src/rdflib/void.py
1from __future__ import annotations
2
3import collections
4from typing import DefaultDict, Dict, Optional, Set
5
6from rdflib.graph import Graph, _ObjectType, _PredicateType, _SubjectType
7from rdflib.namespace import RDF, VOID
8from rdflib.term import IdentifiedNode, Literal, URIRef
9
10
11def generateVoID( # noqa: N802
12 g: Graph,
13 dataset: Optional[IdentifiedNode] = None,
14 res: Optional[Graph] = None,
15 distinctForPartitions: bool = True, # noqa: N803
16):
17 """Returns a new graph with a VoID description of the passed dataset
18
19 For more info on Vocabulary of Interlinked Datasets (VoID), see:
20 http://vocab.deri.ie/void
21
22 This only makes two passes through the triples (once to detect the types
23 of things)
24
25 The tradeoff is that lots of temporary structures are built up in memory
26 meaning lots of memory may be consumed :)
27 I imagine at least a few copies of your original graph.
28
29 the distinctForPartitions parameter controls whether
30 distinctSubjects/objects are tracked for each class/propertyPartition
31 this requires more memory again
32 """
33
34 typeMap: Dict[_SubjectType, Set[_SubjectType]] = ( # noqa: N806
35 collections.defaultdict(set)
36 )
37 classes: Dict[_ObjectType, Set[_SubjectType]] = collections.defaultdict(set)
38 for e, c in g.subject_objects(RDF.type):
39 classes[c].add(e)
40 typeMap[e].add(c)
41
42 triples = 0
43 subjects: Set[_SubjectType] = set()
44 objects: Set[_ObjectType] = set()
45 properties: Set[_PredicateType] = set()
46 classCount: DefaultDict[_SubjectType, int] = collections.defaultdict( # noqa: N806
47 int
48 )
49 propCount: DefaultDict[_PredicateType, int] = collections.defaultdict( # noqa: N806
50 int
51 )
52
53 classProps = collections.defaultdict(set) # noqa: N806
54 classObjects = collections.defaultdict(set) # noqa: N806
55 propSubjects = collections.defaultdict(set) # noqa: N806
56 propObjects = collections.defaultdict(set) # noqa: N806
57
58 for s, p, o in g:
59 triples += 1
60 subjects.add(s)
61 properties.add(p)
62 objects.add(o)
63
64 # class partitions
65 if s in typeMap:
66 for c in typeMap[s]:
67 classCount[c] += 1
68 if distinctForPartitions:
69 classObjects[c].add(o)
70 classProps[c].add(p)
71
72 # property partitions
73 propCount[p] += 1
74 if distinctForPartitions:
75 propObjects[p].add(o)
76 propSubjects[p].add(s)
77
78 if not dataset:
79 dataset = URIRef("http://example.org/Dataset")
80
81 if not res:
82 res = Graph()
83
84 res.add((dataset, RDF.type, VOID.Dataset))
85
86 # basic stats
87 res.add((dataset, VOID.triples, Literal(triples)))
88 res.add((dataset, VOID.classes, Literal(len(classes))))
89
90 res.add((dataset, VOID.distinctObjects, Literal(len(objects))))
91 res.add((dataset, VOID.distinctSubjects, Literal(len(subjects))))
92 res.add((dataset, VOID.properties, Literal(len(properties))))
93
94 for i, c in enumerate(classes):
95 part = URIRef(dataset + "_class%d" % i)
96 res.add((dataset, VOID.classPartition, part))
97 res.add((part, RDF.type, VOID.Dataset))
98
99 res.add((part, VOID.triples, Literal(classCount[c])))
100 res.add((part, VOID.classes, Literal(1)))
101
102 res.add((part, VOID["class"], c))
103
104 res.add((part, VOID.entities, Literal(len(classes[c]))))
105 res.add((part, VOID.distinctSubjects, Literal(len(classes[c]))))
106
107 if distinctForPartitions:
108 res.add((part, VOID.properties, Literal(len(classProps[c]))))
109 res.add((part, VOID.distinctObjects, Literal(len(classObjects[c]))))
110
111 for i, p in enumerate(properties):
112 part = URIRef(dataset + "_property%d" % i)
113 res.add((dataset, VOID.propertyPartition, part))
114 res.add((part, RDF.type, VOID.Dataset))
115
116 res.add((part, VOID.triples, Literal(propCount[p])))
117 res.add((part, VOID.properties, Literal(1)))
118
119 res.add((part, VOID.property, p))
120
121 if distinctForPartitions:
122 entities = 0
123 propClasses = set() # noqa: N806
124 for s in propSubjects[p]:
125 if s in typeMap:
126 entities += 1
127 for c in typeMap[s]:
128 propClasses.add(c)
129
130 res.add((part, VOID.entities, Literal(entities)))
131 res.add((part, VOID.classes, Literal(len(propClasses))))
132
133 res.add((part, VOID.distinctSubjects, Literal(len(propSubjects[p]))))
134 res.add((part, VOID.distinctObjects, Literal(len(propObjects[p]))))
135
136 return res, dataset