Solver-based gradual type migration

Gradually typed languages allow programmers to mix statically and dynamically code, enabling them incrementally reap the benefits of static typing as they add type annotations their code. However, this migration process is typically a manual effort with limited tool support. This paper examines problem automated migration: given dynamic program, infer additional or improved annotations. Existing algorithms prioritize different goals, such maximizing precision, maintaining compatibility unmigrated preserving semantics original program. We argue that involves fundamental compromises: optimizing for single goal often comes at expense others. Ideally, would flexibly accommodate range user priorities. present TypeWhich, new approach gradually-typed lambda calculus some extensions. Unlike prior work, which relies on custom solvers, TypeWhich produces constraints an off-the-shelf MaxSMT solver. allows us easily express objectives, minimizing number necessary syntactic coercions, constraining be compatible first comprehensive evaluation GTLC algorithms, compare four other tools from literature. Our uses benchmarks, set "challenge problems." Moreover, we design methodology highlights subtleties gradual migration. In addition, apply suite benchmarks Grift, programming language based GTLC. able reconstruct all human-written but one