Operation from float to BigDecimal in Java

Question

I'm doing a math operation beetwen hours, but I've a problem with the float (the java virtual machine approximate is not perfect). So, I decided to convert this operation in BigDecimal...but I've some problems with results..

This is the original code:

public float ConvertTo100(float input)
    {
float output = 0.0f;
        int hh;
        float mm;
        hh = (int)input;
        mm = input - hh;

        output = hh + ((input - hh)/60) * 100;

        return output;
}

This is my convertion, but doesn't works:

    public float ConvertTo100(float input)
        {
            BigDecimal inputBD = new BigDecimal(Float.toString(input));

            String inpString = String.valueOf(input);
            String[] inpsplit = inpString.split("\\.");

            BigDecimal hh = new BigDecimal(Float.toString(Integer.parseInt(inpsplit[0])));

            BigDecimal output = hh.add((inputBD.subtract(hh).divide(BigDecimal.valueOf(60))).multiply(BigDecimal.valueOf(1000)));

            return Float.parseFloat(String.valueOf(output));
}

Where is the problem?? Thank you so much!!! :)

Answer 1

Division by 60 cannot be done exactly in decimal for many inputs. To continue using BigDecimal, you will need to pick a scale and rounding mode for the division. Here's one way of doing it, but read the BigDecimal documentation to see all the options.

BigDecimal output = hh.add((inputBD.subtract(hh).divide(
    BigDecimal.valueOf(60), 10, BigDecimal.ROUND_HALF_EVEN))
    .multiply(BigDecimal.valueOf(1000)));

However, the main point of using BigDecimal is to get exact representation of decimal fractions. You are not getting that benefit, but are getting the messy code that results from doing arithmetic using method calls instead of infix operators. Unless you need a specific rounding mode, or some finite precision greater than 16.9 significant digits, you would be much better off using double.