char[] to String sequence mismatching in Java for Unicode characters

Question

I have a method like below (please ignore the code optimization issue.) This method replaces the Unicode character (Bengali characters)

static String swap(String temp, char c) 
{
    Integer length=temp.length();
    char[] charArray = temp.toCharArray();
      for(int u=0;u<length;u++)
        {           
            if(charArray[u]==c)
            {
                    char g=charArray[u];
                    charArray[u]=charArray[u-1];
                    charArray[u-1]=g;
            }                   
        }
    String string2 = new String(charArray);
    return string2;
}

while debugging, i got the values of charArray like the below image:
please note that the characters are in a sequenced format what I want.
But after the execution of the statement, the value stored in String variable is mismatched. like below:
I want to display the string as "রেরেরে" but it is displaying "েরেরের" what i not want. Please tell me what I am doing wrong.

Answer 1

Note - I don't know Bengali, but I know a bit (or a lot, depending on whom you ask) about Unicode and how Java supports it. The answer assumes knowledge of the latter and not the former.

Going by the Unicode 6.0 Bengali chart, রে is a combination of the dependent vowel sign ে (0x09C7) and the consonant র (0x09B0) and is represented as a sequence of two characters in the character array.

If you are getting the dependent vowel sign alone, in the resulting character sequence (and hence the string), then your optimization is likely to be kooky, as it appears to assume that Bengali characters in Unicode can be represented as a single Unicode codepoint or a single char variable in Java; this would result in the scenario where a consonant would be replaced by another consonant, but the dependent vowel preceding the consonant would never be replaced.

I think a correct optimization must therefore consider the presence of dependent vowels, and compare the following consonant in addition to the vowel , i.e. it must compare two characters in the character array, instead of comparing individual characters. This might also imply that your method signature must be changed to allow for a char[] to be passed, instead of a single char, so that Bengali characters can be replaced with the intended Bengali character, instead of replacing a Unicode codepoint with another, which is what is being done currently.

The notes in other answers on the ArrayIndexOutofBoundsException is valid. The following example that uses your character replacement algorithm demonstrates that not only is your algorithm incorrect, but it is quite possible for the exception to be thrown:

class CodepointReplacer
{

    public static void main(String[] args)
    {
        String str1 = "রেরেরে";
            /* 
             * The following is a linguistically invalid sequence,
             * but Java does not concern itself with linguistical correctness
             * if the String or char sequence has been constructed incorrectly.
             */
        String str2 = "েরেরের"; 
            /*
             * replacement character র for our strings
             * It is not রে as one would anticipate.
             */ 
        char c = str1.charAt(1);

        optimizeKookily(str1, c);
        optimizeKookily(str2, c);
    }

    private static void optimizeKookily(String temp, char c)
    {
        Integer length = temp.length();
        char[] charArray = temp.toCharArray();
        for (int u = 0; u < length; u++)
        {
            if (charArray[u] == c)
            {
                char g = charArray[u];
                charArray[u] = charArray[u - 1]; //throws exception on second invocation of this method.
                charArray[u - 1] = g;
            }
        }
    }
}

A better character replacement strategy would therefore be to use the String.replace (the CharSequence variant) or String.replaceAll functions, assuming that you would know how to use these with Bengali characters.