Implement valuable-serde

[taiki-e]

Closes #17

Examples

use valuable::Valuable;
use valuable_serde::Serializable;

#[derive(Valuable)]
struct Point {
    x: i32,
    y: i32,
}

let point = Point { x: 1, y: 2 };

let value = Serializable::new(&point);

assert_eq!(
    serde_json::to_string(&value).unwrap(),
    r#"{"x":1,"y":2}"#,
);

[taiki-e]

Currently, Valuable treats the unit variant/struct as an empty tuple variant/struct. Therefore, I'm not sure if we can provide a behavior that is fully consistent with serde on unit variant/struct.

[carllerche]

This seems related to #21, maybe provide some thoughts there?

[taiki-e]

The problem I encountered here is mainly related to the unit struct, not to tuples. But in any case, we need to think about how to represent tuples...

valuable/valuable-serde/tests/test.rs

Lines 85 to 101 in 9a68fc3

	// TODO:
	// - valuable treats unit struct as an empty tuple struct
	// - serde treats unit struct as unit struct
	#[test]
	fn test_unit_struct() {
	#[derive(Debug, PartialEq, Valuable, Serialize)]
	struct S;
	assert_ser_tokens(
	&Serializable::new(S),
	&[
	Token::TupleStruct { name: "S", len: 0 },
	Token::TupleStructEnd,
	],
	);
	assert_ser_tokens(&S, &[Token::UnitStruct { name: "S" }]);
	}

[taiki-e]

• valuable treats Option<T> as T or unit
• serde treats Option<T> as Some(T) or None

[carllerche]

Ok, we can consider breaking out Some and None in Value?

[taiki-e]

I think this is now ready for review.

Currently, I'm aware of two differences in behavior from serde. See #23 (comment) and #23 (comment) for more.

[carllerche]

Looks great! Could you file follow up issues w/ the remaining unknowns?

[taiki-e]

filed #42 to track remaining todos.