When outputting float values, the external representation uses a format
similar to float literals, except that:
- Values are always written with the character E and the signs
for both mantissa and exponent.
- Values are either 14 or 23 characters long (for 4F and 8F respectively).
- Values are normalized. That is, the decimal point immediately follows
the most significant digit.
- The decimal separator character is either period or comma depending on
the parameter for Control Specification keyword DECEDIT.
Here are some examples of how float values are presented:
+1.2345678E-23
-8.2745739E+03
-5.722748027467392E-123
+1,2857638E+14 if DECEDIT(',') is specified