Remove TS::trasparent
#243
Conversation
ts-rs/src/lib.rs
Outdated
| fn generics() -> impl TypeList | ||
| where | ||
| Self: 'static, | ||
| { | ||
| use std::marker::PhantomData; | ||
| ((PhantomData::<I>,), I::generics()) | ||
| } | ||
| } |
There was a problem hiding this comment.
So, this is equivalent to just returning I::generics().
I'm not sure that's correct, even though the tests pass.
Shouldn't this return ().push::<I>().extend(I::generics())? Right now, I itself would be missing from the TypeList, no?
Ahhh, I got confused by you constructing the TypeList manually here!
Is there a reason you're not doing ().push::<I>().extend(I::generics()) or, even better, I::generics().push::<I>, instead?
There was a problem hiding this comment.
Not really, I just kinda got confused calling methods on () so I did it this way
There was a problem hiding this comment.
I just kinda got confused calling methods on
()so I did it this way
That tripped me up quite a bit in the macros, because the repetition syntax uses parentheses plus the unit type plus the calling of methods
There was a problem hiding this comment.
Hm, yeah, I understand.
I do think it'd be a good idea to keep the implementation of TypeList private, and use the methods on it. That gives us the option to change TypeList is implemented however we like in the future.
In this particular case, I::generics().push::<I>() would probably be the nicest.
To make the code more readable (and keep the implementation detail that the empty TypeList is () hidden), we could add a const EMPTY_TYPELIST: () = () or pub fn new() -> impl TypeList {} to typelist.rs
For the macros, I try to use quote!{} or quote![] to disambiguate them, but yeah, there's still the #()* syntax.
In the end, this is not a big deal, if you've got strong opinions here. Personally, I prefer using the methods on the type.
Edit: I see that you already changed it, thanks! 😆
Being able to call methods on TypeList was the motivation to make dependency_types a function in the first place. It could have been an associated type (e.g type Dependencies = ((A), B), C::Dependencies)), but I liked the idea of being able to pretend that TypeList is just a Vec.
There was a problem hiding this comment.
All of that makes a lot of sense! I'll open another PR to apply this change to the impls I changed in #241, all of which manually build a TypeList
| if T::transparent() { | ||
| // the dependency `T` is "transparent", meaning that our original type depends | ||
| // on the dependencies of `T` as well. | ||
| T::dependency_types().for_each(self); | ||
| } else if let Some(dep) = Dependency::from_ty::<T>() { | ||
| // the dependency `T` is not transparent, so we just add it to the output | ||
| if let Some(dep) = Dependency::from_ty::<T>() { | ||
| self.0.push(dep); | ||
| } |
There was a problem hiding this comment.
I love that this just works. Awesome!
There was a problem hiding this comment.
Yeah, since T::transparent is always false we just get rid of that check
This PR is meant to remove the
TS::transparentmethod, as it was rendered useless by #241, always returnfalsewith the exception ofRangeandRangeInclusive