Ask and you shall receive: Empirically evaluating declarative approaches to finding data in unstructured heaps

This paper reports on experience with the engineering and empirical evaluation of data management software that stores objects in collections like the ArrayList or Vector. While many programs may retrieve an object from a collection by iteratively evaluating each object according to a set of condition(s), this imperative retrieval process becomes more challenging and error-prone as it applies many complex criteria to find the matching objects in multiple collections. Query languages for unstructured Java virtual machine (JVM) heaps present an alternative to the imperative approach for finding the matching objects. Using a benchmarking framework that measures the performance of declarative approaches to identifying certain objects in the JVM heap, this paper empirically evaluates two query languages, JQL and JoSQL. Both the experiences and the experimental results reveal trade-offs in the performance and overall viability of the query languages and the imperative approaches.