|
@withoutboats | |||||
|
MONADS AND RUST A THREAD
- or -
why async/await instead of do notation
|
||||||
|
||||||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
Let's start with borrowing across yield points. We've solved this problem for real yield points with pin, I have no idea how to solve it using a function interface like >>=
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
So that means borrows across statements in do notation are just out of the picture. That's extremely limiting.
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
Rust's imperative control flow statements like `return` and `break` inside of do notation also doesn't make sense, because we do not have TCP preserving closures.
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
So we only know how to have do notation without early return or borrowing. People say this is an open research question, but IMO its just not possible within our runtime constraints.
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
Getting beyond that, this is assuming that "future" implements "monad," so if we could just add HKT to have a Monad trait, everything would be hunkydory. That's not true!
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
the signature of >>= is `m a -> (a -> m b) -> m b`
the signature of Future::and_then is roughly `m a -> (a -> m b) -> AndThen (m a) b`
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
That is, in order to reify the state machine of their control flow for optimization, both Future and Iterator return a new type from their >>= op, not "Self<U>"
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
Also, our functions are not a `->` type constructor; they come in 3 different flavors, and many of our monads use different ones (FnOnce vs FnMut vs Fn).
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
Okay, so Monad can't abstract over Future, but still let's have Monad. Problem: we don't have higher kinded polymorphism, and probably never will.
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
The problem is that without currying at the type level, higher kinded polymorphism makes type inference trivially undecidable. We have no currying.
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
In order to add higher kinded polymorphism, we'd have to restrict the sorts of types you could use in a way that would feel very arbitrary to users.
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
In contrast, generic associated types don't have this problem, and directly solve the expressiveness problems we do have, like Iterable (abstracting over everything with a `.iter(&self)` method)
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
(Don't get me wrong, you can write something to abstract over some monads like Option and Result using generic associated types. But its much less ergonomic than Monad in Haskell, even for those).
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
IN CONCLUSION: a design that works in a pure FP which lazily evaluates and boxes everything by default doesn't necessarily work in an eager imperative language with no runtime.
|
||
|
|
||
|
Saoirse Shipwreckt
@withoutboats
|
9. kol 2018. |
|
This comes up a lot, please link this thread liberally.
|
||
|
|
||