Safe and Performant Programming Language for Intermediate Calculations in Task Orchestration Pipelines

To orchestrate workflows composed of sequentially executed tasks, it is essential to calculate task arguments based on the outcomes of prior tasks. Existing programming languages used in popular orchestrators suffer from either low expressiveness or reduced performance due to necessary safety checks made in runtime. We propose an approach that shifts some of these safety checks to the static analysis phase by incorporating a type system into the language. This optimization allows for maintaining high language expressiveness while achieving competitive performance and ensuring computational safety. In this study, we implemented a prototype programming language leveraging optimizations made possible by the type system. Our findings indicate that, through static analysis, execution of arithmetic operations is sped up by 25%. Furthermore, our efficient implementation is shown to perform common calculations three times faster than the language currently used by the orchestrator in Yandex.Taxi.