Analysis of heap dumps is a key to understanding of critical memory and performance issues in Java, such as memory leaks. In cloud environment, sanitized heapdumps are typically stored in dedicated servers. For practical and security reasons, a preferred way of accessing them is through a web application such as Eclipse Jifa [1], rather than accessing them remotely via a desktop application such as Eclipse MAT [2].

Heap dump analysis is typically done manually, by navigating through an object graph in a UI. To some extent, the analysis can be automated, by defining common analysis steps programmatically and running them for every captured heap dump and generating overview reports for them. However, this approach has its limits, as new types of problems may emerge for which new steps need to be discovered, experimented with, and defined. For their investigation, one would ideally combine manual UI steps with programmatic steps.

The goal of this thesis is to enhance Eclipse Jifa by allowing to define a custom code for heap dump inspection, leverage GraalVM [3] (including its sandboxing features) for its execution, and integrate it within Eclipse Jifa UI. A custom code will be permitted to invoke functions from the core Eclipse MAT API. As an input, it may either use a heap dump as a whole, or a previously precomputed result (e.g. a set of objects) from Jifa. It will generate an output in a (potentially new) format displayable in Jifa. The enhancements should be delivered as a public fork of Eclipse Jifa GitHub project [4].

References:
[1] https://projects.eclipse.org/projects/technology.jifa
[2] https://eclipse.dev/mat/
[3] https://www.graalvm.org/
[4] https://github.com/eclipse/jifa