Converting char[] to byte[]

Arun Abraham picture Arun Abraham · Apr 1, 2011 · Viewed 136.3k times · Source

I would like to convert a character array to a byte array in Java. What methods exists for making this conversion?

Answer

Andrii Nemchenko picture Andrii Nemchenko · Mar 12, 2012

Convert without creating String object:

import java.nio.CharBuffer;
import java.nio.ByteBuffer;
import java.util.Arrays;

byte[] toBytes(char[] chars) {
  CharBuffer charBuffer = CharBuffer.wrap(chars);
  ByteBuffer byteBuffer = Charset.forName("UTF-8").encode(charBuffer);
  byte[] bytes = Arrays.copyOfRange(byteBuffer.array(),
            byteBuffer.position(), byteBuffer.limit());
  Arrays.fill(byteBuffer.array(), (byte) 0); // clear sensitive data
  return bytes;
}

Usage:

char[] chars = {'0', '1', '2', '3', '4', '5', '6', '7', '8', '9'};
byte[] bytes = toBytes(chars);
/* do something with chars/bytes */
Arrays.fill(chars, '\u0000'); // clear sensitive data
Arrays.fill(bytes, (byte) 0); // clear sensitive data

Solution is inspired from Swing recommendation to store passwords in char[]. (See Why is char[] preferred over String for passwords?)

Remember not to write sensitive data to logs and ensure that JVM won't hold any references to it.


The code above is correct but not effective. If you don't need performance but want security you can use it. If security also not a goal then do simply String.getBytes. Code above is not effective if you look down of implementation of encode in JDK. Besides you need to copy arrays and create buffers. Another way to convert is inline all code behind encode (example for UTF-8):

val xs: Array[Char] = "A ß € 嗨 𝄞 🙂".toArray
val len = xs.length
val ys: Array[Byte] = new Array(3 * len) // worst case
var i = 0; var j = 0 // i for chars; j for bytes
while (i < len) { // fill ys with bytes
  val c = xs(i)
  if (c < 0x80) {
    ys(j) = c.toByte
    i = i + 1
    j = j + 1
  } else if (c < 0x800) {
    ys(j) = (0xc0 | (c >> 6)).toByte
    ys(j + 1) = (0x80 | (c & 0x3f)).toByte
    i = i + 1
    j = j + 2
  } else if (Character.isHighSurrogate(c)) {
    if (len - i < 2) throw new Exception("overflow")
    val d = xs(i + 1)
    val uc: Int = 
      if (Character.isLowSurrogate(d)) {
        Character.toCodePoint(c, d)
      } else {
        throw new Exception("malformed")
      }
    ys(j) = (0xf0 | ((uc >> 18))).toByte
    ys(j + 1) = (0x80 | ((uc >> 12) & 0x3f)).toByte
    ys(j + 2) = (0x80 | ((uc >>  6) & 0x3f)).toByte
    ys(j + 3) = (0x80 | (uc & 0x3f)).toByte
    i = i + 2 // 2 chars
    j = j + 4
  } else if (Character.isLowSurrogate(c)) {
    throw new Exception("malformed")
  } else {
    ys(j) = (0xe0 | (c >> 12)).toByte
    ys(j + 1) = (0x80 | ((c >> 6) & 0x3f)).toByte
    ys(j + 2) = (0x80 | (c & 0x3f)).toByte
    i = i + 1
    j = j + 3
  }
}
// check
println(new String(ys, 0, j, "UTF-8"))

Excuse me for using Scala language. If you have problems with converting this code to Java I can rewrite it. What about performance always check on real data (with JMH for example). This code looks very similar to what you can see in JDK[2] and Protobuf[3].