def test_symlink_detect_changes(self):
left = inventory.InventoryLink('123', 'hello.c', ROOT_ID)
left.text_sha1 = 123
left.executable = True
left.symlink_target='foo'
right = inventory.InventoryLink('123', 'hello.c', ROOT_ID)
right.text_sha1 = 321
right.symlink_target='foo'
self.assertEqual((False, False), left.detect_changes(right))
self.assertEqual((False, False), right.detect_changes(left))
left.symlink_target = 'different'
self.assertEqual((True, False), left.detect_changes(right))
self.assertEqual((True, False), right.detect_changes(left))
def _unpack_entry(self, elt):
kind = elt.tag
if not InventoryEntry.versionable_kind(kind):
raise AssertionError('unsupported entry kind %s' % kind)
get_cached = _get_utf8_or_ascii
parent_id = elt.get('parent_id')
if parent_id is not None:
parent_id = get_cached(parent_id)
file_id = get_cached(elt.get('file_id'))
if kind == 'directory':
ie = inventory.InventoryDirectory(file_id, elt.get('name'),
parent_id)
elif kind == 'file':
ie = inventory.InventoryFile(file_id, elt.get('name'), parent_id)
ie.text_sha1 = elt.get('text_sha1')
if elt.get('executable') == 'yes':
ie.executable = True
v = elt.get('text_size')
ie.text_size = v and int(v)
elif kind == 'symlink':
ie = inventory.InventoryLink(file_id, elt.get('name'), parent_id)
ie.symlink_target = elt.get('symlink_target')
else:
raise errors.UnsupportedInventoryKind(kind)
revision = elt.get('revision')
if revision is not None:
revision = get_cached(revision)
ie.revision = revision
return ie
def _unpack_entry(self, elt):
## original format inventories don't have a parent_id for
## nodes in the root directory, but it's cleaner to use one
## internally.
parent_id = elt.get('parent_id')
if parent_id is None:
parent_id = ROOT_ID
kind = elt.get('kind')
if kind == 'directory':
ie = inventory.InventoryDirectory(elt.get('file_id'),
elt.get('name'), parent_id)
elif kind == 'file':
ie = inventory.InventoryFile(elt.get('file_id'), elt.get('name'),
parent_id)
ie.text_id = elt.get('text_id')
ie.text_sha1 = elt.get('text_sha1')
v = elt.get('text_size')
ie.text_size = v and int(v)
elif kind == 'symlink':
ie = inventory.InventoryLink(elt.get('file_id'), elt.get('name'),
parent_id)
ie.symlink_target = elt.get('symlink_target')
else:
raise BzrError("unknown kind %r" % kind)
## mutter("read inventoryentry: %r", elt.attrib)
return ie
def get_sample_inventory(self):
inv = Inventory('tree-root-321', revision_id='rev_outer')
inv.add(inventory.InventoryFile('file-id', 'file', 'tree-root-321'))
inv.add(inventory.InventoryDirectory('dir-id', 'dir', 'tree-root-321'))
inv.add(inventory.InventoryLink('link-id', 'link', 'tree-root-321'))
inv['tree-root-321'].revision = 'rev_outer'
inv['dir-id'].revision = 'rev_outer'
inv['file-id'].revision = 'rev_outer'
inv['file-id'].text_sha1 = 'A'
inv['file-id'].text_size = 1
inv['link-id'].revision = 'rev_outer'
inv['link-id'].symlink_target = 'a'
return inv
def _build_inventory(self, tree_id, ie, path):
assert isinstance(path, str)
tree = self._repository._git.tree(tree_id)
for mode, name, hexsha in tree.entries():
basename = name.decode("utf-8")
if path == "":
child_path = name
else:
child_path = urlutils.join(path, name)
file_id = self.mapping.generate_file_id(child_path)
entry_kind = (mode & 0700000) / 0100000
if entry_kind == 0:
child_ie = inventory.InventoryDirectory(
file_id, basename, ie.file_id)
elif entry_kind == 1:
file_kind = (mode & 070000) / 010000
b = self._repository._git.get_blob(hexsha)
if file_kind == 0:
child_ie = inventory.InventoryFile(file_id, basename,
ie.file_id)
child_ie.text_sha1 = osutils.sha_string(b.data)
elif file_kind == 2:
child_ie = inventory.InventoryLink(file_id, basename,
ie.file_id)
child_ie.text_sha1 = osutils.sha_string("")
