Function einsum_native_subscripts 

pub fn einsum_native_subscripts(
    subscripts: &str,
    operands: &[&Tensor],
) -> Result<Tensor>

Execute a cached einsum over borrowed native tensors using string subscripts.

This is the shared string-notation entry: it parses subscripts into the numeric labels that einsum_native_tensors consumes, checks that the equation describes exactly the supplied operands, and then delegates the contraction. Use it instead of re-implementing the parsing plus label mapping in each crate; use einsum_native_tensors directly when the labels are already numeric.

Arguments

• subscripts - NumPy/PyTorch-style einsum notation such as "ij,jk->ik" or "ii->" for a trace. Parenthesized contraction order is rejected; see [tenferro_einsum::Subscripts::parse].
• operands - One native tensor per comma-separated input term, in term order. Ranks must match the term lengths and repeated labels must have equal extents.

Returns

The contracted native tensor in the promoted common dtype of the operands.

Errors

Returns an error when the notation parser rejects subscripts as invalid notation, when the operand count does not match the number of comma-separated input terms, when an input term’s label count does not match its operand rank, when repeated or contracted labels have incompatible shapes, or when dtype promotion or the backend contraction fails.

Compatibility frontend

Like einsum_native_tensors, this runs on the retained process-global (compatibility) frontend of tensor4all-rs#859: it selects the default execution context and pays one graph dispatch per call. It is meant for cold or per-stage paths, not per-element hot loops; use CpuExecutionContext sessions for those.

Examples

use tensor4all_tensorbackend::einsum_native_subscripts;
use tenferro::Tensor as NativeTensor;

// Column-major A = [[1, 2], [3, 4]] and B = [[5, 6], [7, 8]].
let lhs = NativeTensor::from_vec_col_major(vec![2, 2], vec![1.0_f64, 3.0, 2.0, 4.0])?;
let rhs = NativeTensor::from_vec_col_major(vec![2, 2], vec![5.0_f64, 7.0, 6.0, 8.0])?;
let product = einsum_native_subscripts("ij,jk->ik", &[&lhs, &rhs])?;
assert_eq!(product.shape(), &[2, 2]);
assert_eq!(product.as_slice::<f64>()?, &[19.0, 43.0, 22.0, 50.0]);

let trace = einsum_native_subscripts("ii->", &[&lhs])?;
assert_eq!(trace.as_slice::<f64>()?, &[5.0]);