I think Graham is referring to what would happen if the ideas in section 2 of the following article were implemented.
http://www.jsoftware.com/papers/rank.htm ----- Original Message ----- From: Henry Rich <[EMAIL PROTECTED]> Date: Tuesday, January 9, 2007 12:02 pm Subject: RE: [Jprogramming] Re: Why won't this plot? > I can't quite make J sense of your question, but I suspect > the root of the problem is the meaning of [EMAIL PROTECTED] . > > [EMAIL PROTECTED] is a new verb, with its own rank (which is taken > from the rank of v). So when you write > > [EMAIL PROTECTED] y > > The entire verb [EMAIL PROTECTED] is applied to each atom of y. For each > atom, the interpreter executes +. then {. There is never > any possibility if executing +. (or {.) over the entire y, > because {. never sees the entire y. > > The interpreter could choose to recognize [EMAIL PROTECTED] as a special > combination, in effect making a new primitive out of that > combination, and then using Integrated Rank Support for > the combination. But it doesn't. > > [EMAIL PROTECTED] is a new verb. Important, but often misunderstood. > > Henry Rich > > > > > -----Original Message----- > > From: [EMAIL PROTECTED] > > [EMAIL PROTECTED] On Behalf Of > > Graham Parkhouse > > Sent: Tuesday, January 09, 2007 3:51 AM > > To: [email protected] > > Subject: [Jprogramming] Re: Why won't this plot? > > > > Henry wrote: > > > > >Your analysis is almost right. Actually, {."1 benefits > > >from Integrated Rank Support, which means that the > > >loop over the operand is a tight inner loop rather than > > starting the whole > > verb up anew for each 1-cell. > > > > > >So, [EMAIL PROTECTED] starts {. and +. for each 0-cell, while > > >{."1@:+. starts +. once and {. once. > > > > Thanks, that's really helpful. I like the concept of integrated rank > > support. Using an argument of i.30 40 50 there is a noticably > shorter> response time if it is prefixed by the second verb. Why > > shouldn't integrated > > rank support be extended to the first verb? I guess it's just > > a matter of > > interpreter sophistication, and no doubt there are pros and > > cons to every > > piece of added complexity. ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
