A Modular and Statically Typed Effectful Stack for Custom Graph Traversals

Abstract

Programmers often have to implement custom graph traversals by hand as either there are no suitable text-book algorithms for their tasks, or their problems are too complex for a pure querying language that lacks modularity or static typing. Our new Scala-based graph traversal language uses an effectful stack that adapts monads and type classes. Arbitrary graph effect computations and graph processing rules can be defined and composed in a modular and statically typed way. Custom graph traversals become expressible in a concise notation, run both in-memory and on graph databases, and also allow for parallelization. We evaluate the usability of our approach by detecting occurences of an anti-pattern in a Java source code archive. Our approach outperforms the well-known Gremlin approach due to parallelization.