Class Vectors

Thin wrappers over std::vector<T> for the element types anira’s C++ API exposes. Each Vector* class shares the same surface (size, push, get, destroy) and inherits from VectorBase.

Note

Construct via the Class AniraWeb factory in most cases: aniraWeb.VectorSizeT([2, 1]) rather than new VectorSizeT(...). The factory threads the WASM instance through for you, and most constructors accept a plain JS array as a convenience initializer.

Common Surface

Every Vector* class supports:

• size() → number — element count.

• push(value) — append one element. Element type depends on the class (see table below). Object-vector push accepts either a wrapper instance or its raw pointer.

• get(index) — read one element. Primitive vectors return the element by value; object vectors return a non-owning raw pointer to the underlying C++ element.

• destroy() — free the underlying C++ object. See Lifecycle and Cleanup.

Most constructors also accept a JS array of initial values that is pushed eagerly during construction.

The Vector Family

Class	Underlying C++ type	Element type	get returns
VectorSizeT	std::vector<size_t>	number	number
VectorInt64T	std::vector<int64_t>	bigint	bigint
VectorFloat	std::vector<float>	number	number
VectorUnsignedInt	std::vector<unsigned int>	number	number
VectorVectorInt64	std::vector<std::vector<int64_t>>	VectorInt64T / number[] / pointer	number (raw pointer)
TensorShapeList	alias of VectorVectorInt64	same as VectorVectorInt64	same as VectorVectorInt64
VectorModelData	std::vector<anira::ModelData>	Class ModelData / pointer	n/a — push-only
VectorTensorShape	std::vector<anira::TensorShape>	Class TensorShape / pointer	n/a — push-only
VectorRingBuffer	std::vector<anira::RingBuffer>	Class RingBuffer / pointer	number (raw pointer to the element)
VectorBufferF	std::vector<anira::Buffer<float>>	Class BufferF / pointer	number (raw pointer to the element)

VectorVectorInt64 accepts several convenience forms during construction:

• number[][] — each inner array is auto-converted to bigint and wrapped in a temporary VectorInt64T that is destroyed after the inner copy is taken.

• bigint[][] — same thing, used as-is.

• (VectorInt64T | number)[] — existing wrapper instances or raw pointers; the C++ side copies each inner vector.

TensorShapeList is identical to VectorVectorInt64; it exists purely so call sites that build tensor shapes read more naturally.

Typical Use

// Primitive vectors initialised from a JS array.
const channels = aniraWeb.VectorSizeT([2, 1])
const sizes = aniraWeb.VectorSizeT([512, 0])

// Object vector: pass wrapper instances directly.
const modelData = aniraWeb.ModelData(modelBuffer, aniraWeb.InferenceBackend.ONNX)
const vectorModelData = aniraWeb.VectorModelData([modelData])

// Nested-int64 vector: nested JS array, auto-converted.
const shapes = aniraWeb.TensorShapeList([[1, 2, 512], [1]])

VectorBase

class VectorBase()

Base class shared by every typed vector wrapper.

abstract

Extends:

• BaseWrapper

The abstract base class. Every Vector* extends this and exposes the methods listed under “Common Surface” above.

size()

Number of elements in the vector

Returns:

number