Function parse_einsum_input 

pub fn parse_einsum_input(
    subscripts: &str,
    operands: &[TensorShapeType],
) -> Result<(String, String, Vec<TensorShapeType>), String>

A reproduction of einsum’s parsing logic in Rust (originally from Python/Numpy).

Arguments

• subscripts - The einsum subscript string (e.g., “ij,jk->ik”).
• shapes - Array shapes or tensors to contract.

Returns

• input_subscripts Parsed input subscripts (e.g., “ij,jk”).
• output_subscript Parsed output subscript (e.g., “ik”).
• operands The operands (shapes) to use in contraction.

Examples

let a_shape = vec![4, 4];
let b_shape = vec![4, 4, 4];
let subscripts = "...a, ...a -> ...";
let (input_subscripts, output_subscript, operands) = parse_einsum_input(subscripts, &[a_shape, b_shape]).unwrap();
assert_eq!(input_subscripts, "da,cda");
assert_eq!(output_subscript, "cd");
assert_eq!(operands, vec![vec![4, 4], vec![4, 4, 4]]);

Errors

• Subscripts contain invalid ‘->’ patterns (larger than one ‘->’, or contains separated - or > characters).
• Malformed ellipses (“…”).
• Output characters are non-unique or missing from inputs.
• Operand count doesn’t match subscript terms.