UDF C++ Examples

The following are complete examples of the implementation & execution of User-Defined Functions (UDFs) in the UDF C++ API.

Example UDF (CUDA) - Sum of Squares

Example of the computation of the sum of squares, summing the squares of input table columns and storing the result in an output table. This UDF uses CUDA libraries and must be run on a CUDA build of Kinetica.

Sample Code Block

A simple distributed example of a UDF which copies data from one set of tables to another would look like the following:

/* Sourced from:  gpudb-udf-api-cpp/proc-example/proc-example.cpp */

#include "kinetica/Proc.hpp"

#include <iostream>

int main(int argc, char *argv[])
{
    try
    {
        kinetica::ProcData* procData = kinetica::ProcData::get();
        const kinetica::ProcData::InputDataSet& inputData = procData->getInputData();
        kinetica::ProcData::OutputDataSet& outputData = procData->getOutputData();

        // Loop through input and output tables (assume the same number)
        for (size_t i = 0; i < inputData.getTableCount(); ++i)
        {
            const kinetica::ProcData::InputTable& inputTable = inputData[i];
            kinetica::ProcData::OutputTable& outputTable = outputData[i];
            outputTable.setSize(inputTable.getSize());

            // Loop through columns in the input and output tables (assume the same number and types)
            for (size_t j = 0; j < inputTable.getColumnCount(); ++j)
            {
                const kinetica::ProcData::InputColumn& inputColumn = inputTable[j];
                kinetica::ProcData::OutputColumn& outputColumn = outputTable[j];

                // For each record, copy the data from the input column to the output column
                for (size_t k = 0; k < inputTable.getSize(); ++k)
                {
                    if (inputColumn.isNull(k))
                    {
                        outputColumn.appendNull();
                        continue;
                    }

                    switch (inputColumn.getType())
                    {
                        case kinetica::ProcData::Column::BYTES: outputColumn.appendVarBytes(inputColumn.getVarBytes(k)); break;
                        case kinetica::ProcData::Column::CHAR1: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<1> >(k)); break;
                        case kinetica::ProcData::Column::CHAR2: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<2> >(k)); break;
                        case kinetica::ProcData::Column::CHAR4: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<4> >(k)); break;
                        case kinetica::ProcData::Column::CHAR8: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<8> >(k)); break;
                        case kinetica::ProcData::Column::CHAR16: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<16> >(k)); break;
                        case kinetica::ProcData::Column::CHAR32: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<32> >(k)); break;
                        case kinetica::ProcData::Column::CHAR64: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<64> >(k)); break;
                        case kinetica::ProcData::Column::CHAR128: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<128> >(k)); break;
                        case kinetica::ProcData::Column::CHAR256: outputColumn.appendValue(inputColumn.getValue<kinetica::CharN<256> >(k)); break;
                        case kinetica::ProcData::Column::DATE: outputColumn.appendValue(inputColumn.getValue<kinetica::Date>(k)); break;
                        case kinetica::ProcData::Column::DATETIME: outputColumn.appendValue(inputColumn.getValue<kinetica::DateTime>(k)); break;
                        case kinetica::ProcData::Column::DECIMAL: outputColumn.appendValue(inputColumn.getValue<int64_t>(k)); break;
                        case kinetica::ProcData::Column::DOUBLE: outputColumn.appendValue(inputColumn.getValue<double>(k)); break;
                        case kinetica::ProcData::Column::FLOAT: outputColumn.appendValue(inputColumn.getValue<float>(k)); break;
                        case kinetica::ProcData::Column::INT: outputColumn.appendValue(inputColumn.getValue<int32_t>(k)); break;
                        case kinetica::ProcData::Column::INT8: outputColumn.appendValue(inputColumn.getValue<int8_t>(k)); break;
                        case kinetica::ProcData::Column::INT16: outputColumn.appendValue(inputColumn.getValue<int16_t>(k)); break;
                        case kinetica::ProcData::Column::IPV4: outputColumn.appendValue(inputColumn.getValue<uint32_t>(k)); break;
                        case kinetica::ProcData::Column::LONG: outputColumn.appendValue(inputColumn.getValue<int64_t>(k)); break;
                        case kinetica::ProcData::Column::STRING: outputColumn.appendVarString(inputColumn.getVarString(k)); break;
                        case kinetica::ProcData::Column::TIME: outputColumn.appendValue(inputColumn.getValue<kinetica::Time>(k)); break;
                        case kinetica::ProcData::Column::TIMESTAMP: outputColumn.appendValue(inputColumn.getValue<int64_t>(k)); break;
                    }
                }
            }
        }

        // Copy any parameters from the input parameter map into the output results map (not necessary for table copying, just for illustrative purposes)
        std::map<std::string, std::string> params = procData->getParams();
        procData->getResults().insert(params.begin(), params.end());
        std::map<std::string, std::vector<uint8_t> > binParams = procData->getBinParams();
        procData->getBinResults().insert(binParams.begin(), binParams.end());

        // Inform Kinetica that the proc has finished successfully
        procData->complete();
    }
    catch (const std::exception& ex)
    {
        std::cout << "ERROR: " << ex.what() << std::endl;
        return 1;
    }

    return 0;
}