In Rust I want to assign a given value to all elements of nested vectors

Question

I am trying to build a function similar to np.zeros from Python for Rust.

I would like to do something like:

trait VectorHelper {
  fn fill_with<A>(&mut self, value: A) -> &mut Self;
}
impl<T: VectorHelper> VectorHelper for Vec<T> {
  fn fill_with<A>(&mut self, value: A) -> &mut Self {
    for elem in self.iter() {
      if elem.is_vec() {
        elem.fill_with(value);
      } else {
        *elem = value;
      }
    }
    self
  }
}

But I do not know how to get this to work, because there is no general conversion for A to T. What can I do to make the compiler accept this assignment: "*elem = value"?

Answer 1

While the exact implementation like you want it isn't possible you can still handle all the additional layers but the first by a blanket implementation:

impl<V: Clone, T: VectorHelper<V>> VectorHelper<V> for Vec<T> {
  fn fill_with(&mut self, value: V) -> &mut Self {
    for elem in &mut *self {
        elem.fill_with(value.clone());
    }
    self
  }
}

Then for each base type you can implement the trait once (possibly by a macro to reduce duplication):

macro_rules! impl_vector_helper {
    ($t:ty) => {
        impl VectorHelper<$t> for Vec<$t> {
            fn fill_with(&mut self, value: $t) -> &mut Self {
                self.fill(value);
                self
            }
        }
    };
    ($($t:ty),* $(,)?) => {
        $(impl_vector_helper!{$t})*
    };
}

impl_vector_helper! {
    u8, u16, u32, u64,
    i8, i16, i32, i64,
    f32, f64,
    &'static str,
}

You just have to move the generic from the function to the trait:

trait VectorHelper<V> {
    fn fill_with(&mut self, value: V) -> &mut Self;
}