String Representations of Java Objects: An Empirical Study

Abstract

String representations of objects are used for many purposes during software development, including debugging and logging. In Java, each class can define its own string representation by overriding the toString method. Despite their usefulness, these methods have been neglected by researchers so far. In this paper, we describe an empirical study of toString methods performed on a corpus of Java files. We are asking what portion of classes defines toString, how are these methods called, and what do they look like. We found that the majority of classes do not override the default (not very useful) implementation. A large portion of the toString method calls is implicit (using a concatenation operator). The calls to toString are used for nested string representation building, exception handling, in introspection libraries, for type conversion, and in test code. A typical toString implementation consists of literals, field reads, and string concatenation. Around one third of the string representation definitions is schematic. Half of such schematic implementations do not include all member variables in the printout. This fact motivates the future research direction – fully automated generation of succinct toString methods.