28.14 Handling NaN

NaNs are not numbers: they represent values from computations that produce undefined results. They have a distinctive property that makes them unlike any other floating-point value: they are unequal to everything, including themselves! Thus, you can write a test for a NaN like this:

if (x != x)
  printf ("x is a NaN\n");

This test works in GNU C, but some compilers might evaluate that test expression as false without properly checking for the NaN value. A more portable way to test for NaN is to use the isnan function declared in math.h:

if (isnan (x))
  printf ("x is a NaN\n");

See Floating Point Classes in The GNU C Library Reference Manual.

One important use of NaNs is marking of missing data. For example, in statistics, such data must be omitted from computations. Use of any particular finite value for missing data would eventually collide with real data, whereas such data could never be a NaN, so it is an ideal marker. Functions that deal with collections of data that may have holes can be written to test for, and ignore, NaN values.

It is easy to generate a NaN in computations: evaluating 0.0 / 0.0 is the commonest way, but Infinity - Infinity, Infinity / Infinity, and sqrt (-1.0) also work. Functions that receive out-of-bounds arguments can choose to return a stored NaN value, such as with the NAN macro defined in math.h, but that does not set the invalid operand exception flag, and that can fool some programs.

Like Infinity, NaNs propagate in computations, but they are even stickier, because they never disappear in division. Thus, once a NaN appears in a chain of numerical operations, it is almost certain to pop out into the final results. The programmer has to decide whether that is expected, or whether there is a coding or algorithmic error that needs repair.

In general, when function gets a NaN argument, it usually returns a NaN. However, there are some exceptions in the math-library functions that you need to be aware of, because they violate the NaNs-always-propagate rule:

• pow (x, 0.0) always returns 1.0, even if x is 0.0, Infinity, or a NaN.
• pow (1, y) always returns 1, even if y is a NaN.
• hypot (INFINITY, y) and hypot (-INFINITY, y) both always return INFINITY, even if y is a Nan.
• If just one of the arguments to fmax (x, y) or fmin (x, y) is a NaN, it returns the other argument. If both arguments are NaNs, it returns a NaN, but there is no requirement about where it comes from: it could be x, or y, or some other quiet NaN.

NaNs are also used for the return values of math-library functions where the result is not representable in real arithmetic, or is mathematically undefined or uncertain, such as sqrt (-1.0) and sin (Infinity). However, note that a result that is merely too big to represent should always produce an Infinity, such as with exp (1000.0) (too big) and exp (Infinity) (truly infinite).