual_mode(rev.foreign_revid, path, mode)
367
331
rev.inventory_sha1, inv = repo.add_inventory_by_delta(basis_id,
368
332
inv_delta, rev.revision_id, rev.parent_ids,
370
ret_tree = RevisionTree(repo, inv, rev.revision_id)
371
trees_cache.add(ret_tree)
334
parent_invs_cache[rev.revision_id] = inv
372
335
repo.add_revision(rev.revision_id, rev)
373
336
if "verify" in debug.debug_flags:
374
verify_commit_reconstruction(target_git_object_retriever,
375
lookup_object, o, rev, ret_tree, parent_trees, mapping,
379
def import_git_objects(repo, mapping, object_iter,
380
target_git_object_retriever, heads, pb=None, limit=None):
337
new_unusual_modes = mapping.export_unusual_file_modes(rev)
338
if new_unusual_modes != unusual_modes:
339
raise AssertionError("unusual modes don't match: %r != %r" % (unusual_modes, new_unusual_modes))
340
objs = inventory_to_tree_and_blobs(inv, repo.texts, mapping, unusual_modes)
341
for sha1, newobj, path in objs:
342
assert path is not None
343
oldobj = tree_lookup_path(lookup_object, o.tree, path)
345
raise AssertionError("%r != %r in %s" % (oldobj, newobj, path))
348
def import_git_objects(repo, mapping, object_iter, target_git_object_retriever,
349
heads, pb=None, limit=None):
381
350
"""Import a set of git objects into a bzr repository.
383
352
:param repo: Target Bazaar repository
384
353
:param mapping: Mapping to use
385
354
:param object_iter: Iterator over Git objects.
386
:return: Tuple with pack hints and last imported revision id
356
target_git_object_retriever._idmap.start_write_group() # FIXME: try/finally
388
357
def lookup_object(sha):
390
359
return object_iter[sha]
392
361
return target_git_object_retriever[sha]
362
# TODO: a more (memory-)efficient implementation of this
395
heads = list(set(heads))
396
trees_cache = LRUTreeCache(repo)
366
parent_invs_cache = lru_cache.LRUSizeCache(compute_size=approx_inv_size,
367
max_size=MAX_INV_CACHE_SIZE)
397
368
# Find and convert commit objects
399
370
if pb is not None: