SE-Radio Episode 322: Bill Venners on Property Based Tests

Generator Basics

Generators play a crucial role in property-based testing by producing the necessary data types for testing properties. Bill Venners explains that in languages like Haskell and Scala, generators are based on types, meaning they automatically produce data like doubles or integers based on the function's input type 1. This type-based generation ensures that the data aligns with the expected input, but custom types require custom generators. Matthew Farwell highlights the importance of matching generator types with target class types to avoid issues like generating invalid data 2.

The type of your generator should match the type of your target class.

--- Bill Venners

This approach helps uncover unexpected issues in code, such as improper handling of null values.

   

Structured Data

Generating structured and context-dependent data presents unique challenges in property-based testing. Bill Venners describes how Scala's for expressions can be used to create generators for complex data types, like a person class with a name and age 3. These generators are concise but require careful composition to ensure all necessary data attributes are covered. However, Matthew Farwell points out the risk of combinatorial explosion, where not all data variations are tested due to the vastness of possible combinations 4.

You're pretty much probably not going to explore the whole space.

--- Bill Venners

Despite these challenges, property-based testing remains a valuable tool for identifying edge cases and ensuring robust code.