Algebraic effect handlers are a powerful way to incorporate effects in a programming language. Sometimes perhaps even _too_ powerful. In this article we define a restriction of general effect handlers with _scoped resumptions_. We argue one can still express all important effects, while improving reasoning about effect handlers. Using the newly gained guarantees, we define a sound and coherent evidence translation for effect handlers, which directly passes the handlers as evidence to each operation. We prove full soundness and coherence of the translation into plain lambda calculus. The evidence in turn enables efficient implementations of effect operations; in particular, we show we can execute tail-resumptive operations _in place_ (without needing to capture the evaluation context), and how we can replace the runtime search for a handler by indexing with a constant offset.

@article{XieBHSL20,
  title = {Effect handlers, evidently},
  author = {Ningning Xie and Jonathan Immanuel Brachthäuser and Daniel Hillerström and Philipp Schuster and Daan Leijen},
  journal = {Proceedings of the ACM on Programming Languages},
  volume = {4},
  number = {ICFP},
  year = {2020},
  doi = {10.1145/3408981}
}