Graph Database for Collaborative Communities - Community-Built Databases: Research and Development - page 207

9.2.1.1 Models Based on Simple Node
Data are represented in these models by a (directed or undirected) graph with
simple nodes and edges. Most of these models (GOOD [ 4 ], GMOD [ 5 ], etc.)
represent both schema and instance database as a labeled directed graph. Moreover,
LDM [ 6 ] represents the graph schema as a directed graph where leaves represent
data and whose internal nodes represent connections among the data. LDM
instances consist of two-column tables, one for each node of the schema. Entities
in these model are represented by nodes labeled with type name and also with type
value or object identifier (in the case of instance graph). Some models have nodes
for explicit representation of tuples and sets (PaMaL [ 7 ] and GDM [ 8 ]), and n-ary
relations (GDM). Relations (attributes and relations between entities) are generally
represented in these models by means of labeled edges. LDM and PaMaL use tuple
nodes to describe a set of attributes that are used to define an entity. GOOD defines
edges to distinguish between monovalued (functional edge) and multivalued attri-
butes (nonfunctional edge). Nevertheless, these models do not allow the presenta-
tion of nested relations and are not very suited to modeling complex objects.
9.2.1.2 Models Based on Complex Node
In these models, the basic structure of a graph (node and edge) and the presentation
of entities and relations are based on hypernodes (and hypergraphs). Indeed, a
hypernode is a directed graph in which nodes themselves can be graphs (or
hypernodes) [ 40 ]. Hypernodes can be used to represent simple (flat) and complex
objects (hierarchical, composite, and cyclic) as well as mappings and records. A
hypergraph is a generalized notion of graph where the notion of edge is extended to
hyper edge, which relates to an arbitrary set of nodes. The Hypernode Model and
GGL [ 39 ] emphasize the use of hypernodes for representing nested complex
objects. GROOVY is centered on the use of hypergraphs. The hypernode model
is characterized by using nested graphs at the schema and instance levels. GGL
introduces, in addition to its support for hypernodes (called Master-nodes), the
notion of Master-edge for encapsulation of paths. It uses hypernodes as an abstrac-
tion mechanism consisting in packaging other graphs as an encapsulated vertex,
whereas the Hypernode model additionally uses hypernodes to represent other
abstractions such as complex objects and relations. Most models have explicit
labels on edges. In the hypernode model and GROOVY, labeling can be obtained
by encapsulating edges, which represent the same relation, within one hypernode
(or hyperedge) labeled with the relation name.
9.2.1.3 Discussion
The purpose of this review of graph database models is to find the one most suited to
model many complex data objects and their relationships, such as social networks.