Implement generic assert_impl macro

[Nadrieril]

After squinting for a while at the wrapper trick for assert_impl_any, it turns out it generalizes quite well. I implemented a generic assert_impl macro with type-level booleans that covers most of the cases where we want to check for a type implementing a set of traits.
Examples:

assert_impl!(Cell<u32>: Send && !Sync);
assert_impl!(u32: (Into<u16>) || (Into<u64>));
assert_impl!(for(T) Rc<T>: Clone);
assert_impl!(for(T: Copy) T: Clone);

The error messages are also quite nice:

error[E0308]: mismatched types
  --> tests/trait_impl.rs:66:1
   |
66 | assert_impl!(for(T) T: Clone);
   | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected struct `static_assertions::True`, found struct `static_assertions::False`
   |
   = note: expected type `static_assertions::True`
              found type `static_assertions::False`
   = note: this error originates in a macro outside of the current crate (in Nightly builds, run with -Z external-macro-backtrace for more info)

[Nadrieril]

There are a few limitations to the syntax due to macro shenanigans. Maybe we should make a proc_macro instead.

[Nadrieril]

This subsumes #27.

[Nadrieril]

Since the True and False structs are local to the macro, I believe that this is safe against external blanket traits

[Nadrieril]

I just saw that #26 had a similar idea. I could use const booleans too actually, as long as I can recover pretty error messages

[nvzqz]

This is awesome!! One of my plans for 2.0 was to consolidate the assert_impl*! family of macros into a single one that uses ! for "not", | for "any", and & for "all". I haven't had the time to sit down and attempt it, so thank you very much for putting in the work!

After squinting for a while at the wrapper trick for assert_impl_any

I can only imagine you've been squinting for quite some time. I'm very impressed by you being able to do this. I'll read through the implementation when I have more time this weekend to make sense of your tt munching.

---

In the meantime, could you extract the bits that emit True and False to the assert_impl*! family of macros? I'd like to improve those error messages.

My plan is to adopt this nicer syntax in a 1.x version and deprecate the others to be removed in 2.0.

[Nadrieril]

could you extract the bits that emit True and False to the assert_impl*! family of macros?

I'm not sure what you mean :/

I'd like to improve those error messages

I am in the process of using the newly defined type_level_bool::{True, False} types in assert_impl to get better errors :)

[nvzqz]

So the error message here shows:

expected struct `_::True`, found struct `_::False`

I was saying I'd like that for the old macros.

But it seems that you've redefined the other macros to use this new macro instead. So I imagine you'd get that for free.

[Nadrieril]

Rebased on master and now we got the pretty errors

[nvzqz]

Thanks! I'll definitely try to check this out soon.

Your True and False type stuff got me to create typebool, which I think I'm gonna have this crate use.

[Nadrieril]

We probably don't need that much structure on True and False. I'd rather use const booleans if we can instead, and use True and False only for pretty errors. I just haven't quite figured out how because the deref trick can't be used to generate a const value

[Nadrieril]

Ugh I hit an ICE rust-lang/rust#62708 when trying to use const bool instead of type-level bools :(

[Nadrieril]

Ok, so now it's easy to get a const type-level boolean testing whether a type implements a particular bunch of traits (that's what the hidden does_impl macro does). It's also easy to get a const bool from that. But because of rust-lang/rust#62708, we can't use const_assert on that boolean and remove the intermediate type-level bools :(
If we had another way of implementing to_bool! that might solve this problem.

[nvzqz]

Sorry I haven't gotten back to you on this. Life got in the way. I'll review this soon, I promise!

[nvzqz]

Hey @Nadrieril thanks again for this PR! ❤️

I finally got around to looking through it and I'm happy with the implementation. Apologies for it taking so long to review. A new version will be released soon with this macro!

[nvzqz]

@Nadrieril I rewrote the macro and fixed all of the issues with it! And I decided to put it into its own library at https://github.com/nvzqz/impls. I credited you as co-author in the README.md, in the LICENSE-MIT file, and in the Cargo.toml file. I posted it on Twitter and people seem to like it. Thanks again!

[Nadrieril]

That's great ! You seem to have dropped support for generics though ?

[wycats]

@nvzqz I was surprised to see the missing support for for(...) generics as well. I see that the code is still in assert_impl, but assert_impl seems not to be exposed.

[wycats]

Upon further investigation, it looks like the issue is that the commits after November 3, 2019 were never released to crates.io.

[TTWNO]

@nvzqz Can you please release the latest changes on crates.io? I realize the functionality of this crate is quite stable, but this feature would help me a lot.

Happy to help if needed.