On Wed, Feb 28, 2018 at 3:39 PM, james faure <james.fa...@epitech.eu> wrote: > So... Our generator is infinite, and we cannot fall back to conjuring it up > physically. Does this mean that some operatoions are impossible, and must > return a domain error? Operations must in any event be stored symbolically, > so some errors may go undetected for some time (pending a rigorous proof of > the following conjecture: 'we are able to either guarantee correctness for > all generator links based on i._, or bail out immediately with an error'). > Here is perhaps our biggest problem then: Are there any errors that cannot be > predicted by a generator link on i._ ? Well, boxes aside, we should be able > to see coming all syntax, rank and length errors should be predictable for > any generator link, since we can analyze the complete information we have for > generating the array. As a result then, we need this axiom (for a valid > generator): it must guarantee it's ability to generate an array (short of > system limits)). Should my conjecture about guaranteeing correctness on i._, > prove false, then the offending operations on infinite generators must > trigger a domain error.
You should be aware that addition of certain values can trigger errors: _+__ |NaN error Thanks, -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
