Hypothetically speaking, it would be possible to implement J with certain primitives defined "at run time" by delegating their implementation to some name in some locale.
For that matter, we could design the interpreter to treat its executable as an archive (like maybe a zip file) with some !: command to execute a named element from the archive as a script. Once we have that, it becomes straightforward to do a variety of things with J... Thanks, -- Raul On Sat, Apr 25, 2015 at 8:37 PM, Henry Rich <[email protected]> wrote: > I wouldn't. d. and D. have many peculiarities, which you can see by looking > at the bug list. > > Long ago Roger said IIRC that if he had it to do over he would not implement > d. and D. as primitives. His point was that the analysis needed could be > done just as well in a J script, and then it could be expanded and > maintained by ordinary mortals (my words, not his). > > Since the analysis is done when the conjunction is executed, not when the > derived verb is executed, there's no need to avoid J for the analysis. > > A J model of d. and D. would be a better use of time. > > Henry Rich > > On 4/25/2015 8:28 PM, Raul Miller wrote: >> >> Oops, you are right. >> >> Also, it seems that these work properly: >> >> (^&_2.000000000001) d. _1 >> (^&_1)d._1 >> >> So it seems like a bug. >> >> Does anyone feel like diagnosing why this happens? Or should I do that? >> >> Thanks, >> > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
