Type migration is the process of adding types to untyped code to gain assurance at compile time. TypeScript and other gradual type systems facilitate type migration by allowing programmers to start with imprecise types and gradually strengthen them. However, adding types is a manual effort and several migrations on large, industry codebases have been reported to have taken several years. In the research community, there has been significant interest in using machine learning to automate TypeScript type migration. Existing machine learning models report a high degree of accuracy in predicting individual TypeScript type annotations. However, in this paper we argue that accuracy can be misleading, and we should address a different question: can an automatic type migration tool produce code that passes the TypeScript type checker? We present TypeWeaver, a TypeScript type migration tool that can be used with an arbitrary type prediction model. We evaluate TypeWeaver with three models from the literature: DeepTyper, a recurrent neural network; LambdaNet, a graph neural network; and InCoder, a general-purpose, multi-language transformer that supports fill-in-the-middle tasks. Our tool automates several steps that are necessary for using a type prediction model, including (1) importing types for a project’s dependencies; (2) migrating JavaScript modules to TypeScript notation; (3) inserting predicted type annotations into the program to produce TypeScript when needed; and (4) rejecting non-type predictions when needed. We evaluate TypeWeaver on a dataset of 513 JavaScript packages, including packages that have never been typed before. With the best type prediction model, we find that only 21% of packages type check, but more encouragingly, 69% of files type check successfully.

Large language models have demonstrated the ability to generate both natural language and programming language text. Although contemporary code generation models are trained on corpora with several programming languages, they are tested using benchmarks that are typically monolingual. The most widely used code generation benchmarks only target Python, so there is little quantitative evidence of how code generation models perform on other programming languages. We propose MultiPL-E, a system for translating unit test-driven code generation benchmarks to new languages. We create the first massively multilingual code generation benchmark by using MultiPL-E to translate two popular Python code generation benchmarks to 18 additional programming languages. We use MultiPL-E to extend the HumanEval benchmark [1] and MBPP benchmark [2] to 18 languages that encompass a range of programming paradigms and popularity. Using these new parallel benchmarks, we evaluate the multi-language performance of three state-of-the-art code generation models: Codex [1], CodeGen [3]and InCoder [4]. We find that Codex matches or even exceeds its performance on Python for several other languages. The range of programming languages represented in MultiPL-E allow us to explore the impact of language frequency and language features on model performance. Finally, the MultiPL-E approach of compiling code generation benchmarks to new programming languages is both scalable and extensible, making it straightforward to evaluate new models, benchmarks, and languages.

In order to generate efficient code, dynamic language compilers often need information, such as dynamic types, not readily available in the program source. Leveraging a mixture of static and dynamic information, these compilers speculate on the missing information. Within one compilation unit, they specialize the generated code to the previously observed behaviors, betting that past is prologue. When speculation fails, the execution must jump back to unoptimized code. In this paper, we propose an approach to further the specialization, by disentangling classes of behaviors into separate optimization units. With contextual dispatch, functions are versioned and each version is compiled under different assumptions. When a function is invoked, the implementation dispatches to a version optimized under assumptions matching the dynamic context of the call. As a proof-of-concept, we describe a compiler for the R language which uses this approach. Our implementation is, on average, 1.7× faster than the GNU R reference implementation. We evaluate contextual dispatch on a set of benchmarks and measure additional speedup, on top of traditional speculation with deoptimization techniques. In this setting contextual dispatch improves the performance of 18 out of 46 programs in our benchmark suite.

The R programming language combines a number of features considered hard to analyze and implement efficiently: dynamic typing, reflection, lazy evaluation, vectorized primitive types, first-class closures, and extensive use of native code. Additionally, variable scopes are reified at runtime as first-class environments. The combination of these features renders most static program analysis techniques impractical, and thus, compiler optimizations based on them ineffective. We present our work on PIR, an intermediate representation with explicit support for first-class environments and effectful lazy evaluation. We describe two dataflow analyses on PIR: the first enables reasoning about variables and their environments, and the second infers where arguments are evaluated. Leveraging their results, we show how to elide environment creation and inline functions.

High-performance dynamic language implementations make heavy use of speculative optimizations to achieve speeds close to statically compiled languages. These optimizations are typically performed by a just-in-time compiler that generates code under a set of assumptions about the state of the program and its environment. In certain cases, a program may execute code compiled under assumptions that are no longer valid. The implementation must then deoptimize the program on-the-fly; this entails finding semantically equivalent code that does not rely on invalid assumptions, translating program state to that expected by the target code, and transferring control. This paper looks at the interaction between optimization and deoptimization, and shows that reasoning about speculation is surprisingly easy when assumptions are made explicit in the program representation. This insight is demonstrated on a compiler intermediate representation, named sourir, modeled after the high-level representation for a dynamic language. Traditional compiler optimizations such as constant folding, unreachable code elimination, and function inlining are shown to be correct in the presence of assumptions. Furthermore, the paper establishes the correctness of compiler transformations specific to deoptimization: namely unrestricted deoptimization, predicate hoisting, and assume composition.

We present Flix, a declarative programming language for specifying and solving least fixed point problems, particularly static program analyses. Flix is inspired by Datalog and extends it with lattices and monotone functions. Using Flix, implementors of static analyses can express a broader range of analyses than is currently possible in pure Datalog, while retaining its familiar rule-based syntax. We define a model-theoretic semantics of Flix as a natural extension of the Datalog semantics. This semantics captures the declarative meaning of Flix programs without imposing any specific evaluation strategy. An efficient strategy is semi-naive evaluation which we adapt for Flix. We have implemented a compiler and runtime for Flix, and used it to express several well-known static analyses, including the IFDS and IDE algorithms. The declarative nature of Flix clearly exposes the similarity between these two algorithms.