please answer step by step

Consider the expression 1/1 - x - 1/1 + x, assuming x notequalto plusminus 1. (a) For what range of values of x is it difficult to compute this expression accurately in floating-point arithmetic? (b) Give a rearrangement of the terms such that, for the range of x in part a, the computation is more accurate in floating-point arithmetic