Project Creation

UDF projects are classified into the following types. Create UDF projects based on application scenarios.

Using UDF: In this scenario, users do not call third-party dependent libraries or NN functions that have been implemented. Instead, they compile UDF-related files to implement user-defined functions.
Calling NN functions: In this scenario, the functions required by users can be carried by the existing NN functions. Users can simply call the NN functions through UDF.
Calling a third-party dependency library: In this scenario, the required functions already exist in the third-party dependency library. Users can call related functions in the third-party dependency library through UDF.

Users can directly download the sample project for development. The following UDF sample project directory cover three scenarios. Download the corresponding examples based on the application scenarios, and develop UDF deliverables in the corresponding directories based on the following rules:

├── 01_udf_add
│ ├── CMakeLists.txt    
│ └── add_flow_func.cpp // add function defined by a single-func API
├── 02_udf_call_add_nn
│ ├── CMakeLists.txt
│ └── call_nn_flow_func.cpp // tensorflow model called by UDF
├── 03_udf_add_multi_func
│ ├── CMakeLists.txt
│ └── add_flow_func.cpp // add function defined by multi-func APIs, which can be used together with 04.
├── 04_control_func
│ ├── CMakeLists.txt
│ └── control_func.cpp // activates certain add function defined by multi-func APIs in 03 based on the inputs, which can be used together with 03.
└── README.md

Parent topic: Development Preparation