hint VS klister

Mueval is based on hint, which is in turn based on the ghc library.

Yeah, that project is pretty much at the bottom of my list, unfortunately. My top projects these days are mgmt, klister, recursion-schemes, and hint... And that's already too much!

You can try using hint (instead of ghci) though I'm not sure it has the breakpoint functionality.

Have you tried hint?

with haskell.nix, well, you've found the github issue, you need to put the apecs package in the right nix incantation.

But the use case for that is for using a Haskell program A to catch errors in that same Haskell program A. For your use case, using a Haskell program A to automatically grade a Haskell program B, I recommend using the hint library instead, as it allows you to load code from external source files, run tests on them, and manipulate the error messages produced by ghc. (full disclosure: I am the maintainer of that library)

I am thinking of giving hint the ability to evaluate TemplateHaskell expressions. It would indeed be quite difficult to write an interpreter for all of Haskell, so my plan is to use the Exp's Show instance to produce a program which constructs and then splices that Exp, e.g. $(pure (InfixE (Just (LitE (IntegerL 1))) (VarE GHC.Num.+) (Just (LitE (IntegerL 1))))) is a Haskell expression which is equivalent to 1 + 1, so I should be able to ask hint to evaluate that to get 2 without having to write my own Haskell interpreter.

Yes please! Right now all of my open-source projects (most notably hint and recursion-schemes) are about to drop into barely-updated mode, and while I knew this would happen and have been working towards finding co-maintainers, I am now realizing that it wasn't enough. I think such a website would definitely have helped, and I am hoping that once it launches, I'll be able to use it to find some co-maintainers to tide over my projects until I become available again.

hint's API takes a string, not an AST (I plan to fix this). Internally, hint delegates to the ghc library, which does expose a parser which you can use if you want. hint exists to provide a friendlier API than the ghc library for interpreting Haskell code, but it does not expose a friendlier API for parsing Haskell code.

Yeah, that project is pretty much at the bottom of my list, unfortunately. My top projects these days are mgmt, klister, recursion-schemes, and hint... And that's already too much!

case is a special form, ie a bit of core syntax, but, interestingly, data is not. (It is presumably a macro; typechecking is actually done as a part of macro expansion.) The syntax remains pretty uniform. Or, in Klister, type ascription is done via normal S-expression syntax with a form called the, as (the $type $expression); again, the syntax is uniform.

Shameless plug: we don't have that problem in Klister, because our equivalent to main is a run macro which runs an IO action, and your alternate prelude can define its own run macro which expects an IO action from your alternate prelude.

Klister is very similar to Hackett, but implemented in Haskell instead of Racket, and my most recent PR is from 20 days ago, if that's the metric which counts for you. Still very much of a WIP though.

You might be interested in Klister: https://github.com/gelisam/klister

I've now also found an implementation for klister, which is meant to interleave type checking with macro expansion.

In Klister, which already has Scheme-like macros and Haskell-like types (polymorphism, algebraic types and higher-kinded types, but not yet fancier types like RankNTypes and GADTs), our plan to get the best of both worlds (lexical syntax and typed ASTs) is to separate parsing from macro evaluation. That is, users write their programs using the surface syntax of s-expressions, parsers parse those into typed ASTs, and macros are typed by the type of the ASTs they receive as input and produce as output. At this stage this is only a research idea, I don't know if that's going to work out yet, but I hope so!