++ed by:

1 non-PAUSE user.

Martin Becker
and 1 contributors

NAME

Math::Logic::Ternary - ternary logic and related operations

VERSION

This documentation refers to version 0.004 of Math::Logic::Ternary.

SYNOPSIS

  use Math::Logic::Ternary qw(true false nil);
  use Math::Logic::Ternary qw(word9 word27 word81 ternary_word);
  use Math::Logic::Ternary qw(:all);

  $a = word9(        1234  );
  $a = word9(       '1234' );
  $a = word9(       '%bSS' );
  $a = word9(  '@tffntfnt' );

  $b = word9(       -5678  );
  $b = word9(      '-5678' );
  $b = word9(       '%SfS' );
  $b = word9( '@fnttfnfnt' );

  $c = $a->xor($b);             # word9(962)
  $d = $a->Cmp($b);             # true
  $e = $b->Sign;                # false

  # 1234 * -5678 + 962 == -7005690 == -356 * 3**9 + 1458
  ($f, $g) = $a->Mul($b, $c);   # (word9(1458), word9(-356))

DESCRIPTION

Introduction

Ternary or three-valued logic is a logic in which there are three truth values indicating true, false and some third value, often used to denote uncertainty or indefiniteness.

Ternary logic within a system of reasoning has come up in philosophy as an alternative to classical logic. Some spoken languages such as Aymara allow to express ternary logical propositions more easily than others. Programming languages usually support just binary logic.

Ternary logic and ternary number systems have applications in engineering and computer science. It is possible to build digital systems based on ternary logic using memory units with three distinct states (flip-flap-flops) and logic circuitry with three distinct input/output levels, like positive, negative and zero voltage. A three-valued information unit is called a trit.

This library focuses on computational aspects of ternary logic. It provides a selection of operations a ternary computer might be equipped with, and it addresses some related design considerations.

Objects accessible through Math::Logic::Ternary can represent single trits or data words composed of several trits. The module does not overload Perl builtin operators, nor does it change semantics of existing operators or control flow primitives.

Ternary truth values are called true, false, and nil in this module.

Embedding binary in ternary logic

Ternary logic can be seen as a generalization of binary ("Boolean") logic, with operators working exactly like well-known Boolean operators in the domain of true and false values while hopefully still making some kind of sense in the presence of nil values.

Different authors have used additional truth values for different concepts and thus defined different ternary logic schemes. We use only one of these schemes here, namely Kleene ternary logic, for its simplicity and symmetry.

In Kleene ternary logic, nil can be thougt of as a value that can either be true or false, but about which nothing is known at each time of evaluation. If nil values take part in a logical expression, the result can be calculated using Boolean logic by replacing all occurences of nil independently by true and false in turn (which requires two to the power of the number of nil values evaluations in total). If the result is always the same, that value is also the value of the ternary logical expression. Otherwise, its value is nil.

Note that in this approach, every occurence of nil is treated independently, even if some variables may be repeatedly involved. This is a consequence of seeing nil as a truth value of its own rather than a placeholder for some unknown but fixed boolean value.

Examples:

  +-------+-------+
  |   A   | not A |
  +-------+-------+
  | false | true  |
  | nil   | nil   |
  | true  | false |
  +-------+-------+

  +-------+-------+---------+--------+---------+---------+
  |   A   |   B   | A and B | A or B | A eqv B | A xor B |
  +-------+-------+---------+--------+---------+---------+
  | false | false |  false  | false  |  true   |  false  |
  | false | nil   |  false  | nil    |  nil    |  nil    |
  | false | true  |  false  | true   |  false  |  true   |
  | nil   | false |  false  | nil    |  nil    |  nil    |
  | nil   | nil   |  nil    | nil    |  nil    |  nil    |
  | nil   | true  |  nil    | true   |  nil    |  nil    |
  | true  | false |  false  | true   |  false  |  true   |
  | true  | nil   |  nil    | true   |  nil    |  nil    |
  | true  | true  |  true   | true   |  true   |  false  |
  +-------+-------+---------+--------+---------+---------+

Note that the logical equivalence of two nil values is nil, not true, just as the equivalence of two unknown conditions is unknown. This also figures if "A is equivalent to B" is written as "(A and B) or not (A or B)". Trit equality, on the other hand, yields always true or false, and of course true for two nil trits. Symbolic names for logical equivalence and trit equality relations are eqv and eq, respectively.

