Programming Language Type Systems and Concurrency

点击翻译按钮获取中文摘要

Every aspect of computer science depends on programming languages to transform human ideas into a practical, functioning form executable by a computer. Two crucial aspects of programming-language design are the type system and runtime. This research program will focus on the type system used to verify correct usage and provide code reuse, and on the runtime system to simplify program construction and improve performance. The primary research objective is the design of a novel type system to replace classical object-oriented inheritance for code reuse. Over the past 40 years, the limitations of object-oriented inheritance is well documented. Furthermore, object-oriented inheritance is insufficient for many forms of reuse so other mechanisms are needed in the type system, like overloading and generics (templates), resulting in a confusion of mechanisms to generate reusable programs. The design for the new type-system uses a technique called parametric-polymorphic functions with explicit constraints. Each polymorphic function or type defines a structural type-key needed for its execution; i.e., the set of operations used by a function or type on its parameters/fields. This key set is fulfilled at each call site from the lexical environment and implicitly passed into a polymorphic function. Hence, new lexical scopes and nested functions are used extensively to create local subtypes without having to manage an inheritance hierarchy. Improving a language's type systems makes programmers more productive and results in programs that are reliable and safer. The secondary research objective is to use the new type system to create a hierarchy of high-level, integrated, and efficient control-flow features. Concurrent control-flow is difficult because of temporal ordering and nondeterminism on multiple cores. However, many current systems exacerbate the difficulty through complex presentation mechanisms. The new design would eliminate confusing semantics that are nonintuitive and lead to errors, and by providing customs types for high-level control flow, generators, coroutines, threads, and monitors, versus multiple low-level lock acquisition techniques. Improving a programming language's concurrency allows programmers to take advantage of multiple cores (parallelism) to reduce program execution time (run faster). These research results are being disseminated in a programming language that is backwards compatible with C, fixes well known C problems, while adding safety and modern language-features. C is still the foundational technology for modern computing with billions of lines of code implementing small hobby projects to huge commercial applications (Linux/Windows/MacOS operating systems). This work has a direct benefit to a large segment of software-development companies using the C programming language daily. Without extensions to C, it is unable to cope with the needs of modern programming problems and programmers, putting it in jeopardy.