Domain-specific Abstractions for Algorithmic Graph Processing
Master Project of Jochem Broekhoff

The PGX Algorithm DSL provides high-level abstractions for algorithmic graph processing, with a particular focus on parallel processing of vertices/edges. These abstractions allow users to express a class of graph algorithms concisely. PGX Algorithm also integrates general-purpose data structures such as collections (sets, stacks, queues), maps, priority queues, which are required to express path-finding and community detection algorithms. These algorithms are typically less concise and the essence of the algorithm is cluttered with low-level code to manipulate these general-purpose data structures. The goal of this thesis is to extend PGX Algorithm with high-level abstractions for path-finding and community detection algorithms, which hide the accidental complexity of general-purpose data structures. The new abstractions are mapped to different graph processing runtimes.