Anzen is a multi-paradigm programming language that aims to provide explicit and controllable assignment semantics. It is based on the observation that abstractions over memory management and data representation, as commonly adopted by contemporary programming languages, often transpire relics of the underlying memory model and lead to confusing assignment semantics in the presence of aliases. In response, Anzen’s goal is to offer a modern approach to programming, built on a sound and unambiguous semantics.

This paper reports our experience on implementing Anzen’s compiler, and the challenges thereof. Our implementation transpiles Anzen sources to an intermediate language inspired by the LLVM IR, designed to ease the analysis and evaluation of Anzen’s statements. This intermediate representation is then consumed by a register-based virtual machine. We present the Anzen compiler’s architecture, introduce its intermediate language and describe the latter’s evaluation. Our work aims to set a reference implementation for future developments and extensions of the language.