add numpy.partition like functionality

[NewBornRustacean]

add numpy.partition like functionality

Overview

What's up awesome ndarray community! I searched through old issues and found the related issues below.
There were couple of discussions and PRs, but currently, it seems like ndarray and ndarray-stats lack several key sorting and searching operations that are available in NumPy.

Start with partitioning: API is gonna be like..

impl<A, D> ArrayBase<S, D>
where
    A: Ord,
    D: Dimension,
{
    /// Partially sorts the array around the k-th element.
    /// 
    /// Similar to NumPy's `partition`. Returns a new array where:
    /// - All elements smaller than the k-th element are to its left
    /// - All elements equal or greater than the k-th element are to its right
    /// - The ordering within each partition is undefined
    pub fn partition(&self, kth: usize) -> Array<A, D> { ... }

    /// Returns the indices that would partition the array.
    /// 
    /// Similar to NumPy's `argpartition`. Returns indices that would partition
    /// the array around the k-th element.
    pub fn argpartition(&self, kth: usize) -> Array<usize, D> { ... }
}

Note

• I searched the codebase thoroughly and started implementing this feature, but there might already be an existing implementation that I missed. If that's the case, please let me know!
• will open draft PR soon!

Related Issues

• Argmax/margmin for Array1 #416
• Feedback on Ndarray/Ndarray-linalg (from Reddit) #649
• Implementing an efficient Argsort #1145
• (might be more)

References

• NumPy documentation for argsort
• NumPy documentation for partition
• NumPy documentation for argwhere
• NumPy documentation for where

[NewBornRustacean]

now partition is added. will be back with argpartition or partition_multiple 😃