Struct SparseAxis 

pub struct SparseAxis<T, U = T> {
    pub indices: Vec<usize>,
    pub data: Vec<T>,
    pub data_2: Option<Vec<U>>,
    pub cs_type: CompressedSparseFormat,
    pub len: usize,
}

Generate structure to store sparse rows or columns

Fields

indices: Vec<usize>

The indices of the values/non-zero positions

data: Vec<T>

The values in the row/columns

data_2: Option<Vec<U>>

An optional second data layer

cs_type: CompressedSparseFormat

Is the data stored in Csr or Csc. Csr -> sparse row; Csc -> sparse column

len: usize

Total values in that dimension

Implementations

impl<T, U> SparseAxis<T, U>

where T: BixverseNumeric, U: BixverseNumeric,

pub fn new_csc( indices: Vec<usize>, data: Vec<T>, data_2: Option<Vec<U>>, len: usize, ) -> SparseAxis<T, U>

Generate a new SparseAxis in CSC format

Params

• indices - The indices of the values/non-zero positions
• data - The values in the row/columns
• data_2 - An optional second data layer
• len - Number of rows in this sparse column

pub fn get_indices_data_2(&self) -> Result<(&[usize], &[U]), CoreErrors>

Get references to the indices and second layer

Returns

A tuple of (indices, data_2)

pub fn from_vecs_to_csc( indices: Vec<T>, values: Vec<U>, ) -> Result<SparseAxis<T, U>, CoreErrors>

where T: Into<usize>,

Construct a SparseAxis in CSC format from index and value vectors.

Stores indices (converted to usize) in self.indices, leaves data empty, and places values into data_2. Useful for building single-target test fixtures and lightweight sparse columns where only the float layer is needed.

Params

• indices - Non-zero positions (must be convertible to usize).
• values - Corresponding values, stored in data_2.

Panics

Panics if indices and values differ in length.