Chris, (et al), Referring to your recommendation regarding issue #1, I am having some difficulty. Norman claims that the reason the following computation results in 0's is that numbers like 15^22 are too big for the computer to represent. But 15^22 can be represented with a 64bit computer, at least in my J version, so the 0's do not produce the "signal" that the calculation needs adjustment.
cs=.29 15(cs&|@^)22 5 3 20 15 18 0 10 11 0 0 0 However, the subsequent calculation adjustments that are then employed in the fsoj are all correct as far as I know, and the only thing that I think needs change is a statement that the revised version (which uses powers instead of multiplication) requires a few other changes which are all included in the remaining write up. Such a statement should replace the calculation above and the explanation of the 0's. Maybe there is some other calculation that can be done to signal that a problem exists, but I could not produce such a calculation. Help. On Sat, Jul 28, 2018 at 8:00 AM, chris burke <[email protected]> wrote: > For 1), please use 2 large primes as suggested to fix this for J64. I think > any pair should do for this article. > > For 2) whether you use f"0 or f&> on a list of numbers is a matter of > taste, as also whether you include the "0 or &> in the definition or use it > when it is called. I would leave the style unchanged here, but after the > sentence "The & conjunction ..." it is worth pointing out that "0 could be > used to the same effect. > > > On Fri, Jul 27, 2018 at 3:56 PM, Brian Schott <[email protected]> > wrote: > > > I am now working on fsoj 39 about cyphers. There are 2 separate but > > slightly related issues I need help with. > > > > Issue #1) > > > > The following code in fsoj produces the result shown. > > > > cs=.29 > > 15(cs&|@^)22 5 3 20 15 18 > > 0 10 11 0 0 0 > > > > But current (64bit) J produces the following result. > > > > cs=.29 > > 15(cs&|@^)22 5 3 20 15 18 > > 6 10 11 24 14 9 > > > > Norman uses the fsoj result to lead into a short description of the > problem > > which results and a fix for the problem which involves using the > following > > adjustments. > > > > Instead of using > > > > enc=:cs&|@* > > > > Norman uses > > > > mul=.cs&|@* > > eenc=.mul/@# > > 5 eenc &> 22 5 3 20 15 18 > > 13 22 11 24 10 15 > > > > The issue is that current J does not need the revision, but the rest of > the > > fsoj is based on the such a need. I suspect a good alternative would be > to > > use a well studied product of 2 large primes in place of the cs=.29 to > show > > the problem exists and then use those 2 large primes in place of the 2 > > Norman uses later in the fsoj article ( 3551=53*67) . > > > > Comments on how to handle this question are sought here. If a pair of > large > > primes are part of the suggestion, are there well studied or well > > publicized examples? > > > > Issue #2) > > A smaller detail related to Norman using every (&>) in his development of > > the solution to the problem above when applying the verb eenc. He > justifies > > this choice on eenc NOT being a scalar verb. But if eenc is simply > defined > > with rank 0 as > > > > > > eenc=:mul/@#"0 > > > > > > then every is unnecessary. > > > > > > My question is, should the rank 0 version be used instead? > > > > -- > > (B=) <-----my sig > > Brian Schott > > ---------------------------------------------------------------------- > > For information about J forums see http://www.jsoftware.com/forums.htm > > > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > -- (B=) <-----my sig Brian Schott ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
