Typing annotations for BIP-0340

[ysangkok]

This is on top of #200.

Typing makes the code easier to read. It also makes it easier to see where assumptions about non-infinity are being made.

[real-or-random]

I like the typing annotations, that's neat! I'm not so sure about the assertions though. Since this is a reference implementation, people could get the idea that they need to implement them in their code.

[ysangkok]

I can remove the assertions, but they do have a purpose. The code will not type check with the --strict flag unless they are there. I think the annotations are still an improvement even if it doesn't type check.

[sipa]

@real-or-random I'm not sure that's a problem, actually. I would hope that most implementations of the inner logic used here is implemented in compiled, statically-typed languages, so the typing annotations would actually be a help for that.

Concept ACK

EDIT: I realize I misread, and you were talking about the assertions and not typing annotations. My comment doesn't apply then, but I still think it's not really a problem - real implementations will be very different anyway (at least I hope so).

[sipa]

Maybe rebase on current branch?

[ysangkok]

@sipa Rebased on 1d999cf. debug_print_vars needed adjusting, it now has an extra assert.

[sipa]

What is this int conversion for? Does pow return another type?

[ysangkok]

pow can return float. For example, try pow(2, -1). This is also noted in the common issues section of the Mypy manual. More discussion at mypy issue 285. If you prefer to avoid the coercion (which won't actually coerce), I think it would be possible to use cast, which is always guaranteed to be the identity function at runtime. Which do you prefer?

[sipa]

Let's keep it like this. I suspect cast is more obscure to most readers.

[sipa]

@real-or-random @jonasnick: opinions?

[real-or-random]

ACK

[jonasnick]

Concept ACK

This is a nice improvement to readability. I wasn't aware of these developments in python. I noticed that when you replace assert P is not None with P = None there's no runtime exception thrown by a type checker. How do you actually type check "with --strict"?

[real-or-random]

@jonasnick I think there's a misunderstanding. The (commonly used) type checker is a different binary mypy (http://mypy-lang.org/). It's a different project, so you may even need to install it separately. It accepts --strict.

[jonasnick]

Ah, so more dependencies. Can confirm that this works with mypy and that the asserts are required. We "just" need to remember to run this whenever we change the code. Well I guess someone will speak up eventually if it doesn't pass mypy.

ACK