LongCalcNetCore 0.1.9

This package has no dependencies.

Reversed the changes made in version 0.1.7. Instead, made an effort to ensure that the output from the various public bf functions was accurate to Precision places, counting from first non-zero digit (if output is non-zero). This resulted in changes to the following bf functions: Exp, Cos, ArcTan, ArcCos, ArcSin, Log.

If someone desires to output a bf object x.toString(... Precision, ...) in such a way as that in version 0.1.7, it only requires this code snippet:

string output = "0.";
if (!x.IsZero) { if (x.Abs() lt; new bf(1)) { Precision -= (long)Math.Abs(Log10(x.Abs())); if (Precision gt; 0) output = x.toString(... Precision, ...); } else output = x.toString(... Precision, ...); }

(where lt; is the less than character and gt; is the greater than character)

I should have mentioned in the release notes of the previous version that the internal limitation on x.Exp is that |x| be less than about 1.e+90180. For a value of x about this large, it requires probably several hours to compute x.Exp to 1,000 digits on a rather recent i5 machine with 20Gb memory.

Checks for arithmetic overflow or underflow are again compiler-included in this release.