Ternary number systems

Most suitable for arithmetic on a ternary computer are positional number systems based on three digits. This library provides arithmetic of three different ternary number systems, namely balanced ternary, unbalanced ternary and base(-3).

Balanced ternary uses a base of 3 and the digits -1, 0 and 1. Unbalanced ternary uses a base of 3 and the digits 0, 1 and 2. Base(-3) uses a base of -3 and the digits 0, 1, and 2.

Unfortunately, there seems to be no general agreement on which ternary truth value should be associated with a given numerical value. The choices in this module are as follows:

  • Negative values (in balanced ternary representation) are considered false.

  • Positive values (in balanced ternary representation) are considered true.

  • Zero is associated with the remaining truth value, neither true nor false.

These choices are motivated mostly by symmetry considerations.

Consequently, logical values of false, nil, true correspond to balanced single trit values of -1, 0, +1 respectively. Some examples:

  +--------+-----------------+-------------+------------+
  | number | balanced trits  | truth value | sign value |
  +--------+-----------------+-------------+------------+
  |   -4   | (-1) * 3 + (-1) |   false     |     -1     |
  |   -3   | (-1) * 3 +   0  |   false     |     -1     |
  |   -2   | (-1) * 3 + (+1) |   false     |     -1     |
  |   -1   |   0  * 3 + (-1) |   false     |     -1     |
  |    0   |   0  * 3 +   0  |   nil       |      0     |
  |    1   |   0  * 3 + (+1) |   true      |     +1     |
  |    2   | (+1) * 3 + (-1) |   true      |     +1     |
  |    3   | (+1) * 3 +   0  |   true      |     +1     |
  |    4   | (+1) * 3 + (+1) |   true      |     +1     |
  +--------+-----------------+-------------+------------+

For unbalanced arithmetic, trit value -1 is replaced by 2 and the other values are left alone. This means that addition modulo 3 is the same operation in balanced and unbalanced base 3 arithmetic, whereas comparison operations are quite different.

  +--------+------------------+
  | number | unbalanced trits |
  +--------+------------------+
  |    0   |   0 * 3  +  0    |
  |    1   |   0 * 3  +  1    |
  |    2   |   0 * 3  +  2    |
  |    3   |   1 * 3  +  0    |
  |    4   |   1 * 3  +  1    |
  |    5   |   1 * 3  +  2    |
  |    6   |   2 * 3  +  0    |
  |    7   |   2 * 3  +  1    |
  |    8   |   2 * 3  +  2    |
  +--------+------------------+

Note that this particular implementation of unbalanced ternary arithmetic does not deal with negative numbers, avoiding another controversial choice among different possible representations.

Possible solutions include treating the most significant trit as balanced ternary digit, or using an extra trit to hold just the sign, or implicitly shifting the unsigned number range down by exactly half of its maximum. All of these result in complications that are absent in balanced arithmetic, so they should not be missed very much.

The negative base case, finally, shares the trit values of the unbalanced case but assigns different weights. The sign of a negative base number is determined by the position of the most significant nonzero trit.

  +--------+------------------+------------+
  |        |                  |  base(-3)  |
  | number | unbalanced trits | sign value |
  +--------+------------------+------------+
  |   -6   |  2 * (-3) +  0   |     -1     |
  |   -5   |  2 * (-3) +  1   |     -1     |
  |   -4   |  2 * (-3) +  2   |     -1     |
  |   -3   |  1 * (-3) +  0   |     -1     |
  |   -2   |  1 * (-3) +  1   |     -1     |
  |   -1   |  1 * (-3) +  2   |     -1     |
  |    0   |  0 * (-3) +  0   |      0     |
  |    1   |  0 * (-3) +  1   |     +1     |
  |    2   |  0 * (-3) +  2   |     +1     |
  +--------+------------------+------------+

Objects

The core of this library revolves around two basic object classes, Math::Logic::Ternary::Trit and Math::Logic::Ternary::Word. A trit is the smallest ternary information unit. It has a range of three values. A word is a container for many trits. Words come in different sizes: typical are 9, 27, or 81 trits. Words of all sizes are implemented in the Word class.

