On 4/5/07, Roger Hui <[EMAIL PROTECTED]> wrote:
There is an executable model (algorithmic
description) of the rank conjunction in my APL95
paper "Rank and Uniformity".
http://www.jsoftware.com/papers/rank.htm
Also, a somewhat simpler model (which relies on
the native " implementation for positive rank values):
rank=:2 :0
tripleR=.3 $&.|. n
neg=.0>tripleR
if.0=+/neg do. u"tripleR return.end.
(u 2 :('u"(0>.({.n)+#$y)y';':';'x u"(0>.(}.n)+x,&([EMAIL PROTECTED])y)y')
tripleR)"(tripleR+neg{0 _)
)
+/rank _1 i. 2 2
1 5
That said, this probably uses features of J which are not covered
in J for C at the point where it initially discusses rank.
--
Raul
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
