def calculate_flat_provenance_types(
prov_doc: ProvDocument,
to_level: int = 0,
including_primitives_types: bool = True,
counting_wdf_as_two: bool = False,
ignored_types: Iterable[str] = ϕ,
) -> MultiLevelTypeDict:
# flatten all the bundles, if any
prov_doc = prov_doc.flattened()
# initialise index structures
level0_types = defaultdict(
set) # type: Dict[QualifiedName, Set[QualifiedName]]
predecessors = defaultdict(
set
) # type: Dict[QualifiedName, Set[Tuple[QualifiedName, QualifiedName]]]
types_to_ignore: FrozenSet[str] = frozenset(ignored_types)
# indexing node types and relations
for rec in prov_doc.get_records(): # type: ProvRecord
if rec.is_element():
level0_types[rec.identifier] |= get_element_types(
rec, including_primitives_types, types_to_ignore)
elif rec.is_relation():
rel_type = rec.get_type()
attrs, values = zip(*rec.formal_attributes)
# expecting a QualifiedName from the first argument of a relation
predecessor, successor = values[:2]
if predecessor is not None and successor is not None:
predecessors[successor].add((rel_type, predecessor))
# the type map for this graph
fp_types = defaultdict(dict) # type: MultiLevelTypeDict
# converting type sets to FlatProvenanceType level 0
fp_types[0] = {
node: (frozenset(level0_types[node]), )
for node in level0_types
}
# propagating level-0 types to the specified level
for k in range(1, to_level + 1):
# only propagating (k-1) types from nodes that have them
for node, types in fp_types[k - 1].items():
# propagating the types to the predecessors
for rel_type, predecessor in predecessors[node]:
k_type = types + (frozenset({rel_type}),
) # type: FlatProvenanceType
if counting_wdf_as_two and (rel_type == PROV_DERIVATION):
k_p1_type = k_type + (frozenset({rel_type}),
) # type: FlatProvenanceType
fp_types[k + 1][predecessor] = (
join_flat_types(fp_types[k +
1][predecessor], k_p1_type)
if predecessor in fp_types[k + 1] else k_p1_type)
else:
fp_types[k][predecessor] = (join_flat_types(
fp_types[k][predecessor], k_type) if predecessor
in fp_types[k] else k_type)
return fp_types
def declare_directory(self, value: CWLObjectType) -> ProvEntity:
"""Register any nested files/directories."""
# FIXME: Calculate a hash-like identifier for directory
# so we get same value if it's the same filenames/hashes
# in a different location.
# For now, mint a new UUID to identify this directory, but
# attempt to keep it inside the value dictionary
dir_id = cast(str, value.setdefault("@id", uuid.uuid4().urn))
# New annotation file to keep the ORE Folder listing
ore_doc_fn = dir_id.replace("urn:uuid:", "directory-") + ".ttl"
dir_bundle = self.document.bundle(self.metadata_ns[ore_doc_fn])
coll = self.document.entity(
dir_id,
[
(PROV_TYPE, WFPROV["Artifact"]),
(PROV_TYPE, PROV["Collection"]),
(PROV_TYPE, PROV["Dictionary"]),
(PROV_TYPE, RO["Folder"]),
],
)
# ORE description of ro:Folder, saved separately
coll_b = dir_bundle.entity(
dir_id,
[(PROV_TYPE, RO["Folder"]), (PROV_TYPE, ORE["Aggregation"])],
)
self.document.mentionOf(dir_id + "#ore", dir_id, dir_bundle.identifier)
# dir_manifest = dir_bundle.entity(
# dir_bundle.identifier, {PROV["type"]: ORE["ResourceMap"],
# ORE["describes"]: coll_b.identifier})
coll_attribs = [(ORE["isDescribedBy"], dir_bundle.identifier)]
coll_b_attribs = [] # type: List[Tuple[Identifier, ProvEntity]]
# FIXME: .listing might not be populated yet - hopefully
# a later call to this method will sort that
is_empty = True
if "listing" not in value:
get_listing(self.fsaccess, value)
for entry in cast(MutableSequence[CWLObjectType],
value.get("listing", [])):
is_empty = False
# Declare child-artifacts
entity = self.declare_artefact(entry)
self.document.membership(coll, entity)
# Membership relation aka our ORE Proxy
m_id = uuid.uuid4().urn
m_entity = self.document.entity(m_id)
m_b = dir_bundle.entity(m_id)
# PROV-O style Dictionary
# https://www.w3.org/TR/prov-dictionary/#dictionary-ontological-definition
# ..as prov.py do not currently allow PROV-N extensions
# like hadDictionaryMember(..)
m_entity.add_asserted_type(PROV["KeyEntityPair"])
m_entity.add_attributes({
PROV["pairKey"]: entry["basename"],
PROV["pairEntity"]: entity,
})
# As well as a being a
# http://wf4ever.github.io/ro/2016-01-28/ro/#FolderEntry
m_b.add_asserted_type(RO["FolderEntry"])
m_b.add_asserted_type(ORE["Proxy"])
m_b.add_attributes({
RO["entryName"]: entry["basename"],
ORE["proxyIn"]: coll,
ORE["proxyFor"]: entity,
})
coll_attribs.append((PROV["hadDictionaryMember"], m_entity))
coll_b_attribs.append((ORE["aggregates"], m_b))
coll.add_attributes(coll_attribs)
coll_b.add_attributes(coll_b_attribs)
# Also Save ORE Folder as annotation metadata
ore_doc = ProvDocument()
ore_doc.add_namespace(ORE)
ore_doc.add_namespace(RO)
ore_doc.add_namespace(UUID)
ore_doc.add_bundle(dir_bundle)
ore_doc = ore_doc.flattened()
ore_doc_path = str(PurePosixPath(METADATA, ore_doc_fn))
with self.research_object.write_bag_file(
ore_doc_path) as provenance_file:
ore_doc.serialize(provenance_file,
format="rdf",
rdf_format="turtle")
self.research_object.add_annotation(dir_id, [ore_doc_fn],
ORE["isDescribedBy"].uri)
if is_empty:
# Empty directory
coll.add_asserted_type(PROV["EmptyCollection"])
coll.add_asserted_type(PROV["EmptyDictionary"])
self.research_object.add_uri(coll.identifier.uri)
return coll
