How to deal with a huge, one-line file in Java

Question

I need to read a huge file (15+GB) and perform some minor modifications (add some newlines so a different parser can actually work with it). You might think that there are already answers for doing this normally:

• Reading a very huge file in java
• How to read a large text file line by line using Java?

but my entire file is on one line.

My general approach so far is very basic:

char[] buffer = new char[X];
BufferedReader reader = new BufferedReader(new ReaderUTF8(new FileInputStream(new File("myFileName"))), X);
char[] bufferOut = new char[X+a little];
int bytesRead = -1;
int i = 0;
int offset = 0;
long totalBytesRead = 0;
int countToPrint = 0;
while((bytesRead = reader.read(buffer)) >= 0){
    for(i = 0; i < bytesRead; i++){
        if(buffer[i] == '}'){
            bufferOut[i+offset] = '}';
            offset++;
            bufferOut[i+offset] = '\n';
        }
        else{
            bufferOut[i+offset] = buffer[i];
        }
    }
    writer.write(bufferOut, 0, bytesRead+offset);
    offset = 0;
    totalBytesRead += bytesRead;
    countToPrint += 1;
    if(countToPrint == 10){
        countToPrint = 0;
        System.out.println("Read "+((double)totalBytesRead / originalFileSize * 100)+" percent.");
    }
}
writer.flush();

After some experimentation, I've found that a value of X larger than a million gives optimal speed - it looks like I'm getting about 2% every 10 minutes, while a value of X of ~60,000 only got 60% in 15 hours. Profiling reveals that I'm spending 96+% of my time in the read() method, so that's definitely my bottleneck. As of writing this, my 8 million X version has finished 32% of the file after 2 hours and 40 minutes, in case you want to know how it performs long-term.

Is there a better approach for dealing with such a large, one-line file? As in, is there a faster way of reading this type of file that gives me a relatively easy way of inserting the newline characters?

I am aware that different languages or programs could probably handle this gracefully, but I'm limiting this to a Java perspective.

Answer 1

You are making this far more complicated than it should be. By just making use of the buffering already provided by the standard classes you should get a thorughput of at least several MB per second without any hassles.

This simple test program processes 1GB in less than 2 minutes on my PC (including creating the test file):

import java.io.BufferedInputStream;
import java.io.BufferedOutputStream;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.InputStream;
import java.io.InputStreamReader;
import java.io.OutputStream;
import java.io.OutputStreamWriter;
import java.io.Reader;
import java.io.Writer;
import java.nio.charset.Charset;
import java.nio.charset.StandardCharsets;
import java.util.Random;

public class TestFileProcessing {

    public static void main(String[] argv) {

        try {
            long time = System.currentTimeMillis();
            File from = new File("C:\\Test\\Input.txt");
            createTestFile(from, StandardCharsets.UTF_8, 1_000_000_000);
            System.out.println("Created file in: " + (System.currentTimeMillis() - time) + "ms");

            time = System.currentTimeMillis();
            File to = new File("C:\\Test\\Output.txt");
            doIt(from, to, StandardCharsets.UTF_8);
            System.out.println("Converted file in: " + (System.currentTimeMillis() - time) + "ms");
        } catch (IOException e) {
            throw new RuntimeException(e.getMessage(), e);
        }
    }

    public static void createTestFile(File file, Charset encoding, long size) throws IOException {
        Random r = new Random(12345);
        try (OutputStream fout = new FileOutputStream(file);
                BufferedOutputStream bout = new BufferedOutputStream(fout);
                Writer writer = new OutputStreamWriter(bout, encoding)) {
            for (long i=0; i<size; ++i) {
                int c = r.nextInt(26);
                if (c == 0)
                    writer.write('}');
                else
                    writer.write('a' + c);
            }
        }
    }

    public static void doIt(File from, File to, Charset encoding) throws IOException {
        try (InputStream fin = new FileInputStream(from);
                BufferedInputStream bin = new BufferedInputStream(fin);
                Reader reader = new InputStreamReader(bin, encoding);
                OutputStream fout = new FileOutputStream(to);
                BufferedOutputStream bout = new BufferedOutputStream(fout);
                Writer writer = new OutputStreamWriter(bout, encoding)) {
            int c;
            while ((c = reader.read()) >= 0) {
                if (c == '}')
                    writer.write('\n');
                writer.write(c);
            }
        }
    }

}

As you see no elaborate logic or excessive buffer sizes are used. What is used is simply buffering the streams closest to the hardware, the FileInput/OutputStream.