Using the RE.match(CharacterIterator,int) function
with a "CharacterArrayCharacterIterator", then calling "getParen(int)"
often returns a string of the incorrect length, or throws an exception.

This is due to the implementation of "substring(int,int)" in the
CharacterArrayCharacterIterator class and/or the mis-documentation of
the CharacterIterator.substring interface.

The confusion is in whether the second argument to substring represents
the endIndex or the length. The API docs say it's the length, but the
RE implementation, and the StringCharacterIterator implementation both
treat it as the endIndex.
[Note, the standard java string has,
java.lang.String.substring(int beginIndex, int endIndex)
but the constructor is java.lang.String(char[] src, int off, int len)]

Secondly, there is no check that the requested substring stays within the
bounds of the sequence length specified at construction time.
An IndexOutOfBoundsException should be thrown in that case.

I think the best solution is to first update the API docs to specify
that it is infact (beginIndex, endIndex), and then to update the 
CharacterArrayCharacterIterator.substring functions to be something like this:

 public String substring(int beginIndex, int endIndex)
 {
   if (endIndex > len)
     throw new IndexOutOfBoundsException("endIndex=" + endIndex +
	"; sequence size=" + len);
   if (beginIndex < 0)
     throw new IndexOutOfBoundsException("beginIndex=" + beginIndex);
   return new String(src, off + beginIndex, endIndex - beginIndex);
 }

 public String substring(int beginIndex)
 {
   if (beginIndex > len)
     throw new IndexOutOfBoundsException("index=" + beginIndex +
	"; sequence size=" + len);
   return new String(src, off + beginIndex, len - beginIndex);
 }