classify 0.2.2

use num_traits::ToPrimitive;

/// Represents a unique value found within a sorted dataset along with the indices of its first and last occurrences in the dataset
pub struct UniqueVal {
    pub val: f64,
    pub first: usize,
    pub last: usize,
}

/// Represents a single bin in a classification, including the bin's lowest (inclusive) and highest (exclusive) values and the number of points within it
pub struct Bin {
    pub bin_start: f64,
    pub bin_end: f64,
    pub count: u64,
}

impl PartialEq for Bin {
    fn eq(&self, other: &Self) -> bool {
        let starts_eq: bool = self.bin_start == other.bin_start;
        let ends_eq: bool = self.bin_end == other.bin_end;
        let counts_eq: bool = self.count == other.count;
        starts_eq && ends_eq && counts_eq
    }
}

/// Represents a full classification, which is a collection of Bin objects
pub type Classification = Vec<Bin>;

/// Translates generic numeric vectors to Vec<f64> using the ToPrimitive trait from the num crate
///
/// # Arguments
///
/// * `data` - A reference to a vector of generic type T where T implements the ToPrimitive trait
pub fn to_vec_f64<T: ToPrimitive>(data: &[T]) -> Vec<f64> {
    let mut result: Vec<f64> = vec![];
    for item in data {
        result.push(item.to_f64().unwrap());
    }
    result
}

/// Populates an empty vector of UniqueVal objects for each unique value in the dataset in the format (value, first occurrence index, last occurrence index)
///
/// # Arguments
///
/// * `unique_val_map` - A mutable reference to an empty vector of UniqueVals
/// * `vals` - A reference to the data (sorted, ascending) to use in populating unique_val_map
pub fn create_unique_val_mapping(unique_val_map: &mut Vec<UniqueVal>, vals: &[f64]) {
    unique_val_map.clear();
    let mut idx: i64 = -1;

    for (i, item) in vals.iter().enumerate() {
        if unique_val_map.is_empty() {
            idx += 1;
            unique_val_map.push(UniqueVal {
                val: *item,
                first: i,
                last: i,
            });
        } else if unique_val_map[idx as usize].val != *item {
            unique_val_map[idx as usize].last = i - 1;
            idx += 1;
            unique_val_map.push(UniqueVal {
                val: *item,
                first: i,
                last: i,
            });
        }
    }
}

/// Adjusts break indices from unique value breaks to normal breaks, accounting for repeated values, given unique value breaks, a unique value map, and an empty vector for normal breaks
///
/// # Arguments
///
/// * `u_val_breaks` - A reference to a vector of uniquely valued breaks (sorted, ascending)
/// * `u_val_map` - A reference to a map of unique values to their first and last occurrences in the dataset
/// * `normal_breaks` - A mutable reference to an empty vector to populate with adjusted break indices
pub fn unique_to_normal_breaks(
    u_val_breaks: &Vec<usize>,
    u_val_map: &[UniqueVal],
    normal_breaks: &mut Vec<usize>,
) {
    if normal_breaks.len() != u_val_breaks.len() {
        normal_breaks.resize(u_val_breaks.len(), 0);
    }
    for i in 0..u_val_breaks.len() {
        normal_breaks[i] = u_val_map[u_val_breaks[i]].first;
    }
}

/// Returns a Classification object given a set of breaks between bins and the original dataset
///
/// # Arguments
///
/// * `breaks` - A reference to a vector of breaks (f64) generated through any classification function or manually
/// * `data` - A reference to a vector of unsorted data points (f64) used to count the points in each bin
///
/// # Examples
///
/// ```
/// use classify::{breaks_to_classification};
/// use classify::{Classification, Bin};
///
/// let data: Vec<f64> = vec![1.0, 2.0, 4.0, 5.0, 7.0, 8.0];
/// let breaks: Vec<f64> = vec![2.0, 5.0];
///
/// let result: Classification = breaks_to_classification(&breaks, &data);
/// let expected: Classification = vec![
///     Bin{bin_start: 1.0, bin_end: 2.0, count: 1},
///     Bin{bin_start: 2.0, bin_end: 5.0, count: 2},
///     Bin{bin_start: 5.0, bin_end: 8.0, count: 3}
/// ];
///
/// assert!(result == expected);
/// ```
pub fn breaks_to_classification<T: ToPrimitive>(breaks: &Vec<f64>, data: &[T]) -> Classification {
    let data = to_vec_f64(data);

    let mut min_value = data[0];
    let mut max_value = data[0];
    for item in &data {
        if *item < min_value {
            min_value = *item;
        }
        if *item > max_value {
            max_value = *item;
        }
    }

    let mut bounds: Vec<f64> = vec![min_value];
    for item in breaks {
        bounds.push(*item);
    }
    bounds.push(max_value);

    let mut results: Classification = vec![];
    for i in 0..(bounds.len() - 1) {
        results.push(Bin {
            bin_start: bounds[i],
            bin_end: bounds[i + 1],
            count: 0,
        });
    }

    let num_bins = results.len();
    for bin in results.iter_mut().take(num_bins) {
        for item in &data {
            if bin.bin_start <= *item && *item < bin.bin_end {
                bin.count += 1;
            }
        }
    }
    results[num_bins - 1].count += 1;

    results
}

/// Returns an Option<usize> containing the index of the Bin within which a value should fall given the value and a Classification (returns None if the value is outside of the Classification's range)
///
/// # Arguments
///
/// * `val` - Data value to classify
/// * `class` - Classification object
///
/// # Examples
///
/// ```
/// use classify::{classify_val};
/// use classify::{Classification, Bin};
///
/// let vals: Vec<f64> = vec![0.0, 1.5, 3.5];
/// let class: Classification = vec![
///     Bin{bin_start: 0.0, bin_end: 1.0, count: 5},
///     Bin{bin_start: 1.0, bin_end: 2.0, count: 5},
///     Bin{bin_start: 2.0, bin_end: 3.0, count: 5}
/// ];
///
/// let mut results: Vec<Option<usize>> = vec![];
/// for val in vals {results.push(classify_val(val, &class))}
///
/// assert_eq!(results, vec![Some(0), Some(1), None])
/// ```
pub fn classify_val(val: f64, class: &Classification) -> Option<usize> {
    if val < class[0].bin_start || val > class[class.len() - 1].bin_end {
        return None;
    }
    for (i, _item) in class.iter().enumerate() {
        if class[i].bin_start <= val && val < class[i].bin_end {
            return Some(i);
        }
    }
    Some(class.len() - 1) // Accounts for case where val is maximum within dataset
}