Maths Functions in Ginger

Number Types

Ginger currently supports the following numerical types - in the future complex numbers with be added.

  • “Small” integers: broadly speaking these correspond to the native machine representation of integers, with a couple of bits reserved for tagging. e.g. -1, 0, 1, 2, 77, -23456. These automatically overflow into big-nums.
  • “Bignums”, which are integral values of unlimited precision. e.g. 12345678900. These automatically convert into small integers when needed.
  • Rational numbers, which are the ratio of two integral values called the numerator and the denomiator. e.g. 1/2, -113/355. Strictly speaking, rationals will be normalised so that the numerator and denominator are coprime, so that they do not share any prime factors, and the denominator is positive.
  • Double precision floats: these are the same as the native repesentation of doubles. The clean rules of `transreal arithmetic`_ are used for all double precision arithmetic. All arithmetic is ‘total’, meaning that it always returns a single, meaningful value - and it is safe and correct to divide by zero, raise zero to the power of zero, etc.
  • Strictly transreal numbers: nullity, +infinity and -infinity. These are cleaned up, safe versions of the floating point NaN and infinities that can be used with integers, rationals and floats.

Total Arithmetic

Arithmetic operators are “total”; this means they always return results and never raise exceptions. This follows the underlying rules of transreal arithmetic.

Because arithmetic is total, numbers are automatically converted from one type to another as necessary. The general rule is that conversions maximise accuracy. (Note, however, that the “/” operator always returns a float - if that is unsuitable there are alternatives, see div and ./.)

Working with Non-finite Numbers

Ginger provides three special values for representing non-finite numbers: infinity, -infinity and nullity. Infinity and -infinity can be intuitively understood as standing for numbers that are too large, positively or negatively, to be represented exactly. Nullity, on the other hand, can be understood as the situation where you have no information about the value.

Unlike the normal floating point versions, there is nothing ‘bad’ about these values. They are generated predictably, work the way you might expect given these intuitive meanings, can be compared reliably with themselves and with other non-finite numbers, and can be stored in variables and so on.

The two infinities are generated when dividing any non-zero finite value by zero, much as in normal floating point arithmetic. They are also used for representing the overflow of arithmetic operations that would otherwise fail.

Nullity is generated when dividing zero by zero and also arises naturally in many problematic situations, such as adding plus and minus infinities. It work a good deal like NaN in ordinary floating point arithmetic but can be safely comparsed with itself. It often comes in handy when writing functions such as the arithmetic mean - there’s no need to check for an empty list of values:

define arithmeticMean( list ) =>>
        var sofar := 0;
        for i in list do
                sofar ::= sofar + i
        endfor;
        sofar / list.length
enddefine;

>>> arithmeticMean( [] );
There is one result.
1.      nullity

Basic Arithmetic Functions

The basic arithmetic functions are +, -, *, and /.

Scientific Constants

TO BE DONE

Scientific Functions

TO BE DONE

See Also

To learn more about Transreal arithmetic, take a look at this tutorial http://www.bookofparagon.com/Mathematics/Evolutionary_Revolutionary_2011_web.pdf . ((Not tutorial enough))

More academically minded readers may enjoy learning the theory.

Axioms of Transreal arithmetic. http://www.bookofparagon.com/Mathematics/PerspexMachineVIII.pdf

Transreal analysis inc. scientific functions.

http://www.bookofparagon.com/Mathematics/PerspexMachineIX.pdf