Most efficient way to convert InputStream into byte[]? [duplicate]

Question

ChannelBufferInputStream responseStream = (ChannelBufferInputStream) response.getBodyAsStream();
ArrayList<Byte> arrayList = new ArrayList<Byte>();
try {
    while (responseStream.available() > 0) {
        arrayList.add(responseStream.readByte());
    }
} catch (IOException e) {
    e.printStackTrace();
    return internalServerError();
}
Iterator<Byte> iterator = arrayList.iterator();
byte[] bytes = new byte[arrayList.size()];
int i = 0;
while (iterator.hasNext()) {
    bytes[i++] = iterator.next();
}

This code is called on every page load of my web app. It seems to be running pretty fast, but is there anything that could make this run faster?

Edit - Updated using byte array output stream

ChannelBufferInputStream responseStream = (ChannelBufferInputStream) response.getBodyAsStream();
ByteArrayOutputStream byteArrayOutputStream = new ByteArrayOutputStream();
try {
    int read = responseStream.read();
    while (read != -1) {
        byteArrayOutputStream.write(read);
        read = responseStream.read();
    }
} catch (IOException e) {
    e.printStackTrace();
    return internalServerError();
}
byte[] bytes = byteArrayOutputStream.toByteArray();
return ok(bytes).as(response.getHeader("Content-type"));

Edit - Benchmark test code

ChannelBufferInputStream responseStream = (ChannelBufferInputStream) response.getBodyAsStream();
long t1 = System.nanoTime();

ByteArrayOutputStream byteArrayOutputStream = new ByteArrayOutputStream();
try {
    int read = responseStream.read();
    while (read != -1) {
        byteArrayOutputStream.write(read);
        read = responseStream.read();
    }
} catch (IOException e) {
    e.printStackTrace();
    return internalServerError();
}
byte[] bytes = byteArrayOutputStream.toByteArray();

long t2 = System.nanoTime();
System.out.println(t2-t1);
return ok(bytes).as(response.getHeader("Content-type"));

Average Time after 100+ request - 46873

ChannelBufferInputStream responseStream = (ChannelBufferInputStream) response.getBodyAsStream();
long t1 = System.nanoTime();

ArrayList<Byte> arrayList = new ArrayList<Byte>();
try {
    while (responseStream.available() > 0) {
        arrayList.add(responseStream.readByte());
    }
} catch (IOException e) {
    e.printStackTrace();
    return internalServerError();
}
Iterator<Byte> iterator = arrayList.iterator();
byte[] bytes = new byte[arrayList.size()];
int i = 0;
while (iterator.hasNext()) {
    bytes[i++] = iterator.next();
}

long t2 = System.nanoTime();
System.out.println(t2-t1);
return ok(bytes).as(response.getHeader("Content-type"));

Average Time after 100+ request - 522848

long t1 = System.nanoTime();
byte[] bytes;
try {
    bytes = org.apache.commons.io.IOUtils.toByteArray(responseStream);
} catch (Exception e) {
    return internalServerError();
}

long t2 = System.nanoTime();
System.out.println(t2-t1);

Average Time after 100+ request - 45088

long t1 = System.nanoTime();
byte[] bytes;
try {
    bytes = sun.misc.IOUtils.readFully(responseStream, -1, true);
} catch (Exception e) {
    return internalServerError();
}

long t2 = System.nanoTime();
System.out.println(t2 - t1);

Average Time after 100+ request - 20180

Answer 1

Yes. Use a ByteArrayOutputStream rather than an ArrayList. Then read chunks of bytes from the InputStream (without using available(), which should almost always never used) and write these chunks to the ByteArrayOutputStream, until the read() method returns -1. Then call toByteArray() on your ByteArrayOutputStream.

You could use Guava's ByteStreams.toByteArray() method, which does all that for you, or you could read its source code to have a better idea of how it does it. Reading the IO tutorial might also help.