approx ( epsilon float -- a/b )

Vocabulary

math.continued-fractions

Inputs

epsilon	a positive floating point number representing the absolute acceptable error
float	a positive floating point number to approximate

Outputs

a/b	a fractional number containing the approximation

Word description

Give a rational approximation of float with a precision of epsilon using the smallest possible denominator.

Definition

USING: kernel math vectors ;

IN: math.continued-fractions

: approx ( epsilon float -- a/b )
    dup 1vector
    [ 3dup >ratio - abs < ] [ dup next-approx ] while 2nip
    >ratio ;