http://www.jsoftware.com/jwiki/Essays/Odometer
Fastest of these is the sparse solution
(4 $. $.)@($&1)
To fill your table assuming you've used c notation:
odometer =: (4 $. $.)@($&1)
i=:{.
j=:1&{
k=:{:
($ ([: ((i*j) - (i*k) - (j*k))"1 odometer))2 3 4
0 0 0 0
0 1 2 3
0 2 4 6
0 _1 _2 _3
1 1 1 1
2 3 4 5
Or
'I J K' =: ($"1 ([: |: odometer)) 2 3 4
(I * J) - (I * K) - (J * K)
0 0 0 0
0 1 2 3
0 2 4 6
0 _1 _2 _3
1 1 1 1
2 3 4 5
($"1 ([: |: odometer)) is the hook (u v) identify
$"1 as u
([: |: odometer) as v, a fork.
In a hook v always evaluates as a monad.
The capped fork (u v w) evaluates v as a monad on the result of w.
($"1 ([: |: odometer)) independently reshapes the components of the
index matrix, increasing the rank by 1, and then
'I J K' =: RHS assigns the items accordingly.
Hi.
How can someone create list of indexes of 3d table (N x M x K)?
I have written this simple code:
ind3d =: 3 : 0
'n m k' =. y
p0 =. k&|
p1 =. (m&|)@:<.@:(%&k)
p2 =. <.@:(%&(m*k))
(p2 , p1 , p0)"0 (i. (n*m*k))
)
but it's very very slow. Does any tacit solution for this problem exist?
Another question is how to fill the table with values depending on indexes
of cell? For example, F[i,j,k] = (i * j) - (i * k) + (j * k).
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
