viva_tensor - v2.1.0

viva_tensor/core/error

Centralized error types for tensor operations.

Philosophy: fail fast, fail loud, fail informatively.

Each error variant carries enough context to debug the issue without printf-debugging. ShapeMismatch tells you both shapes. IndexOutOfBounds tells you the index AND the size. No more “index out of bounds” with no context.

Why a single error type instead of operation-specific ones?

Types

TensorError

</>

All the ways tensor operations can fail. Tried to keep it minimal but expressive.

pub type TensorError {
  ShapeMismatch(expected: List(Int), got: List(Int))
  InvalidShape(reason: String)
  DimensionError(reason: String)
  BroadcastError(shape_a: List(Int), shape_b: List(Int))
  IndexOutOfBounds(index: Int, size: Int)
  DtypeError(reason: String)
}

Constructors 

  • ShapeMismatch(expected: List(Int), got: List(Int))

    Shape mismatch between two tensors.

    Example

    Trying to add [2, 3] tensor with [4, 5] tensor.

  • InvalidShape(reason: String)

    Invalid shape specification.

    Example

    Data size doesn’t match shape dimensions.

  • DimensionError(reason: String)

    Dimension-related error (axis out of bounds, etc.).

    Example

    Accessing axis 3 on a 2D tensor.

  • BroadcastError(shape_a: List(Int), shape_b: List(Int))

    Broadcasting incompatibility.

    Example

    Cannot broadcast [2, 3] with [4, 5].

  • IndexOutOfBounds(index: Int, size: Int)

    Index out of bounds.

    Example

    Accessing index 10 in tensor of size 5.

  • DtypeError(reason: String)

    Invalid dtype for operation.

    Example

    Using INT8 operation on Float32 tensor.

Values

shape_to_string

</> 
pub fn shape_to_string(shape: List(Int)) -> String

Pretty-print a shape like [2, 3, 4].

to_string

</> 
pub fn to_string(error: TensorError) -> String

Human-readable error message. Useful for debugging.

✨ Search Document