Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.

Non-WSU users: Please talk to your librarian about requesting this dissertation through interlibrary loan.

Access Type

WSU Access

Date of Award

January 2017

Degree Type


Degree Name



Computer Science

First Advisor

Marwan Abi-Antoun


Object graphs help explain the runtime structure of a system. To make object graphs convey design intent, one insight is to use

abstraction by hierarchy, i.e., to show objects that are implementation details as children of architecturally-relevant objects from the application domain. But additional information is needed to express this object hierarchy, using ownership type qualifiers in the code. Adding qualifiers after the fact involves manual overhead, and requires developers to switch between adding qualifiers in the code and looking at abstract object graphs to understand the object structures that the qualifiers describe. We propose an approach where developers express their design intent by refining an object graph directly, while an inference analysis infers valid qualifiers in the code. We present, formalize and implement the inference analysis. Novel features of the inference analysis compared to closely related work include a larger set of qualifiers to support less restrictive object hierarchy (logical containment) in addition to strict hierarchy (strict encapsulation), as well as object uniqueness and object borrowing. A separate extraction analysis then uses these qualifiers and extracts an updated object graph. We evaluate the approach on two subject systems. One of the subject systems is reproduced from an experiment using related techniques and another ownership type system, which enables a meaningful comparison. For the other subject system, we use its documentation to pick refinements that express design intent. We compute metrics on the refinements (how many attempts on each subject system) and classify them by their type. We also compute metrics on the inferred qualifiers and metrics on the object graphs to enable quantitative comparison. Moreover, we qualitatively compare the hierarchical object graphs with the flat object graphs and with each other, by highlighting how they express design intent. Finally, we confirm that the approach can infer from refinements valid qualifiers such that the extracted object graphs reflect the design intent of the refinements.

Off-campus Download