leadzero - Flexible F0 format
Simulate the F0 format with more options, for compilers that implement NaN, Inf. (f0withIEEE.f90) and for compilers that do not implement these non-normal numbers (f0noIEEE.f90). Both versions define a Fortran 95/2003 module leadzero and have a corresponding test program in this file. The code was developed by J F Harper, Mathematics, Victoria University, Wellington, NZ 20 Jan 2012.
Compilers vary in their own treatments of leading zero and optional + sign because Fortran standards f90,f95,f2003 differ, and f95 was ambiguous. The module lets users choose their own treatment. The test program assumes that NAN and INF are valid inputs to a READ statement for a real variable, as f2003 requires for processors that support IEEE 754 features. If your processor does not support them, use f0noIEEE.f90 instead of this program. The modules in f0noIEEE.f90 and fotest.f90 are identical but their test programs differ. IEEE intrinsic modules are not used.
The module contains one public function and four public parameters to aid in their usage:
Function f0(x,n,s) or f0(x,n) (the s argument is optional) returns the shortest possible string of characters containing x in F0.n format, with leading zero and + sign controlled by the character string s if it is present, or by SS,F0.n format as in f2003 if not, so if the processor supports separate positive and negative zeros then f0(-0.0,n) shall begin with a - sign.
It returns a string of the minimum required length, which is at most rspace+n.
Value to be written by f0
Number of digits to be written after the decimal point
Specifying treatment of + sign and leading zero
If s is absent or contains neither LZ nor SP the effect is that of format SS,F0.n as in the f2003 standard. If s contains SP the effect is that of SP,F0.n, i.e. a + sign if x>0 or x is +0.0, a - sign if x<0 or -0.0. If the processor does not support separate -0.0 and +0.0 then 0.0 is treated as +0.0. A zero before the decimal point is inserted if the output would otherwise have had no digits at all, or if s contains LZ and there would otherwise be no digits before the decimal point. If x is infinite or not a number, f0 returns the appropriate one of +Inf,Inf,-Inf,NaN.
Kind for single-precision reals - p1 = kind(1.0).
Kind for double-precision reals - p2 = kind(1d0).
The integer parameter ispace is the space needed for -huge(1) in I0 format
The integer parameter rspace is the space needed for -huge(1d0) in F0.0 format,
Copyright © 2012 John Harper