Mark wrote:
> The only exception that I know of is with ?. which pretends to be a
> rank-0 verb, but actually behaves like a rank-_ verb:
Raul responded:
> This is not a rank issue -- it's a state issue.
The verbs ? and ?. share the "stateful" aspect, but the former treats each
atom uniquely, the latter treats each total
argument uniquely. That is, fixedÂ-seed is sensitive to the rank of its
argument (it "sees" arguments of ranks greater than its
advertised rank of zero).
Fixed-seed says it generates the same random number every time it is fed the
same input, and it says it decomposes inputs of rank
> 0 into arguments of rank 0, these statements cannot both be true and
> reconciled with the observed behavior:
?. 6
0
?. 6 NB. Aleways gives the same random output for input 6
0
?. 6 NB. For the first 6, the second 6, and the third six...
0
?. 6 6 6 NB. Except when they're part of the same array
0 5 5
?. 4 4 $ 6
0 5 5 4
2 3 2 1
4 2 0 2
1 4 1 2
Conclusion: despite the dictionary definition, fixed-seed is a verb with
unbounded rank.
-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
