J’s "extended" arithmetic sometimes makes programming simpler but some other
times more difficult; and I am afraid this "feature" is here to stay. You can
circumvent it by using a cover verb for (dyadic) %. For example, if full rank
is sufficient and some loss of performance is not an issue the following might
help:
over=. [ % [:^:(+./@:(0 = ,@:]))
I=. >:@i. 10 100 1000
J=. *:I
I (% -: over) J
1
st=. 7!:2 , 6!:2
st'I over J'
16778112 0.094218958
st'I % J'
8389312 0.0358302522
0 1 over 2 3
0 0.333333333
0 1 over 0 3
|domain error: over
| 0 1 over 0 3
|[-0]
0 1 over 2 0
|domain error: over
| 0 1 over 2 0
|[-0]
% b.0
0 0 0
over b.0
_ _ _
From: Eldon Eller <[EMAIL PROTECTED]>
Forgive me for not being more clear. I am a super superannuated
engineer. I and (by unwarranted projection) other engineers tend to use
computers, and J in particular, as a superlative calculator. We may have
occasion to evaluate an expression such as my example that results in
0%0 without realizing it. I.e., we do not separately evaluate numerator
and denominator. In such cases, J returns 0, which is the wrong answer,
as demonstrated by application of l'Hospital's rule. J gives no warning
that there is anything pathological in the expression. As a result,
bridges are collapsing, planes are falling from the sky, cell phones are
exploding, and wheels are coming off wagons.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
