Maybe this illustrates where "too many caps" comes from.
',{:+ [: {:}:y'(13 : )
[: , [: {: [: + [: [: [: {: }:
It breaks my heart to replace all the Caps with Ats!
You change a nice right to left tacit expression back to a left to right
parenthesized expression.
Is it possible to improve on how the expression above inserts the Caps?
Linda
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Jose Mario
Quintana
Sent: Thursday, May 09, 2013 6:26 PM
To: [email protected]
Subject: Re: [Jprogramming] rosettacode
Skip, I finally was able to take a look at your version. The following should
be a series of equivalent definitions of your version. I do not code explicitly
thus the control words are alien to me but they probably are
correct:
jose1=: 4 : 0
k =. y - 1{x NB. k is the number of "kills"
L =. i. y NB. L is the position list
whilst. k =. k-1 do.
L =. (<:0{x) |. L NB. Rotate by the skip number
L =. }. L NB. Remove the dead index.
end.
)
Substituting L in the first line of the whilst block into the second line,
jose10=: 4 : 0
k =. y - 1{x NB. k is the number of "kills"
L =. i. y NB. L is the position list
whilst. k =. k-1 do.
L =. }. (<:0{x) |. L NB. Removing the dead index after rotating by the skip
number
end.
)
Replacing the whilst by a for form,
jose11=: 4 : 0
k =. y - 1{x NB. k is the number of "kills"
L =. i. y NB. L is the position list
for. i.k do.
L =. }. (<:0{x) |. L NB. Removing the dead index after rotating by the skip
number
end.
)
Eliminating L by reusing y,
jose12=: 4 : 0
k =. y - 1{x NB. k is the number of "kills"
y =. i. y NB. L is the position list
for. i.k do.
y =. }. (<:0{x) |. y NB. Removing the dead index after rotating by the skip
number
end.
)
Finding a tacit form for the for block...
'}. (<:0{x) |. y' (13 :)
[: }. ] |.~ [: <: 0 { [
Replacing Caps by Ats because Caps are overused! ;)
}.@:(] |.~ <:@:(0 { [))
}.@:(] |.~ <:@:(0 { [))
Replacing the for form by a power form,
jose13=: 4 : 0
k =. y - 1{x
y =. i. y
x (}.@:(] |.~ <:@:(0 { [))) ^: k y
)
Making a one-liner by substituting k and y in the first and second lines into
the code of the last line,
jose14=: 'x (}.@:(] |.~ <:@:(0 { [)))^:(y - 1{x) i. y' (4 :)
Trying to find a tacit expression,
'x (}.@:(] |.~ <:@:(0 { [)))^:(y - 1{x) i. y' (13 :)
4 : 'x (}.@:(] |.~ <:@:(0 { [)))^:(y - 1{x) i. y'
No, J is not smart enough; not yet anyway. Replacing (y - 1{x) by __ to find
the form,
'x (}.@:(] |.~ <:@:(0 { [)))^:__ i. y' (13 :) [ }.@:(] |.~ <:@:(0 { [))^:__
[: i. ]
Replacing the offending Cap,
[ (}.@:(] |.~ <:@:(0 { [))^:__) i.@]
[ }.@:(] |.~ <:@:(0 { [))^:__ i.@]
Consulting the Dictionary for power... Yes, the gerund form fits the bill!
jose15=. }.@:(] |.~ <:@:(0 { [))^:((] - 1 { [)`(i.@]))
3 7 (jose1 ; jose15) 10000
┌────────────────────────────────┬────────────────────────────────┐
│8922 111 930 2691 3850 6154 6624│8922 111 930 2691 3850 6154 6624│
└────────────────────────────────┴────────────────────────────────┘
They compare as follows,
st=. 7!:2@:] ; 6!:2
10 st '3 7 jose1 10000'
┌──────┬────────┐
│133440│0.100217│
└──────┴────────┘
10 st '3 7 jose15 10000'
┌──────┬────────┐
│263808│0.184226│
└──────┴────────┘
The explicit version is leaner and faster but my attitude is to wait for the
implementation to catch up (I would not bet on it though). Wait!
Apparently the issue is more complicated than just explicit versus tacit,
((<10) (((,@<@]) , st) &> ) (('3 7 ' , ] , ' 10000'"_)each @: ;:))
'jose1 jose10 jose11 jose12 jose13 jose14 jose15'
┌────────────────┬──────┬────────┐
│3 7 jose1 10000 │133440│0.100398│
├────────────────┼──────┼────────┤
│3 7 jose10 10000│199104│0.183993│
├────────────────┼──────┼────────┤
│3 7 jose11 10000│265152│0.174459│
├────────────────┼──────┼────────┤
│3 7 jose12 10000│265152│0.173494│
├────────────────┼──────┼────────┤
│3 7 jose13 10000│265728│0.181894│
├────────────────┼──────┼────────┤
│3 7 jose14 10000│265920│0.180529│
├────────────────┼──────┼────────┤
│3 7 jose15 10000│263808│0.180721│
└────────────────┴──────┴────────┘
This is puzzling to me; then again, I do not know much (to say the least) about
the performance of explicit code.
On Sat, May 4, 2013 at 1:42 AM, Skip Cave <[email protected]> wrote:
> Very nice Elton and Jose! I had a bug in my explicit formula, as well
> as having some extra variables I didn't need. So I simplified my
> formula and changed the index to the "Mayan Way" (zero index).
>
> jose1=: 4 : 0
> k =. y - 1{x NB. k is the number of "kills"
> L =. i. y NB. L is the position list
> whilst. k =. k-1 do.
> L =. (<:0{x) |. L NB. Rotate by the skip number
> L =. }. L NB. Remove the dead index.
> end.
> )
>
> So now my results match Jose and Elton's tacit versions.
>
> 3 4 jose1 20
> 19 1 7 12
>
> 5 1 jose1 200
> 112
>
> 7 10 jose1 345
> 283 287 295 319 72 92 93 137 175 224
>
> Interestingly:
>
> Ts '7 10 jose1 345'
> 0.0028602 6464
>
> Ts '7 10 jose 345'
> 0.0010386 65664
>
> Ts '3 7 jose1 10000'
> 0.0823621 133440
>
> Ts '3 7 jose 10000'
> |stack error: jose
> | 3 7 jose 10000
>
> Where jose1 is my explicit version, and jose is Elton's tacit version.
>
> Elton, I'm trying to parse through your tacit version to understand
> how it works. Could you break it appart, and explain what is going on in
> there?
>
> Is it possible to implement my explicit algorithm in a tacit
> expression that won't get a stack error with large numbers?
>
> Skip
>
>
> On Fri, May 3, 2013 at 7:13 PM, elton wang <[email protected]> wrote:
>
> > modified from Jose's recursive form:
> >
> > jose =. i.@{:@[`(] | {.@[ + [ $: _1 + ])@.({:@[ < ])
> >
> >
> --
> Skip Cave
> Cave Consulting LLC
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