The top level module Math::Logic::Ternary wraps up constructors for both object classes in a single interface. It also contains the introductory documentation you are reading right now.

Trit and Word objects generally represent constants and have therefore no alterable attributes. Computations yielding new values will create new objects.

Trit and Word object classes share a couple of methods through their common Math::Logic::Ternary::Object role. This makes it possible to use trits and words interchangeably in many places, most notably as second, third or fourth operands. Note however that leftmost operands acting as method invocants of course do have to be in the class defining the desired operator as method. (We might have taken this further, even thrown both classes together, but felt keeping the distinction to be the more natural choice.)

Comparison of word sizes

  +-------------+-------------+--------------+
  | ternary     | binary      | decimal      |
  +-------------+-------------+--------------+
  |     9 trits |  14.26 bits |  4.29 digits |
  |    27 trits |  42.79 bits | 12.88 digits |
  |    81 trits | 128.38 bits | 38.65 digits |
  +-------------+-------------+--------------+
  | 10.09 trits |     16 bits |  4.82 digits |
  | 20.19 trits |     32 bits |  9.63 digits |
  | 40.38 trits |     64 bits | 19.27 digits |
  | 80.76 trits |    128 bits | 38.53 digits |
  +-------------+-------------+--------------+
  | 12.58 trits |  19.93 bits |     6 digits |
  | 18.86 trits |  29.90 bits |     9 digits |
  | 25.15 trits |  39.86 bits |    12 digits |
  | 37.73 trits |  59.79 bits |    18 digits |
  | 50.30 trits |  79.73 bits |    24 digits |
  | 75.45 trits | 119.59 bits |    36 digits |
  +-------------+-------------+--------------+

  +-------+----------------------------------------------------+
  | trits |                   balanced range                   |
  +-------+----------------------------------------------------+
  |   9   |                   -9841 ..  9841                   |
  +-------+----------------------------------------------------+
  |  27   |        -3_812798_742493 .. 3_812798_742493         |
  +-------+----------------------------------------------------+
  |  81   | -221_713244_121518_884974_124815_309574_946401 ..  |
  |       |  221_713244_121518_884974_124815_309574_946401     |
  +-------+----------------------------------------------------+

  +-------+----------------------------------------------------+
  | trits |                  unbalanced range                  |
  +-------+----------------------------------------------------+
  |   9   |                       0 .. 19682                   |
  +-------+----------------------------------------------------+
  |  27   |                       0 .. 7_625597_484986         |
  +-------+----------------------------------------------------+
  |  81   |                                              0 ..  |
  |       |  443_426488_243037_769948_249630_619149_892802     |
  +-------+----------------------------------------------------+

  +-------+----------------------------------------------------+
  | trits |                   base(-3) range                   |
  +-------+----------------------------------------------------+
  |   9   |                   -4920 .. 14762                   |
  +-------+----------------------------------------------------+
  |  27   |        -1_906399_371246 .. 5_719198_113740         |
  +-------+----------------------------------------------------+
  |  81   | -110_856622_060759_442487_062407_654787_473200 ..  |
  |       |  332_569866_182278_327461_187222_964362_419602     |
  +-------+----------------------------------------------------+

String Representations of Ternary Data

This library supports several formats for input and output of ternary data. Special prefix characters help to disambiguate among these. Single trits can be represented by their names as '$true', '$false', or '$nil', respectively.

All objects have numeric values in all three implemented number systems. These numeric values can be represented as decimal integer numbers.

