Fix CharArrayBuffer.subSequence(beganIndex,endIndex) to return right result.

[ufolr]

CharArrayBuffer implemented the CharSequence interface,
and CharSequence define method:
public CharSequence subSequence(final int beginIndex, final int endIndex)

Threre are two problem in original implementation of CharArrayBuffer.subSequence

1. It call CharBuffer.wrap(buffer.array, beginIndex, endIndex) directly，but the 3rd parameter of the CharBuffer.wrap method is length, it means how many chars need to copy, not endIndex. It makes some problem such as pass CharArrayBuffer as a CharSequence in to Regex's matcher method like following:

        charSequence.append("aabb123c");   
        Pattern p = Pattern.compile("\\D*(\\d+)\\D*");   
        Matcher m = p.matcher(charSequence);   
        Assert.assertEquals("123", m.find()?m.group(1):""); 

2. CharArrayBuffer's sub-sequence should return the result as same class of CharArrayBuffer. But' the old implementation use CharBuffer.wrap to make a result which instance of HeapCharBuffer.

This problem exists in both 4.x and 5.x. I'd like see this problem to be fixed.

[ok2c]

CharArrayBuffer's sub-sequence should return the result as same class of CharArrayBuffer. But' the old implementation use CharBuffer.wrap to make a result which instance of HeapCharBuffer.

@ufolr Where is this requirement stated anywhere? I could not find any evidence of such requirement in CharSequence javadocs.

[ufolr]

CharArrayBuffer's sub-sequence should return the result as same class of CharArrayBuffer. But' the old implementation use CharBuffer.wrap to make a result which instance of HeapCharBuffer.

@ufolr Where is this requirement stated anywhere? I could not find any evidence of such requirement in CharSequence javadocs.

@ok2c There is no specification. Just my suggestion. I think its return should be intuitive. Such as sub noodle is a noodle lol.
There maybe some method which accept a CharArrayBuffer as parameter, but can't take the sub-sequence result of CharArrayBuffer.
This will cause me confusion.
Anyway, just my suggestion.

[ok2c]

@ufolr While "sub noodle is a noodle" is conceptually nice, it does cause an unnecessary array allocation and an array copy, which is a very expensive operation. It completely defeats the purpose of having efficient access to a sub sequence of characters contained in the buffer.
Could we fix the bug without causing unnecessary buffer allocation?

[ufolr]

@ok2c Yep. I saw that waste of performance. So I make CharArrayBuffer can reuse the existing char array by adding an offset pos tag. If there is no write option, the CharArrayBuffer will not do allocation or array copy. Just like a HeapCharBuffer do.
Too do this, I add an constructor CharArrayBuffer(final char[] b, final int off, final int len) to CharArrayBuffer.

[ok2c]

@ufolr I am sorry but this is too much so that "sub noodle could be a noodle". Let's just fix the bug and let the rest stay the same.

[ufolr]

@ufolr I am sorry but this is too much so that "sub noodle could be a noodle". Let's just fix the bug and let the rest stay the same.

@ok2c OK, I revert the changes about "sub noodle", just fix the wrong result value problem.
"sub noodle could be a noodle" is just a suggestion, and does not have to be modified.

[ok2c]

@ufolr Cool! Many thanks for contributing the fix!