Fraction

Arduino library to implement a Fraction data type (experimental).

Description

The fraction library implements fractional numbers a.k.a. Q, (integers are Z and floats/doubles are R), and the conversion to floats.

The code is working with a number of limitations a.o.:

• denominator is max 4 digits to keep code for multiply and divide simple
• Fractions are not exact, even floats are not exact.
• the range of numbers supported is limited.
• code is experimental.

That said, the library is useful e.g. to display float numbers as a fraction. From programming point of view the fractionize(float) function, converting a double / float into a fraction is a nice programming problem, how to do it fast while minimizing the error.

In short, use fractions with care otherwise your sketch might get broken ;)

Notes on natural order

Depending on fractionize(float) algorithm used the natural order of numbers might be broken. This means that if two floats are very close

float f  <  float g  does not imply  Fraction(f)  <  Fraction(g)
float f  >  float g  does not imply  Fraction(g)  >  Fraction(f)

The minimalistic fractionize keeps the natural order due its simplicity. It does have a lower accuracy as only limited number of denominators are used. This means that if two floats are very close

float f  <  float g  implies  Fraction(f)  <=  Fraction(g)
float f  >  float g  implies  Fraction(g)  >=  Fraction(f)

0.2.0 Breaking change 1

When testing with the array implementation it became evident that some Fractions were incorrect (not just inaccurate).

An analysis lead to using reciproke values for fractions larger than 1. By excluding the integerPart the problem looks solved in most cases. For very small values there are still problems as the fraction cannot be determined.

A test sketch fraction_extensive.ino has been added to test all floats with up to five decimals 0.00000 .. 0.99999. Results looking good but it is no proof of correctness or guarantee there are no issues left. In fact the well known fraction for PI = 355/113 is not found in 0.2.0 any more. This will be investigated in the future.

0.2.0 Breaking change 2

The 0.1.x version implemented the Printable interface to allow direct printing of a Fraction object. However it became clear that this costs 2 extra bytes per element, which adds up when creating arrays of fractions.

So the Printable interface is removed and replaced by a toString() function.

Fraction fr(PI);
Serial.print(fr.toString());

Fractions can also be printed by using toFloat() or toDouble()

Interface

#include "fraction.h"

Constructors

• explicit Fraction(double)
• explicit Fraction(float)
• Fraction(int32_t nominator = 0, int32_t denominator = 1) Default zero constructor
• explicit Fraction(int32_t p)
• explicit Fraction(int16_t p)
• explicit Fraction(int8_t p)
• explicit Fraction(uint32_t p)
• explicit Fraction(uint16_t p)
• explicit Fraction(uint8_t p)
• Fraction(const Fraction &f)

Equalities

The Fraction library implements ==, !=, >=, >, <, <=

Basic Math

The Fraction library implements:

• addition: + and +=
• subtraction: - and -+
• multiplication: * and *=
• division: / and /=
• negation: -

Conversion

• double toDouble() converts the fraction to a double.
• float toFloat() converts the fraction to a float.
• String toString() converts the fraction to a String. The format is "(n/d)", where n has optionally the sign.
• bool isProper() absolute value < 1.
• float toAngle() returns 0..360 degrees.
• int32_t nominator() returns the nominator.
• int32_t denominator() returns the denominator.

Miscellaneous (static)

• Fraction mediant(const Fraction&, const Fraction&)
• Fraction middle(const Fraction&, const Fraction&)
• Fraction setDenominator(const Fraction&, uint16_t) (might be simplified still)

Use with care

The library is reasonably tested. If problems arise please open an issue.

Future

Must

• improve documentation
• test test test ...

Should

• investigate the fraction of PI (0.2.0 does not find 355/113)
• performance testing
• investigate better fractionize()
  • see fraction_fast.ino for faster fractionize (price == accuracy)
  • a good start value ?
  • depends on nominator / denominator size
• float fractionize() returns the error.
• investigate divide by zero errors
  • NAN in fraction? => 0/0 ?
  • INF in fraction? => 1/0 and -1/0?

Could

• extend unit tests
• experiment with bigger nominator/denominator using all of 32767 possibilities ?
• add famous constants as Fraction e.g
  • FRAC_PI = 355/113
  • FRAC_E = 3985/1466
  • FRAC_GOLDEN_RATIO = (2584/1597)
• add parameters to toString() to set () and separator?

Wont

Support

If you appreciate my libraries, you can support the development and maintenance. Improve the quality of the libraries by providing issues and Pull Requests, or donate through PayPal or GitHub sponsors.

Thank you,