Reputation: 2013
I'm trying to make an assertion that the type returned by a function in someone else's crate is big enough, and don't feel like changing my code if they change theirs to something that's still ok.
Here's a simplified version of what I'm trying to do (playground):
const NANOSECONDS_IN_DAY:f64 = 8.64E13;
trait HasMax<T> {
fn max_value() -> T;
}
fn max_of<T:HasMax>(_: &T) -> T {
T::max_value()
}
fn main() {
let mv = max_of(&1f64);
assert!(mv > NANOSECONDS_IN_DAY, "type is too small");
}
However I get the following error:
<anon>:5:13: 5:19 error: wrong number of type arguments: expected 1, found 0 [E0243]
<anon>:5 fn max_of<T:HasMax>(_: &T) -> T {
^~~~~~
Upvotes: 1
Views: 506
Reputation: 2013
extern crate num;
extern crate time;
#[cfg(test)]
mod tests {
use num::Bounded;
use time::{self,Tm};
#[test]
fn size_test() {
let NANOSECONDS_IN_DAY:f64 = 8.64E13;
fn max_of<T:Bounded>(_: T) -> T {
T::max_value()
}
let type_v = time::Duration::zero().num_nanoseconds().unwrap();
let mv = max_of(type_v);
assert!(mv as f64 > NANOSECONDS_IN_DAY, "type is too small");
}
}
This works.
Upvotes: 1
Reputation: 213817
Trying to make your code work in spite of unknown future breaking API changes in a library you are using is misguided at best. The short answer is that if the API of a library you are using changes, it is better that your code will fail to compile instead of failing in some more subtle or difficult to detect way. But I can answer why your code isn't compiling.
The problem is in your trait definition, which uses generics. Your trait should be HasMax
, not HasMax<T>
.
const NANOSECONDS_IN_DAY:f64 = 8.64E13;
trait HasMax {
fn max_value() -> Self;
}
fn max_of<T:HasMax>(_: &T) -> T {
T::max_value()
}
impl HasMax for f64 {
fn max_value() -> Self {
std::f64::MAX
}
}
fn main() {
let mv = max_of(&1f64);
assert!(mv > NANOSECONDS_IN_DAY, "type is too small");
}
Upvotes: 2