I feel like I should elaborate on my usecase a little bit more. The arcitecture of my game engine is that there's a hypothetical giant struct that contains all of your game data, and then there's a giant pure function that takes an input state and the previous game state, and gives you the game state for the new frame. The problem is that writing a big function like that is a bit unwieldly, so I had the idea of separating each part out into a tiny script containing a pure function. My initial idea was that if I could strip out the io and os modules from the environment, each lua module (file) would be considered pure as a whole. So this type system was designed such that I could: Get a value of a type -> Convert it to lua types -> call lua code with it -> Convert into my type system -> Type check the result I don't know if I needed the TypedValue; but I felt it was convenient for a value to carry around it's type for some reason, my (in foresight flawed) vision was that lua code would call into haskell code, and I would keep on type checking in the interim, but I now realize that I don't need to do that if I'm checking at the end anyways. Now that you mention it, implementing my own statically typed language would probably be better than trying to rig a strict type system on a traditionally dynamically typed language. I have to think about this for a little bit. Thank you for the notes, I'll be looking them over. -------- Original Message -------- On Jul 10, 2023, 7:01 PM, Brent Yorgey wrote:

So to make sure I'm understanding correctly, you want to create your own statically typed scripting language for your game engine? Implementing your own statically-typed language is definitely a fun and worthwhile learning experience. You might find helpful the notes from my programming languages course, which is taught in Haskell: https://hendrix-cs.github.io/csci360/

Having `Type` and `Value` ADTs to represent types and values makes sense. They often do correspond closely since there is often a canonical kind of value for each type, but they are not necessarily exactly the same.

You didn't mention an ADT for expressions/terms, i.e. syntax. Typically you will parse concrete syntax to produce a Term, and then you will typecheck a Term to make sure it has a valid Type. Then you can interpret the Term to produce a Value.

I am not sure what the point of TypedValue is. Typically, once you are finished with typechecking, you no longer need to keep track of types while interpreting.

-Brent

On Mon, Jul 10, 2023 at 7:36 AM Talha Qamar via Haskell-Cafe wrote:

...

Hello everyone, I'm working on a game engine in Haskell. The thing is, while I want to write the architecture in Haskell, I'm thinking of using something else as the actual scripting language.

Long story short, how would I represent a type system in Haskell? I've gotten _something_ together that works a bit like this:

``` data Type = -- ...all possible types data Value = -- ...all possible types again data TypedValue = TypedValye Type Value ``` But I'm concerned if this is the ideal way of doing things, particularly when I get to complex types like structs or tagged unions.

Any help would be appreciated.

Sent from Proton Mail mobile

_______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.