shape_optional.h#

Shape-inference functions for ONNX operators in the optional family.

namespace onnx_light

Alias that makes onnx-light headers compatible with code that references ONNX_LIGHT_NAMESPACE (the macro used in the standard onnx package).

Set to ONNX_LIGHT_NAMESPACE so both names resolve to the same namespace.

Symbol-visibility attribute for the public onnx-light C++ API.

Defined as empty because onnx-light does not require explicit __declspec(dllexport) or __attribute__((visibility("default"))) annotations — visibility is controlled at the shared-library level. The macro is provided so that vendored ONNX headers that decorate their declarations with ONNX_API compile without modification.

Namespace alias so that ONNX C++ code (and consumers such as onnxruntime) that refers to the literal onnx namespace — rather than the ONNX_NAMESPACE macro — resolves to the onnx-light namespace. The standard onnx package lives in namespace onnx; onnx-light uses onnx_light (via ONNX_LIGHT_NAMESPACE), so this alias keeps onnx-light a true drop-in. It is only introduced when the onnx-light namespace differs from onnx.

namespace onnx_optim
namespace shapes
namespace optional#

Functions

void ComputeShapeOptional(ShapesContext &ctx, const NodeProto &node)#

Computes the output :cpp:class:OptimTensor of an Optional node and stores it in ctx.

Optional (since opset 15) wraps a value into an optional-type output. The wrapped element type is determined either by the input value (when an input is provided) or by the type TypeProto attribute (when no input is provided). Since :cpp:class:OptimTensor does not model optional or sequence types, this implementation only supports the tensor-element path: the output descriptor mirrors the dtype and shape of the wrapped tensor.

Supported cases:

  • node has one input: the output dtype and shape are copied from the input descriptor stored in ctx.

  • node has no input and the type attribute wraps a tensor type (either directly or as the elem_type of an optional_type): the output dtype and shape are taken from the attribute.

Parameters:

  • ctx – In/out context. When the node has an input, ctx must already contain an entry for it; on return ctx contains an entry for node.output(0).

  • node – The Optional NodeProto whose output should be described. node.op_type() must be "Optional" and node must declare exactly one output.

Throws:

  • std::invalid_argument – if node.op_type() is not "Optional", if node has no output or more than one input, if the type attribute wraps a sequence or non-tensor element type, or if neither an input nor a valid type attribute is provided.

  • std::out_of_range – if the input name is missing from ctx.

void ComputeShapeOptionalGetElement(ShapesContext &ctx, const NodeProto &node)#

Computes the output descriptor of an OptionalGetElement node and stores it in ctx.

OptionalGetElement (since opset 15 in the ai.onnx domain) extracts the element from an optional-type input. Since opset 18 the operator also accepts non-optional tensor or sequence inputs as a no-op. Because :cpp:class:OptimTensor does not model optional values, this implementation forwards the input descriptor verbatim:

  • if the input name is bound to an :cpp:class:OptimSequence in ctx, the output is registered as the same sequence;

  • otherwise the input must be bound to an :cpp:class:OptimTensor and the output is registered as a tensor with the same dtype and shape.

Parameters:

  • ctx – In/out context. Must already contain an entry for node.input(0); on return it also contains an entry for node.output(0).

  • node – The OptionalGetElement NodeProto.

Throws:

  • std::invalid_argument – if node.op_type() is not "OptionalGetElement", if node does not declare exactly one input and one output, or if the input name is empty.

  • std::out_of_range – if the input name is missing from ctx.

void ComputeShapeOptionalHasElement(ShapesContext &ctx, const NodeProto &node)#

Computes the output descriptor of an OptionalHasElement node and stores it in ctx.

OptionalHasElement (since opset 15 in the ai.onnx domain) produces a scalar boolean tensor indicating whether the input optional contains an element. Since opset 18 the input may also be a non-optional tensor or sequence, and the input may be omitted entirely (in which case the output is false). The output is always a scalar :cpp:class:OptimTensor of dtype :cpp:enumerator:TensorType::kBool.

Parameters:

  • ctx – In/out context. On return it contains an :cpp:class:OptimTensor entry for node.output(0).

  • node – The OptionalHasElement NodeProto.

Throws:

std::invalid_argument – if node.op_type() is not "OptionalHasElement", if node declares more than one input or does not declare exactly one output.