Humanizes numbers. Makes them human-readable

DoubleExtensions.cs

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using static System.Math;

namespace Foo {

    public static class DoubleExtensions {
        
        public static double RoundToSignificantFigures(this double value, int numSignificantFigures) {
            if (value == 0) return 0.0;
            var scale = Pow(10, Floor(Log10(Abs(value))) + 1);
            // Perform the last step using decimals to prevent double-arithmetic re-introducing tiny errors (and more figures to the result)
            return (double)((decimal)scale * (decimal)Math.Round(value / scale, numSignificantFigures));
        }
    }
}

HumanReadableDoubles.cs

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using static System.Math;

namespace Foo {

    public static class HumanReadableDoubles {
        
        // Created with thanks to http://stackoverflow.com/questions/16083666/make-big-and-small-numbers-human-readable/16091580#16091580
        static readonly string[] humanReadableSuffixes = { "f", "a", "p", "n", "μ", "m", "", "k", "M", "G", "T", "P", "E" };
        public static string ToHumanReadable(this double value, int numSignificantDigits) {

            // Deal with special values
            if (double.IsInfinity(value) || double.IsNaN(value) || value == 0 || numSignificantDigits <= 0)
                return value.ToString();

            // We deal only with positive values in the code below
            var isNegative = Sign(value) < 0;
            value = Abs(value);

            // Calculate the exponent as a multiple of 3, ie -6, -3, 0, 3, 6, etc
            var exponent = (int)Floor(Log10(value) / 3) * 3;

            // Find the correct suffix for the exponent, or fall back to scientific notation
            var indexOfSuffix = exponent / 3 + 6;
            var suffix = indexOfSuffix >= 0 && indexOfSuffix < humanReadableSuffixes.Length
                ? humanReadableSuffixes[indexOfSuffix]
                : "·10^" + exponent;

            // Scale the value to the exponent, then format it to the correct number of significant digits and add the suffix
            value = value * Pow(10, -exponent);
            var numIntegerDigits = (int)Floor(Log(value, 10)) + 1;
            var numFractionalDigits = Min(numSignificantDigits - numIntegerDigits, 15);
            var format = $"{new string('0', numIntegerDigits)}.{new string('0', numFractionalDigits)}";
            var result = value.ToString(format) + suffix;

            // Handle negatives
            if (isNegative)
                result = "-" + result;

            return result;
        }

        public static double ParseHumanReadableDouble(this string expression) {
            var multiplier = 1.0;
            if (expression.Contains("·10^")) {
                var indexOfCaret = expression.LastIndexOf('^');
                multiplier = Pow(10, int.Parse(expression.Substring(indexOfCaret + 1)));
                expression = expression.Substring(0, indexOfCaret - 3);
            } else {
                var suffix = humanReadableSuffixes.SingleOrDefault(s => s.Length > 0 && expression.EndsWith(s, StringComparison.InvariantCulture)) ?? "";
                var suffixIndex = humanReadableSuffixes.IndexOf(suffix);
                multiplier = Pow(10, 3 * (suffixIndex - 6));
                expression = expression.Replace(suffix, string.Empty);
            }
            return double.Parse(expression) * multiplier;
        }
    }
}

I just saw it. It is an interesting approach. Do you have any references that describe in more detail this trick? I can see if the number is decomposed into this form X 10^n that we can use decimal for X which has more bits than the mantissa of double.

Sorry to dump more code at you, but how about these numeric manipulation functions:

    /// <summary>
    /// Return 1 only when values in between min and max, 0 otherwise.
    /// </summary>
    /// <param name="value">The value to evaluate</param>
    /// <param name="min_value">The min value</param>
    /// <param name="max_value">The max value</param>
    /// <returns></returns>
    [Pure]
    public static double Chi(this double value, double min_value, double max_value)
    {
        double dx = Math.Abs(max_value-min_value);
        min_value=Math.Min(min_value, max_value);
        max_value=min_value+dx;
        return value<min_value ? 0 : (value>max_value ? 0 : 1);
    }
    /// <summary>
    /// Return a saw-tooth value between a minimum and a maxmimum value.
    /// <remarks>Sorts the min/max values from lowest to highest</remarks>
    /// </summary>
    /// <param name="value">The value to wrap</param>
    /// <param name="min_value">The lower limit</param>
    /// <param name="max_value">The upper limit</param>
    /// <returns>A scalar value</returns>
    [Pure]
    public static double WrapAround(this double value, double min_value, double max_value)
    {
        double dx = Math.Abs(max_value-min_value);
        min_value=Math.Min(min_value, max_value);
        return value-dx*Math.Floor((value-min_value)/dx);
    }
    /// <summary>
    /// Return a saw-tooth value between zero an a maximum value.
    /// </summary>
    /// <param name="x">The value to use</param>
    /// <param name="x_high">The maximum value allowed</param>
    /// <returns>A scalar value</returns>
    [Pure]
    public static double WrapAround(this double value, double max_value)
    {
        return WrapAround(value, 0, max_value);
    }
    /// <summary>
    /// Return x when between min and max value, otherwise clamp at limits
    /// </summary>
    /// <example>    
    ///     ClapMinMax(-0.33, 0.0, 1.0) = 0.00
    ///     ClapMinMax( 0.33, 0.0, 1.0) = 0.33
    ///     ClapMinMax( 1.33, 0.0, 1.0) = 1.00
    /// </example>
    /// <param name="x">The value to clamp</param>
    /// <param name="min_value">The minimum value to use</param>
    /// <param name="max_value">The maximum value to use</param>
    /// <returns>A scalar value</returns>
    [Pure]
    public static double ClampMinMax(this double value, double min_value, double max_value)
    {
        return value>max_value ? max_value : value<min_value ? min_value : value;
    }
    /// <summary>
    /// Return x when more than min, otherwise return min
    /// </summary>
    /// <param name="x">The value to clamp</param>
    /// <param name="min_value">The minimum value to use</param>
    /// <returns>A scalar value</returns>
    [Pure]
    public static double ClampMin(this double value, double min_value)
    {
        return value<min_value ? min_value : value;
    }
    /// <summary>
    /// Return x when less than max, otherwise return max
    /// </summary>
    /// <param name="x">The value to clamp</param>
    /// <param name="max_value">The maximum value to use</param>
    /// <returns>A scalar value</returns>
    [Pure]
    public static double ClampMax(this double value, double max_value)
    {
        return value>max_value ? max_value : value;
    }

Fore me WrapAround and ClampMinMax are used all the time. Chi not so often. Examples would be getting angle results between -180 and +180. or 0 to 360 for usage in Sin() or Cos() in order to maintain precision. Try to see that Sin(1) != Sin(1+2*Math.PI). They differ by some 10^-16. But as the angles go up that error accumulates rapidly.