Don't use null objects for error handling

The article critiques the practice of using null objects for error handling in programming, particularly in lower-level languages like C. It argues that returning a harmless default value, such as zero, when an operation fails can lead to misleading results and a poor user experience. The author emphasizes that this approach, often referred to as Zero Is Initialization (ZII), fails to address the underlying error and instead allows the program to continue operating with potentially invalid data. An example is provided where a function, `find_bone`, returns zero for an invalid input, which is treated as a valid output, leading to further erroneous calculations. The author suggests that true error handling should involve recording the error and aborting the erroneous task rather than propagating a null object through the system.

The article outlines various strategies for handling errors, including restricting input sets, using assertions, and expanding output sets. It advocates for immediate error handling by aborting the task that caused the error and suggests that programmers should choose the appropriate error handling strategy based on the context of the function's use. The author concludes that while null objects may seem convenient, they often mask errors rather than resolve them, leading to garbage data and a subpar user experience.