opeltre/fp

Functional programming in python

fp's Issues

Docs

Keep track of the documentation here:

🆗 fill the index and make it friendly

• 📦 install with pip install funprogram ✨
• 📝 the manual will get you started with the basics of fp 🛩️
• 📚 the docstrings for any deeper understanding of the internals! ⚙️

🍞 Finish the manual

• 🆗 The Type Category
• 🍞 Functor and Monads
• 🍞 The Tens Category

Casts

Do not cast automatically, or at least not with python builtins.

Right now, anything in Show i.e. implementing __repr__ is cast to string, and the following is accepted:

f = Arrow(str, int)(len)
f([0, 1, 2, 3])
#out: length of string representation

Curry

Implement:

curry   :: ((a, b) -> c) -> a -> b -> c
uncurry :: (a -> b -> c) -> (a, b) -> c

As methods of the Arrow class.

Then, given arity of function f, return partial application (slice) of f with smaller number of arguments.

To do this properly, we should have product types, so that:

Arrow(A * B, C) ~ Arrow(A, Arrow(B, C))

And arity 2 is infered from the input type A * B.

Uncurry

Curryfication works well by inheritance from ArrowMeta with 28bc2da

=> Implement Arrow.uncurry for (recursively) unfolding tgt attribute when tgt = Arrow(B, C) !

Multilinear instances

Support tuple of shapes input when creating Linear instances.

Contracting against n vectors is lighter than applying a flattened matrix to the tensor product.