Struct CompressedSparse 

pub struct CompressedSparse<T> {
    pub data: Vec<T>,
    pub indices: Vec<u32>,
    pub indptr: Vec<u32>,
    pub format: SparseFormat,
    pub shape: (usize, usize),
}

A sparse matrix in compressed row or column form.

Fields are public so buffers can go straight to a kernel, but only CompressedSparse::validate guarantees consistency. Prefer the constructors.

Fields

data: Vec<T>

Non-zero values, ordered by the compressed axis then by indices.

indices: Vec<u32>

Index along the non-compressed axis for each value in data.

indptr: Vec<u32>

Offsets into data/indices, one per compressed slice plus a trailing total.

format: SparseFormat

Whether indptr walks rows or columns.

shape: (usize, usize)

Logical shape as (n_rows, n_cols), independent of format.

Implementations

impl<T: Copy> CompressedSparse<T>

pub fn from_parts( data: Vec<T>, indices: Vec<u32>, indptr: Vec<u32>, format: SparseFormat, shape: (usize, usize), ) -> Result<Self, EdgeErrors>

Assembles a matrix from its parts without copying.

Params

• data - Non-zero values
• indices - Index along the non-compressed axis per value
• indptr - Offsets per compressed slice, length outer_len + 1
• format - Which axis indptr walks
• shape - Logical (n_rows, n_cols)

Returns

The matrix, or EdgeErrors::MalformedSparse if the parts disagree.

pub fn validate(&self) -> Result<(), EdgeErrors>

Checks the structural invariants.

data and indices must match in length, indptr must have the right length, be non-decreasing and end at the non-zero count, and every index must be inside the non-compressed axis.

Returns

Ok(()) when consistent, otherwise EdgeErrors::MalformedSparse.

pub fn nrows(&self) -> usize

Number of rows.

Returns

The first component of the shape.

pub fn ncols(&self) -> usize

Number of columns.

Returns

The second component of the shape.

pub fn nnz(&self) -> usize

Number of stored non-zeros.

Returns

The length of data.

pub fn outer_len(&self) -> usize

Length of the compressed axis, that is, how many slices indptr walks.

Returns

Row count for CSR, column count for CSC.

pub fn inner_len(&self) -> usize

Length of the non-compressed axis, that is, the range indices spans.

Returns

Column count for CSR, row count for CSC.

pub fn outer(&self, outer: usize) -> (&[u32], &[T])

Borrows one slice along the compressed axis.

The per-gene accessor: for a gene-major matrix, that gene’s non-zero cells and counts, no copy.

Params

• outer - Index along the compressed axis

Returns

The indices and values of that slice, in indices order.

pub fn transpose(self) -> Self

The transposed matrix, for free.

A CSR over (m, n) describes the same bytes as a CSC over (n, m).

Returns

A matrix representing the transpose, sharing this one’s layout.

pub fn convert(&self) -> Self

where T: Default,

The same matrix stored in the other format.

A pass over the non-zeros plus a counting sort. Sequential; meant to be called once at ingest.

Returns

An equivalent matrix with the opposite format.

pub fn from_dense<F>( dense: &[T], n_rows: usize, n_cols: usize, format: SparseFormat, is_zero: F, ) -> Result<Self, EdgeErrors>

where F: Fn(&T) -> bool,

Builds a compressed matrix from a dense row-major buffer.

For tests and small inputs.

Params

• dense - n_rows * n_cols values, row-major
• n_rows - Number of rows
• n_cols - Number of columns
• format - Storage format to build
• is_zero - Predicate deciding which values to drop

Returns

The compressed matrix, or EdgeErrors::LengthMismatch if dense does not match the stated shape.

pub fn to_dense(&self, zero: T) -> Vec<T>

Materialises a dense row-major buffer.

Tests and interop only.

Params

• zero - Value to fill the structural zeros with

Returns

An n_rows * n_cols row-major vector.

Trait Implementations

impl<T: Clone> Clone for CompressedSparse<T>

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<T: Debug> Debug for CompressedSparse<T>

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

Formats the value using the given formatter.