How do programmers use optional typing?

Abstract

The recent popularization of dynamically typed languages, such as Ruby and JavaScript, has brought more attention to the discussion about the impact of typing strategies on development. Types allow the compiler to find type errors earlier and potentially improve the readability and maintainability of code. On the other hand, "untyped" code may be easier to change and require less work from programmers. This paper tries to identify the programmers' point of view about these tradeoffs. An analysis of the source code of 6638 projects written in Groovy, a programming language which features optional typing, shows in which scenarios programmers prefer to type or not to type their declarations. Our results show that types are popular in the definition of module interfaces, but are less used in scripts, test classes and frequently changed code. There is no correlation between the size and age of projects and how their constructs are typed. Finally, we also found evidence that the background of programmers influences how they use types.

The behavior of gradual types: a user study

“…Groovy has optional type annotation checking. Souza and Figueiredo [29] study how programmers use optional typing in a repository of Groovy programs. Type annotation checking was a recent feature for Groovy at the time, so type annotations in the code served mostly as documentation.…”

Exploiting Type Hints in Method Argument Names to Improve Lightweight Type Inference

“…These type annotations usually serve as a kind of documentation and they are often exercised by developers [35], [65]. Developers with less background in dynamically-typed languages tend to use type annotations more often, mostly in method definitions.…”