Yes, you're right, my proposed !.. would have to produce a verb of infinite rank.
I think your "(r;f) and my "r!..f are equivalent. You argue that "r!..f couldn't work because the filling required by u"r would be over and done with before !.. executed, but surely that understates Roger's ingenuity. We could simply say that the fill described in the Dictionary is derived from the type of the noun unless overridden by a !.. attached to the executing verb. A possible point in favor of !.. is that u"n has rank n while u"(n;f) has infinite rank, an asymmetry that would disappear if u"n!..f were used. A better point in favor of !.. is that the framing-fill mechanism is not part of the rank conjunction but rather a basic feature of all verb execution; a verb v has a rank, which I may or may not want to override, and if I want to change the fill it would be better for me to be able to write v!..f rather than having to assign a rank as well. But this is all nugatory. I have never needed to override fill, and a mechanism to do so is not likely to be added to the language. Henry Rich On 9/8/2010 5:31 AM, [email protected] wrote: > Henry Rich wrote (slightly re-ordered): >> Isn't it plain that you could define a conjunction !.. (rank mu lu ru) >> such that u!..n is the same as u, but with n used as the framing fill? > > Frame filling during result assembly happens outside of a verb > working at its rank mlr. You simply cannot have your u!..n at the > same rank as u because it wouldn't stand a chance to look at "fill > space". It would have to be at rank _ _ _ . > >> I wonder what he said. I remember his saying that the language has no >> way to specify the framing fill, but I don't see that it could not be >> extended to have one. > > I proposed that, too. Just let me nroff that part of the original > article again. > > "Steps 1/2/3" below refer to (1) argument disection into cells and > pairing/replication according to "II.B Agreement", (2) verb executions > at cell level, (3) re-assembly of the individual results into common > shape/frame, i.e. the filling we are talking about here. > > What I would consider to be a step in the right direction is > having a conjunction at the level of the rank ". Just like > " gives control over argument dissection and cell pairings > prior to verb evaluation, it could allow for control of the > fill value for shape assimilation during the reassembly of > the intermediate results. Without compromising existing > behavior, the rank conjunction could accept a boxed pair > containing rank and fill specifications. For example: > > > "(0; 9) 1 ; 2 3 ; 4 5 6 > 1 9 9 > 2 3 9 > 4 5 6 > > Details like the order of that boxed pair and the defaulting > of specifications are an issue left to the language lawyers > and designers. The important point is that it should really > be rank itself that should be extended. A separate, new > fill conjunction defined to control step 3 could not do the > job in concert with rank: applying both conjunctions to a > verb would result in nested verbs so that steps 1 and 3 > would lie on different levels. No matter what the order of > the nesting were, there would never be anything to fill. > > End of quote. The ranks of the proposed>"(0;9) notation would have > to be _ _ _ as they would have to be for your !.. notation. > I wouldn't stick this functionality onto !. because a rank change > not associated with this modifier otherwise. > > Martin > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
