diff --git a/README.md b/README.md --- a/README.md +++ b/README.md @@ -188,25 +188,45 @@ * Effect System: For a term representing an effectful program, is it possible to statically decidable a type that enumerates all the effects the program may produce? * Purely Monadic: Is an effectful program represented as a transparent data structure that is a monad, and can it be interpreted into other data types using only pure operations without side effects or `unsafePerformIO`? * Dynamic Effect Rewriting: Can an effectful program have its internal effects altered afterwards (by functions typically referred to as `handle with`, `intercept`, `interpose`, `transform`, `translate`, or `rewrite`) ? + + For example, would it be possible to apply `interpose` as many times as the number of values input by the user at runtime? + +* Semantics: Classification of behaviors resulting from the interpretation of effects. + + * continuation-based: The same as Algebraic Effects and Handlers. + * IO-based: IO + Reader pattern. + * carrier dependent: The behavior depends on the specific type inference result of the monad. Tagless-final style. + * Performance: Time complexity or space complexity. -| Library or Language | Higher-Order Effects | Delimited Continuation | Effect System | Purely Monadic | Dynamic Effect Rewriting | Performance (TODO) | -| ------------------- | -------------------- | ---------------------- | --------------| --------------------------------- | ------------------------ | ------------------ | -| Heftia | Yes | Multi-shot | Yes | Yes (also Applicative and others) | Yes | ? | -| freer-simple | No | Multi-shot | Yes | Yes | Yes | ? | -| Polysemy | Yes | No | Yes | Yes | Yes | ? | -| Effectful | Yes | No | Yes | No (based on the `IO` monad) | Yes | ? | -| eff | Yes | Multi-shot? | Yes | No (based on the `IO` monad) | Yes | Fast | -| mtl | Yes | Multi-shot (`ContT`) | Yes | Yes | No | ? | -| fused-effects | Yes | No? | Yes | Yes | No | ? | -| koka-lang | No [^2] | Multi-shot | Yes | No (language built-in) | Yes | ? | -| OCaml-lang 5 | ? | One-shot | No [^3] | No (language built-in) | ? | ? | +| Library or Language | Higher-Order Effects | Delimited Continuation | Effect System | Purely Monadic | Dynamic Effect Rewriting | Semantics | Performance (TODO) | +| ------------------- | -------------------- | ---------------------- | --------------| --------------------------------- | ------------------------ | -------------------------------- | ------------------ | +| Heftia | Yes | Multi-shot | Yes | Yes (also Applicative and others) | Yes | continuation-based | ? | +| freer-simple | No | Multi-shot | Yes | Yes | Yes | continuation-based | ? | +| Polysemy | Yes | No | Yes | Yes | Yes | weaving-based (functorial state) | ? | +| Effectful | Yes | No | Yes | No (based on the `IO` monad) | Yes | IO-based | ? | +| eff | Yes | Multi-shot? | Yes | No (based on the `IO` monad) | Yes | continuation-based (IO-fused) | ? | +| in-other-words | Yes | Multi-shot? | Yes | Yes | No? | carrier dependent | ? | +| mtl | Yes | Multi-shot (`ContT`) | Yes | Yes | No | carrier dependent | ? | +| fused-effects | Yes | No? | Yes | Yes | No | carrier dependent & weaving-based (functorial state) | ? | +| koka-lang | No [^2] | Multi-shot | Yes | No (language built-in) | Yes | continuation-based | ? | +| OCaml-lang 5 | ? | One-shot | No [^3] | No (language built-in) | ? | continuation-based? | ? | [^2]: https://gist.github.com/ymdryo/6fb2f7f4020c6fcda98ccc67c090dc75 [^3]: Effects do not appear in the type signature and can potentially cause unhandled errors at runtime Heftia can simply be described as a higher-order version of freer-simple. This is indeed true in terms of its internal mechanisms as well. + +Additionally, this library provides a consistent *continuation-based* semantics that is independent of carriers and effects. +On the other hand, in libraries like in-other-words, mtl, and fused-effects, the semantics of the code depend on the effect and, in part, the carrier inferred by type inference. +Fixing the semantics to a *continuation-based* model helps improve the predictability of the behavior (interpretation result) of the code. + +Carrier-dependent semantics can lead to unexpected behavior for code readers, particularly in situations where the types become implicit. +Particularly, attention should be given to the fact that due to type inference, semantic changes may propagate beyond the blocks enclosed by `interpret` or `interpose`. +In the case of carrier-independent semantics, especially with Freer-based effects, `interpret` and `interpose` do not alter the semantics by intervening in type inference or instance resolution of the carrier. +Instead, they function as traditional functions, simply transforming the content of the data structure. +This results in minimal surprise to the mental model of the code reader. ### Compatibility with other libraries #### Representation of effects diff --git a/heftia-effects.cabal b/heftia-effects.cabal --- a/heftia-effects.cabal +++ b/heftia-effects.cabal @@ -1,6 +1,6 @@ cabal-version: 2.4 name: heftia-effects -version: 0.3.0.0 +version: 0.3.1.0 -- A short (one-line) description of the package. synopsis: higher-order effects done right @@ -20,7 +20,7 @@ maintainer: Yamada Ryo -- A copyright notice. -copyright: 2023 Yamada Ryo +copyright: 2023-2024 Yamada Ryo category: Control, Effect, Monads extra-source-files: @@ -55,8 +55,8 @@ build-depends: base ^>= 4.16, ghc-typelits-knownnat ^>= 0.7, - data-effects ^>= 0.1.1, - heftia ^>= 0.3, + data-effects ^>= 0.1.2, + heftia ^>= 0.3.1, ghc-options: -Wall -fplugin GHC.TypeLits.KnownNat.Solver @@ -191,7 +191,7 @@ hs-source-dirs: Example/Logging build-depends: heftia-effects, - text ^>= 1.2.5, + text >= 2.0 && < 2.2, time ^>= 1.11.1, executable Continuation