Native C++ Library

We also provide C++ library for you to link to in case:

• You are only interested in one or two specific algorithms implemented in some operators.

• You don’t like the concept or methodology that PyIS imposes on you, e.g., defination of a model, deployment with libtorch or onnxruntime. It is a pity. But we understand that and are with you.

Use the Operator in C++

// Copyright (c) Microsoft Corporation. All rights reserved.
// Licensed under the MIT license.

#include <cstdio>
#include "pyis/ops/text/cedar_trie.h"

int main() {
    pyis::ops::CedarTrie trie;
    trie.Insert("the answer", 42);
    auto answer = trie.Lookup("the answer");
    printf("the answer is %d", answer.value());
    return 0;
}

Build the Example as a C++ Project with CMake

All core algorithms are implemented in the CMake target pyis_operators.

cmake_minimum_required(VERSION 3.12 FATAL_ERROR)

project(cpp_library_test VERSION 0.1)

include(FetchContent)
FetchContent_Declare(
    pyis
    GIT_REPOSITORY https://github.com/microsoft/python-inference-script
    GIT_TAG main)
set(PYTHON_BACKEND
    OFF
    CACHE BOOL "" FORCE)
set(TORCH_BACKEND
    OFF
    CACHE BOOL "" FORCE)
set(ONNX_BACKEND
    OFF
    CACHE BOOL "" FORCE)

add_executable(cpp_library_test cpp_library_test.cpp)
target_link_libraries(cpp_library_test PRIVATE pyis_operators)

To optimize the binary size, PyIS allows you to disable C++ exceptions or RTTI.

# disable C++ exceptions
set(PYIS_NO_EXCEPTIONS
    ON
    CACHE BOOL "" FORCE)
# disable C++ RTTI
set(PYIS_NO_RTTI
    ON
    CACHE BOOL "" FORCE)

Run

cmake -S . -B build -A x64
cmake --build build

Now we can run the successfully built executable:

> .\build\Debug\cpp_library_test.exe
the answer is 42