optirustic 1.2.3

pub use algebra::{
    dot_product, perpendicular_distance, solve_linear_system, vector_magnitude,
    LinearSolverTolerance,
};
pub use fast_non_dominated_sort::{fast_non_dominated_sort, NonDominatedSortResults};
pub use reference_points::{DasDarren1998, NumberOfPartitions, TwoLayerPartitions};
use std::cmp::Ordering;
use std::collections::{BTreeSet, HashMap};
use std::hash::{BuildHasher, Hash};

use crate::core::OError;

mod algebra;
mod fast_non_dominated_sort;
mod reference_points;

/// Define the sort type
#[derive(PartialEq)]
pub enum Sort {
    /// Sort values in ascending order
    Ascending,
    /// Sort values in descending order
    Descending,
}

/// Returns the indices that would sort an array in ascending order.
///
/// # Arguments
///
/// * `data`: The vector to sort.
/// * `sort_type`: Specify whether to sort in ascending or descending order.
///
/// returns: `Vec<usize>`. The vector with the indices.
pub fn argsort(data: &[f64], sort_type: Sort) -> Vec<usize> {
    let mut indices = (0..data.len()).collect::<Vec<_>>();
    indices.sort_by(|a, b| data[*a].total_cmp(&data[*b]));

    if sort_type == Sort::Descending {
        indices.reverse();
    }
    indices
}

/// Calculate the vector minimum value.
///
/// # Arguments
///
/// * `v`: The vector.
///
/// returns: `Result<f64, OError>`
///
/// # Examples
pub fn vector_min(v: &[f64]) -> Result<f64, OError> {
    Ok(*v
        .iter()
        .min_by(|a, b| a.total_cmp(b))
        .ok_or(OError::Generic(
            "Cannot calculate vector min value".to_string(),
        ))?)
}

/// Calculate the vector maximum value.
///
/// # Arguments
///
/// * `v`: The vector.
///
/// returns: `Result<f64, OError>`
pub fn vector_max(v: &[f64]) -> Result<f64, OError> {
    Ok(*v
        .iter()
        .max_by(|a, b| a.total_cmp(b))
        .ok_or(OError::Generic(
            "Cannot calculate vector max value".to_string(),
        ))?)
}

/// Returns `true` if two arrays are element-wise equal within a tolerance. This behaves as the
/// numpy implementation at <https://numpy.org/doc/stable/reference/generated/numpy.allclose.html>.
///
/// # Arguments
///
/// * `a`: First vector to compare.
/// * `b`: Second vector to compare.
/// * `r_tol`: The relative tolerance parameter.
/// * `a_tol`: The absolute tolerance parameter.
///
/// returns: `bool`
pub fn all_close(a: &[f64], b: &[f64], r_tol: Option<f64>, a_tol: Option<f64>) -> bool {
    let r_tol = r_tol.unwrap_or(1e-05);
    let a_tol = a_tol.unwrap_or(1e-08);

    a.iter()
        .zip(b)
        .any(|(v1, v2)| (v1 - v2).abs() <= (a_tol + r_tol * v2.abs()))
}

/// Return the index of the minimum values of the vector.
///
/// # Arguments
///
/// * `vector`: The vector.
///
/// returns: `(usize, f64)`: The index of minimum value and the minimum value.
pub fn argmin(vector: &[f64]) -> (usize, f64) {
    let mut min_value = f64::INFINITY;
    let mut min_index: usize = 0;
    for (index, value) in vector.iter().enumerate() {
        if *value < min_value {
            min_value = *value;
            min_index = index;
        }
    }

    (min_index, min_value)
}

pub fn hasmap_eq_with_nans<K, S>(a: &HashMap<K, f64, S>, b: &HashMap<K, f64, S>) -> bool
where
    K: Eq + Hash,
    S: BuildHasher,
{
    if a.len() != b.len() {
        return false;
    }

    a.iter().all(|(key, value)| {
        b.get(key)
            .map_or(false, |v| (v.is_nan() && value.is_nan()) || (*value == *v))
    })
}

/// Return the vector items and its index corresponding to the minimum value returned by the closure.
///
/// # Arguments
///
/// * `vector`: The vector.
/// * `f`: The closure that receives the vector item and its index and returns a number to minimise.
///
/// returns: `Option<(usize, &I)>`: The vector index and its value or `None` if `vector` is
/// empty.
pub fn argmin_by<I, F>(vector: &[I], f: F) -> Option<(usize, &I)>
where
    I: Sized,
    F: FnMut((usize, &I)) -> f64,
{
    #[inline]
    fn key<T>(
        mut f: impl FnMut((usize, &T)) -> f64,
    ) -> impl FnMut((usize, &T)) -> (f64, (usize, &T)) {
        move |(index, it)| (f((index, it)), (index, it))
    }

    #[inline]
    fn compare<T>((x_p, _): &(f64, T), (y_p, _): &(f64, T)) -> Ordering {
        x_p.total_cmp(y_p)
    }

    let (_, (index, x)) = vector.iter().enumerate().map(key(f)).min_by(compare)?;
    Some((index, x))
}

/// Search for an element in an iterator, returning its index.
///
/// # Arguments
///
/// * `vector`: The vector.
/// * `item`: The item to look for.
///
/// returns: `Option<usize>`
pub fn index_of<I: Sized + PartialEq>(vector: &[I], item: &I) -> Option<usize> {
    vector.iter().position(|i| i == item)
}

/// Check whether a vector contains unique elements.
///
/// # Arguments
///
/// * `vector`: The vector.
/// * `f`: The function that returns the keys to check.
///
/// returns: `bool`
pub fn has_unique_elements_by_key<I: Sized, J: Sized + Ord, F: Fn(&I) -> J>(
    vector: &[I],
    f: F,
) -> bool {
    let values: Vec<J> = vector.iter().map(f).collect();
    let mut unique = BTreeSet::new();
    values.iter().all(move |x| unique.insert(x))
}

#[cfg(test)]
mod test {
    use crate::utils::{argmin_by, argsort, Sort};

    #[test]
    fn test_argsort() {
        let vec = vec![99.0, 11.0, 456.2, 19.0, 0.5];

        assert_eq!(argsort(&vec, Sort::Ascending), vec![4, 1, 3, 0, 2]);
        assert_eq!(argsort(&vec, Sort::Descending), vec![2, 0, 3, 1, 4]);
    }

    #[test]
    fn test_armin_by() {
        #[derive(PartialEq, Debug)]
        struct A {
            distance: f64,
            other: f64,
        }

        let values = [10.0, -99.0, 55.2, -1.0];
        let a_values: Vec<A> = values
            .iter()
            .map(|v| A {
                distance: *v,
                other: 0.0,
            })
            .collect();
        let (min_index, min_distance) = argmin_by(&a_values, |(_, a)| a.distance).unwrap();
        assert_eq!(min_index, 1);
        assert_eq!(min_distance, &a_values[1]);
    }
}