Abstract
The paper is concerned with those properties of software that can be statically surmised from the source code. Many such properties have been extensively studied from the perspective of compiler construction technology. However, live variable analysis, alias analysis and the such are too low level to be of interest to the software engineer. The authors identify a family of statically checkable properties that should represent a higher level abstraction, and reach the detailed design level. Properties in this family which is defined by five precise distinguishing criteria are called traits. Some examples of traits include mutability, const correctness, ownership, and pure functions. In fact, in many ways, traits are non-standard types. They argue that traits should bring about similar benefits to these of static typing in terms of clarity, understandability, adherence to design decisions, and robustness. They further argue that traits can be used for better checking of substitutability in inheritance relationships. Having made the case for traits, they proceed to describing a taxonomy for classifying and understanding traits and show how it can be used to better understand previous work on this topic. The paper also discusses the abstract computational complexity of traits and compares previous research from that perspective.