.. blogpost::
    :title: Xop, easy to create onnx graph
    :keywords: tips, tensorflow, tensorflow-onnx
    :date: 2022-02-27
    :categories: xop, onnx

    :epkg:`onnx` package has a very verbose API to create ONNX
    graph. Could you imagine a user to directly write the syntax tree
    of a program instead of some python code? Creating a ONNX graph is
    very similar to that task except ONNX language is more simple
    than python.

    We could start writing a compiler for ONNX language but it should be
    defined first. Another option consists in using an existing API,
    such as :epkg:`numpy` API (see :ref:`l-numpy2onnx-tutorial`).
    But it is not always easy to keep the same simplicity when numpy is
    not strongly typed and ONNX is. Another direction is to implement
    :epkg:`ONNX Operators` as function. Adding an operator into a graph
    becomes similar to a function call. This API is introduced in
    :ref:`l-xop-api`.