Graph database systems (GDBS) have become popular for representing real-world entities and their relationships, and offering convenient query languages based on graph pattern matching. As graphs increase in size and complexity, GDBS need to provide the appropriate support for abstraction for which views have demonstrated to be an effective tool, facilitating query writing and improving query execution time via materialization techniques. This paper explores how views can be defined and used in GDBS. We propose view-based extensions to the widely used graph query language Cypher, explore a wide range of possible view types, and outline several implementation strategies for view materialization. Using a set of micro- and macro-benchmarks, we provide insight into how expressive different view types are and how effective the proposed implementation strategies are for different GDBS. Our results show that views can be a powerful tool for GDBS, offering great flexibility in query expression and providing performance improvements if materialized.