Ternary words can also be represented as sequences of 't', 'f', or 'n' characters with a '@' prefix. Like decimal numbers, these are written left to right from most to least significant unit. Note that trits within word objects and words within arrays of word objects are addressed with the least significant element at index zero, though.

Finally, a more compact format for ternary words uses 27 symbols, namely 26 ASCII letters and the underscore. This format is called base27 here and uses a prefix of '%'. In base27, the underscore represents a value of zero, lower case letters 'a' through 'm' represent values of one through 13, and upper case letters 'N' through 'Z' represent values of -13 through -1. In input, case is ignored. In output, case is chosen such that lexical sorting in the 'C' locale will preserve order. Again, most significant units are written leftmost.

Examples for all word to string conversions can be found in Math::Logic::Ternary::Word.

Exports

None by default.

These functions can be explicitly imported:

nil, true, false, ternary_trit, word9, word27, word81, and ternary_word.

The export tag :all is a symbol for all of them.

Convenience Functions

nil

The nil trit.

true

The true trit.

false

The false trit.

ternary_trit($arg)

The trit identified by $arg.

The argument can be a small integer, 0 for nil, 1 for true, -1 or 2 for false. It can also be one of the strings '$true', '$false', or '$nil'.

word9(@args)

Equivalent to ternary_word(9, @args).

word27(@args)

Equivalent to ternary_word(27, @args).

word81(@args)

Equivalent to ternary_word(81, @args).

ternary_word($n, $int)

Ternary word with $n trits, given as an integer.

ternary_word($n, $str)

Ternary word with $n trits, given as a string of 't'/'n'/'f' characters with prefix '@', or a string of base27 characters with prefix '%'.

ternary_word($n, @trits)

Ternary word with $n trits, given as a list of trit values, read from lowest to highest significance, to be padded with nil trits.

ROADMAP

  • Binary/ternary conversions.

  • Fully implement floating point arithmetic.

  • Efficient multiplication and division.

  • Tcalc file I/O and batch mode.

  • Math::Logic::Ternary::Hardware - modelling ternary digital circuitry

DEPENDENCIES

This version of Math-Logic-Ternary requires these other modules and libraries to run:

  • perl version 5.8.0 or higher

  • Math::BigInt version 1.78 or higher (usually bundled with perl)

  • Role::Basic (available on CPAN)

Additional requirements to run the test suite are:

Recommended modules for increased functionality are:

BUGS AND LIMITATIONS

As of version 0.004, the list of numerical operators and corresponding low-level trit operators is not at all complete.

For example, operators to divide a value by two and compute the arithmetic mean of two values are already in the pipeline but did not make it into the release.

This library in general is a work in progress and in the state of beta testing. Feel free to send comments, observations and suggestions to the author. He would especially like to hear about features actually implemented in ternary digital hardware, whether addressed in this library or not.

Version 0.004 is the first official beta release of the library. Earlier versions should no longer be used. Upcoming development will aim to extend the API without radically changing existent functionality, unless of course there is a compelling reason to do so. At the present stage, nothing should be considered final.

Please submit bug reports and suggestions through the CPAN RT, https://rt.cpan.org/Public/Bug/Report.html?Queue=Math-Logic-Ternary .

PROVIDED MODULES AND SCRIPTS

The distribution provides 16 modules and one pod-only file in the Math::Logic::Ternary namespace.

Additionally, an interactive ternary calculator is provided as executable script tcalc.

RESOURCES

Project Homepage

https://vera.in-ulm.de/ternary-logic/

Bug Tracker

https://rt.cpan.org/Public/Bug/Report.html?Queue=Math-Logic-Ternary

SEE ALSO

AUTHOR

Martin Becker <becker-cpan-mp@cozap.com>

COPYRIGHT AND LICENSE

Copyright (c) 2006-2017 by Martin Becker. All rights reserved.

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.8.0 or, at your option, any later version of Perl 5 you may have available.

DISCLAIMER OF WARRANTY

This library is distributed in the hope that it will be useful, but without any warranty; without even the implied warranty of merchantability or fitness for a particular purpose.