Flink: Write tuples with CSV header into file

问题

I did some data processing using Flink (1.7.1 with Hadoop). At the end I'd like to write the dataset consisting of 2-tuples into a file. Currently, I am doing it like this:

<Tuple2<Integer, Point>> pointsClustered = points.getClusteredPoints(...);
pointsClustered.writeAsCsv(params.get("output"), "\n", ",");

However, I would like to have the CSV headers written into the first line. The Flink's Javadoc API doesn't state any options for this. Furthermore, I couldn't find any solution googling for it.

Could you kindly advise on how to accomplish that. Thanks a lot!

回答1:

Flink's own CsvOutputFormat does not support this functionality. What you could do is to extend the CsvOutputFormat and override the open method which writes the header when the format is opened. Then you would use DataSet#output to specify the newly created output format:

public static void main(String[] args) throws Exception {
    final ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

    DataSource<Integer> input = env.fromElements(1, 2, 3);
    DataSet<Tuple3<Integer, String, Double>> result = input.map((MapFunction<Integer, Tuple3<Integer, String, Double>>) integer -> Tuple3.of(integer, integer.toString(), 42.0));

    Path outputPath = new Path("hdfs:///foobar");
    result.output(new MyCsvOutputFormat(outputPath));

    env.execute();
}

private static class MyCsvOutputFormat<T extends Tuple> extends CsvOutputFormat<T> {

    public MyCsvOutputFormat(Path outputPath) {
        super(outputPath);
    }

    @Override
    public void open(int taskNumber, int numTasks) throws IOException {
        try (PrintWriter wrt = new PrintWriter(stream)) {
            wrt.println("Foo|bar|foobar");
        }
        super.open(taskNumber, numTasks);
    }
}

来源：https://stackoverflow.com/questions/54530755/flink-write-tuples-with-csv-header-into-file