Quick and dirty (very dirty) big decimal writable. This is much uglier than it needs to be because Hadoop's API causes an extra copy to be required.

BigDecimalWritable

  public static class BigDecimalWritable implements Writable {
    private BigDecimal value;

    public BigDecimalWritable(BigDecimal value) {
      this.value = value;
    }
    
    public BigDecimal value() {
      return value;
    }

    @Override
    public void write(DataOutput dataOutput) throws IOException {
      ByteArrayOutputStream bos = new ByteArrayOutputStream();
      ObjectOutputStream out = new ObjectOutputStream(bos);
      out.writeObject(value);
      out.close();
      byte[] bytes = bos.toByteArray();
      dataOutput.write(bytes.length);
      dataOutput.write(bytes);
    }

    @Override
    public void readFields(DataInput dataInput) throws IOException {
      int n = dataInput.readInt();
      if (n < 0 || n > 1000) {
        throw new IllegalArgumentException("Invalid representation for BigDecimal ... length is " + n);
      }
      byte[] bytes = new byte[n];
      dataInput.readFully(bytes);
      try {
        value = (BigDecimal) new ObjectInputStream(new ByteArrayInputStream(bytes)).readObject();
      } catch (ClassNotFoundException e) {
        throw new RuntimeException("Unable to read serialized BigDecimal value, can't happen", e);
      }
    }
  }