else:
raise AssertionError("Unknown file kind, perms=%o." %
(mode, ))
child_ie.text_id = b.id
child_ie.text_size = len(b.data)
else:
raise AssertionError("Unknown blob kind, perms=%r." % (mode, ))
fs_mode = mode & 0777
child_ie.executable = bool(fs_mode & 0111)
child_ie.revision = self.revision_id
self._inventory.add(child_ie)
if entry_kind == 0:
self._build_inventory(hexsha, child_ie, child_path)
def test_link_has_text(self):
link = inventory.InventoryLink('123', 'hello.c', ROOT_ID)
self.failIf(link.has_text())
def test_link_kind_character(self):
dir = inventory.InventoryLink('123', 'hello.c', ROOT_ID)
self.assertEqual(dir.kind_character(), '')
def unpack_inventory_entry(elt, entry_cache=None, return_from_cache=False):
elt_get = elt.get
file_id = elt_get('file_id')
revision = elt_get('revision')
# Check and see if we have already unpacked this exact entry
# Some timings for "repo.revision_trees(last_100_revs)"
# bzr mysql
# unmodified 4.1s 40.8s
# using lru 3.5s
# using fifo 2.83s 29.1s
# lru._cache 2.8s
# dict 2.75s 26.8s
# inv.add 2.5s 26.0s
# no_copy 2.00s 20.5s
# no_c,dict 1.95s 18.0s
# Note that a cache of 10k nodes is more than sufficient to hold all of
# the inventory for the last 100 revs for bzr, but not for mysql (20k
# is enough for mysql, which saves the same 2s as using a dict)
# Breakdown of mysql using time.clock()
# 4.1s 2 calls to element.get for file_id, revision_id
# 4.5s cache_hit lookup
# 7.1s InventoryFile.copy()
# 2.4s InventoryDirectory.copy()
# 0.4s decoding unique entries
# 1.6s decoding entries after FIFO fills up
# 0.8s Adding nodes to FIFO (including flushes)
# 0.1s cache miss lookups
# Using an LRU cache
# 4.1s 2 calls to element.get for file_id, revision_id
# 9.9s cache_hit lookup
# 10.8s InventoryEntry.copy()
# 0.3s cache miss lookus
# 1.2s decoding entries
# 1.0s adding nodes to LRU
if entry_cache is not None and revision is not None:
key = (file_id, revision)
try:
# We copy it, because some operations may mutate it
cached_ie = entry_cache[key]
except KeyError:
pass
else:
# Only copying directory entries drops us 2.85s => 2.35s
if return_from_cache:
if cached_ie.kind == 'directory':
return cached_ie.copy()
return cached_ie
return cached_ie.copy()
kind = elt.tag
if not inventory.InventoryEntry.versionable_kind(kind):
raise AssertionError('unsupported entry kind %s' % kind)
file_id = get_utf8_or_ascii(file_id)
if revision is not None:
revision = get_utf8_or_ascii(revision)
parent_id = elt_get('parent_id')
if parent_id is not None:
parent_id = get_utf8_or_ascii(parent_id)
if kind == 'directory':
ie = inventory.InventoryDirectory(file_id, elt_get('name'), parent_id)
elif kind == 'file':
ie = inventory.InventoryFile(file_id, elt_get('name'), parent_id)
ie.text_sha1 = elt_get('text_sha1')
if elt_get('executable') == 'yes':
ie.executable = True
v = elt_get('text_size')
ie.text_size = v and int(v)
elif kind == 'symlink':
ie = inventory.InventoryLink(file_id, elt_get('name'), parent_id)
ie.symlink_target = elt_get('symlink_target')
elif kind == 'tree-reference':
file_id = elt.attrib['file_id']
name = elt.attrib['name']
parent_id = elt.attrib['parent_id']
revision = elt.get('revision')
reference_revision = elt.get('reference_revision')
ie = inventory.TreeReference(file_id, name, parent_id, revision,
reference_revision)
else:
raise errors.UnsupportedInventoryKind(kind)
ie.revision = revision
if revision is not None and entry_cache is not None:
# We cache a copy() because callers like to mutate objects, and
# that would cause the item in cache to mutate as well.
# This has a small effect on many-inventory performance, because
# the majority fraction is spent in cache hits, not misses.
entry_cache[key] = ie.copy()
return ie
