Wrapping a ByteBuffer with an InputStream

There seem to be some bugs with the implementation referred to by Thilo, and also copy and pasted on other sites verbatim:

1. ByteBufferBackedInputStream.read() returns a sign extended int representation of the byte it reads, which is wrong (value should be in range [-1..255])
2. ByteBufferBackedInputStream.read(byte[], int, int) does not return -1 when there are no bytes remaining in the buffer, as per the API spec

ByteBufferBackedOutputStream seems relatively sound.

I present a 'fixed' version below. If I find more bugs (or someone points them out) I'll update it here.

Updated: removed synchronized keywords from read/write methods

InputStream

public class ByteBufferBackedInputStream extends InputStream {

    ByteBuffer buf;

    public ByteBufferBackedInputStream(ByteBuffer buf) {
        this.buf = buf;
    }

    public int read() throws IOException {
        if (!buf.hasRemaining()) {
            return -1;
        }
        return buf.get() & 0xFF;
    }

    public int read(byte[] bytes, int off, int len)
            throws IOException {
        if (!buf.hasRemaining()) {
            return -1;
        }

        len = Math.min(len, buf.remaining());
        buf.get(bytes, off, len);
        return len;
    }
}

OutputStream

public class ByteBufferBackedOutputStream extends OutputStream {
    ByteBuffer buf;

    public ByteBufferBackedOutputStream(ByteBuffer buf) {
        this.buf = buf;
    }

    public void write(int b) throws IOException {
        buf.put((byte) b);
    }

    public void write(byte[] bytes, int off, int len)
            throws IOException {
        buf.put(bytes, off, len);
    }

}

If it's backed by a byte array, you can use a ByteArrayInputStream and get the byte array via ByteBuffer.array(). This will throw an exception if you're trying it on a native ByteBuffer.

Nothing in the JDK, but there are lots of implementations out there, google for ByteBufferInputStream. Basically they wrap one or more ByteBuffers and keep track of an index into them that records how much has already been read. Something like this comes up a lot, but apparently is buggy, see @Mike Houston's answer for an improved version).