Name

---

      WHERENAN()

Purpose

---

      Find the indices of all big-endian NaN values in an array. OBSOLETE

Explanation

---

      Find the positions of all values within an array that correspond to the
      big-endian NaN (not-a-number) special values.
      THIS PROCEDURE ONLY IDENTIFIES BIG_ENDIAN NaN VALUES. DO NOT USE IT
      TO IDENTIFY NaN VALUES IN GENERAL. Instead, to identify NaN values on
      the host machine use the FINITE() function
      IDL> result = where( finite(array,/NAN) )
     
      The main purpose of this routine is to catch NaN special values
      written in big_endian format (e.g. FITS data) on a little endian
      machine prior to conversion with e.g. IEEE_TO_HOST. It was needed
      many years ago because VMS machines could not handle big-endian
      special values, but this routine is now kept only for backwards
      compatibility.

Calling Sequence

---

      Result = WHERENAN( ARRAY [, COUNT ] )

Input Parameters

---

      ARRAY = Array to test against the IEEE NaN special values. Must be
                of either floating point, double-precision, or complex type.

Outputs

---

      The result of the function is the indices of all values of ARRAY
      corresponding to the IEEE NaN specification, similar to the IDL WHERE
      function.

Optional Output Parameters

---

      COUNT = Number of values found corresponding to IEEE NaN.

Side Effects

---

      If no NaN values are found, or if ARRAY is not of type float, double
      precision, or complex, then -1 is returned, and COUNT is set to 0.

Restrictions

---

      ARRAY must be of type float, double-precision, or complex.

Procedure

---

      The bit patterns of the numbers being tested are compared against the
      IEEE NaN standard.

Modification History

---

      William Thompson, Feb. 1992.
      William Thompson, Oct. 1992, fixed bug regarding order of bytes on VAX
              machines.
      Converted to IDL V5.0 W. Landsman September 1997