Allow somehow validate the decoded output

[xlc]

For example, we want a custom integer type can only old number less than 100:

#[derive(Encode, Decode)]
struct UIntLessThan100(u8);

impl UIntLessThan100 {
  pub new(val: u8) -> Option<Self> { if val < 100 { Some(Self(val)) } else { None } 
  pub add(self, other: Self) -> u8 { self.0 + other.0 } // this can never overflow because we know 100 + 100 < 255
}

But this will result an invalid struct and break the invariant and causing panics.

let bad: UIntLessThan100 = Decode::decode(&mut &[255u8]).unwrap();

BoundedVec have similar issue and the solution is write a custom decode implementation. But it is tedious for large struct.

[xlc]

Could be something like this serde-rs/serde#642 (comment)

[xlc]

From @rphmeier

You could make a PR to add a trait parity_scale_codec::Validated and then have a new #[codec(validated)] annotation which applies to a struct which is #[derive(Decode)]
Add a CodecError::ValidationFailed
Then in the autogenerated Decode impl, if the codec(validated) is present, it adds the extra check
#[codec(validate_with = "fn_path")] would avoid the need for a new trait

[rphmeier]

Yes, so an example of this would be

#[derive(Decode)]
#[codec(validate_with = "check_trit")]
struct Trit(u8);

// Probably some more concrete error type in practice
fn check_trit(trit: &Trit) -> Result<(), String> {
    match trit.0 {
        0 | 1 | 2 => Ok(()),
        _ => Err("byte-pattern out of range for trit".into()),
    }
}

If the "validation function" fails, that would manifest as a CodecError at a higher level.