This site is a static rendering of the Trac instance that was used by R7RS-WG1 for its work on R7RS-small (PDF), which was ratified in 2013. For more information, see Home. For a version of this page that may be more recent, see NegativeRationalize in WG2's repo for R7RS-large.

Negative­Rationalize

cowan
2016-07-01 13:31:29
1history
source

The correct value of (rationalize 20 1) is 19, because 19/1 is the simplest rational that differs from 20 by an amount no more than 1. But it is not clear what the value of (rationalize 20 -1) should be. Is it also 19, or should an error be signaled?

(rationalize 20 1) => 19 and (rationalize 20 -1) => 19: Racket, MIT, Chicken with the numbers egg, Scheme48/scsh, Guile, Kawa, SISC, Chez, Vicare, Larceny, Ypsilon, Mosh, IronScheme, STklos, KSi, S7, Sagittarius, Foment, Chibi

(rationalize 20 1) => 19 but (rationalize 20 -1) signals an error: Gauche, Gambit

rationalize unsupported: Chicken, Bigloo, Detroit, Stalin, Scheme->C, SCM, NexJ, JScheme, SigScheme, Shoe, Mini-Scheme, TinyScheme, Scheme 9, RScheme, Unlikely, SIOD, BDC, XLisp, Rep, Schemik, Llava, Sizzle, FemtoLisp, Dfsch, Inlab, Picrin

(rationalize 20 1) => 20 and (rationalize 20 -1) => 20: UMB, SXM, Owl Lisp

Note that for a procedure that's been in the language since R2RS there are a remarkable number of Schemes that don't support it. I suspect this is because implementers assume that if they don't support arbitrary exact ratnums there is no point in implementing rationalize, though its arguments can be either exact or inexact.