sintrastes / hafly Goto Github PK

Currently these have to be specified one-by one, e.x. a forEach for IO, a forEach for List, etc...

It would be really nice if we could specify these polymorphically somehow. I'm not entirely sure how that would work.

Currently, we have:

> [1,2,3,4.2]
[<<Int>>,<<Int>>,<<Int>>,<<Double>>]

whereas it would be nice if we had:

> [1, 2, 3, 4.2]
[1, 2, 3, 4.2]

Maybe the best approach for this would be to have the interpreter see if there is a show instance for each element of the list, and try to display it that way -- otherwise, falling-back to just displaying the type?

We'll need some way to disambiguate these from function calls, maybe something like:

For an expression x.y:

1. If x is a record, first check to see if it has any fields y. If so, use y as a record accessor for that field.
2. Otherwise, check to see if there is a function y that could act on x.

e.x. [1, 2, 3].map \x -> x + 2

The reason for this is likely that . is not defined to use flexible application.

This would allow us to do things like define /= in terms of ==.

To be fully general, we should also allow expressions to depend on operators, and operators to depend on expressions.

Currently, examples such as foldr (\x y -> x + y) 0 [1,2,3,4,5] do not evaluate properly.

I believe this has to do with lambdas always having type Dynamic -> Dynamic.

Currently, currying seems to be causing issues, e.x.

f = \x y -> x
g = f 2

> g 3
3 -- (expected 2)

e.x. if I start typing:

main = {

I want the next line of the interpreter to be indented by four spaces.

In order to properly act on Dynamic data with higher-order functions, we'll probably need to consider variance (e.x. implicitly lifting a [Int] to a [Dynamic]. We want this to be able to be specified for an arbitrary Haskell data type to keep things extensible.

This may also be relevant for features like functor/applicative/monad polymorphism. e.x. if f : Int -> Int and x: [Int], we could say f x, and have that automatically lifted to map f x. This I think is the general idea behind what I was thinking of as "reactive polymorphism".

Currently multiple dispatch only works on library-defined functions.

I believe one reason for this is that the user currently has no way of being able to specify type signatures.

We should probably make our type signatures mirror that of Haskell's.