Struct SparseTensor 

pub struct SparseTensor<const N: usize> {
    pub shape: [usize; N],
    pub entries: Vec<SparseTensorEntry<N>>,
}

Sparse N-dimensional tensor in coordinate (COO) format.

The tensor stores only its non-zero entries together with their multi-indices. Entries with the same index can be present multiple times; call SparseTensor::canonicalize to merge them.

Type parameters

• N – Tensor order (number of dimensions), fixed at compile time.

Fields

shape: [usize; N]

Shape of the tensor (length-N array of dimension sizes).

entries: Vec<SparseTensorEntry<N>>

Non-zero entries.

Implementations

impl<const N: usize> SparseTensor<N>

pub fn new(shape: [usize; N]) -> Self

Create an empty sparse tensor with the given shape.

pub fn push(&mut self, indices: [usize; N], value: f64)

Add an entry to the tensor (does not check for duplicates).

pub fn nnz(&self) -> usize

Number of stored (non-zero) entries.

pub fn size(&self) -> usize

Total number of elements (product of all dimensions).

pub fn nnz_ratio(&self) -> f64

Fraction of non-zero elements: nnz / total_elements.

Returns 0.0 if the tensor has zero total elements.

pub fn canonicalize(&mut self)

Sort entries lexicographically by index and merge duplicates by summing their values.

After calling this function, entries are in sorted order with no duplicate indices.

Trait Implementations

impl<const N: usize> Clone for SparseTensor<N>

fn clone(&self) -> SparseTensor<N>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<const N: usize> Debug for SparseTensor<N>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.