Terrence,
You might consider that you can move your index up by more than one at every
iteration.
When you're done with 2&pp you already know that 4 is not a prime candidate.
So, at the end of 3&pp you could advance to 5&pp . That is, you could
shorten:
5 pp 4 pp 3 pp 2 pp A
to:
5 pp 3 pp 2 pp A
But, of course, you cannot know this in advance; you have to update your index
at the end of every pp . So you couldn't use scan.
But you could, for example, recurse:
rec =: {. , 1 , { (] (i.&1@:] $:^:(<#) *) 0 ~: |) (}.~ >:)~
soe =: 2 rec 0: 1} i.
SoE =: I.@:soe
SoE 20
2 3 5 7 11 13 17 19
-Dan
PS to Roger: There's another bug in f. wrt $: :
dyd =. [ $: ]
bob =. 'x' dyd ]
bob f.
'x' 4 : 'x [ $: ] y' ]
The fixed form of bob should have parenthesis around the verb within the
explicit definition, like so:
'x' 4 : 'x([ $: ])y' ]
The problem is the original dyd is defined with a train of verbs, but its
fixed counterpart is not.
(bug report entitled: TRAIN DERAILED or FIXED VERBS BROKEN :)
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
