use the unsigned fixed type for balances: U64F64
#353
Conversation
| if let Some(i) = nominal_income { | ||
| validate_nominal_income(&i).map_err(|_| <Error<T>>::InvalidNominalIncome)?; | ||
| } |
There was a problem hiding this comment.
Validate only validated if the value is negative; type safety gives us this guarantee now already.
| @@ -215,26 +218,26 @@ mod tests { | |||
|
|
|||
| #[test] | |||
| fn demurrage_works() { | |||
| let bal = BalanceEntry::<u32>::new(1.into(), 0); | |||
| let bal = BalanceEntry::<u32>::new(1u32.into(), 0); | |||
There was a problem hiding this comment.
Implicit type is i32, which can obviously not be into-ed.
primitives/src/balances.rs
Outdated
| return None | ||
| } | ||
|
|
||
| Some(U64F64::from_le_bytes(source.to_le_bytes())) |
There was a problem hiding this comment.
I don't like this because byte order is not what makes this work and it confused me. what you do could be done with scale encode and decode and it would be at least obvious that it is a hacky typecast
There was a problem hiding this comment.
I agree that the LE is confusing, I have made it more trivial and just use to_bits and from_bits (much better) and also made a remark about the conversion/
I don't think it is good to to use the scale codec because it is much more inefficient than using the #[inline] bit conversions that operate on a stack instead of on the heap.
With the from_bits and to_bits conversions, I don't even consider this hacky anymore.
No description provided.