Write List<CSVRecord> to HDFS

Question

I have a list of CSV records.

import org.apache.commons.csv.CSVRecord
List<CSVRecord> records = getRecords();

I format records into a stringbuilder.

private StringBuilder formatRecords(final List<CSVRecord> records, final List<String> headersList) {
        final StringBuilder sb = new StringBuilder();

        final String headers = String.join(",", headersList);

        sb.append(headers + "\n");

        if (!records.isEmpty()) {
            for (final CSVRecord r : records) {
                for (int i = 0; i < r.size(); i++) {
                    sb.append(r.get(i));
                    // add comma if not last element
                    if (i < (r.size() - 1)) {
                        sb.append(",");
                    }
                }
                sb.append("\n");
            }
        }
        return sb;
    }

Then write string to HDFS file.

DataOutputStream outputStream = getHdfsOutputStream(destPath);
outputStream.writeBytes(records.toString());

This works fine with a small list.

But with a large list (say > 100000) the JVM has crashed with an OOME.

Administratively Yielded for 1 sec: java.lang.OutOfMemoryError: Java heap space
java.lang.OutOfMemoryError: Java heap space
        at java.util.Arrays.copyOfRange(Arrays.java:3664)
        at java.lang.String.<init>(String.java:207)
        at java.lang.StringBuilder.toString(StringBuilder.java:407)

Is it possible to bypass creation of the string builder and stream the contents of the list to HDFS?

This could involve 2 steps.

1. create a stream of formatted record strings
2. write stream (line-by-line?) to HDFS

Pseudo code

outputStream.write(records.stream().map(<format to add commas and new line>))

But I'm not sure what is best for step 1.

This returns a list of strings so it's not streaming anything.

recs.stream().map(v -> v.get(0) + "," + v.get(1)).collect(toList());

This returns a stream of objects.

records.stream().map(v -> v.get(0) + "," + v.get(1) + "\n"));

How to feed this stream into HDFS?

Any help/tips would be greatly appreciated.

**** EDIT: ****

This approach seems to works for step 2 (write stream) but without the formatting of record.

final FSDataOutputStream outputStream = getHdfsOutputStream(destPath);
BufferedWriter bufferedWriter = new BufferedWriter(new OutputStreamWriter(outputStream, StandardCharsets.UTF_8));

for(CSVRecord r: records) {
    bufferedWriter.write(r.toString());
    bufferedWriter.newLine();
}
bufferedWriter.close();
configOutputStream.close();

Answer 1

Here the final code. Creates a string builder p/record and writes one string at a time.

    private void persistRecords(final List<CSVRecord> records, final String targetDir, final String targetFile,
                                final List<String> headers) throws IOException {
        final String targetPath = targetDir + targetFile;

        try (final FSDataOutputStream outputStream = getHdfsOutputStream(targetPath);
             final BufferedWriter bufferedWriter = new BufferedWriter(new OutputStreamWriter(outputStream, StandardCharsets.UTF_8))) {

            bufferedWriter.write(String.join(",", headers));
            bufferedWriter.newLine();

            for (final CSVRecord r : records) {
                final StringBuilder sb = format(r);
                bufferedWriter.write(sb.toString());
                bufferedWriter.newLine();
            }

        }
    }

    private StringBuilder format(final CSVRecord r) {
        final StringBuilder sb = new StringBuilder();
        for (int i = 0; i < r.size(); i++) {
            sb.append(r.get(i));
            // add comma if not last element
            if (i < (r.size() - 1)) {
                sb.append(",");
            }
        }
        return sb;
    }