SwarmGuide: Towards Multiple-Query Optimization in Graph Databases

Preliminaries. A graph database G is a finite, directed, edge-labeled, multigraph defined by G = 〈N,Σ,E〉, where N is a finite set of nodes (vertices), Σ is a set of labels, E is a set of directed, labeled edges, and E ⊆ N ×Σ×N . A path p in G is defined as a sequence of n0a0n1 · · · nk−1ak−1nk where ni ∈ N , ai ∈ Σ, and 〈ni, ai, ni+1〉 ∈ E for 0 ≤ i ≤ k. We call the sequence of edge labels Σ∗ of a particular path p the word, ω(p) that p induces. A regular path rp is a path in the graph where ω(rp) is a word in a given regular language L(reg) (e.g., ω(rp) ∈ L(reg)). A regular path query (RPQ) [2] is a triple 〈x, reg, y〉 in which x and y are free variables over the domain of nodes, and reg is a regular expression. An answer of an RPQ is a node-pair 〈s, t〉 (s, t ∈ N) such that there is a path p in G between s and t and ω(p) ∈ L(reg). The answer set of an RPQ is the set of all its answers. The RDF data model and SPARQL query language instantiate these concepts. With the introduction of property paths in SPARQL 1.1, the query language encompasses RPQs. For our examples in this paper, we adopt SPARQL syntax.