monadic do-over fails when n is applied as a function argument

[spreadLink]

I'm not sure if this occurs due to how the evaluation rules work, but I'm p sure it is indeed a bug. Specifically, the over form of "n m/ x", where n is an integer, m a monadic function and x some starting argument, should apply m to x n times. And it does, unless n is supplied as an argument in a function (please see the screenshot for when specifically the error occurs). (id here is "2 2 # 1 1 1 0", the function as a whole is supposed to be the fibonacci Q matrix thingy)

[tavmem]

Yes, it's definitely a bug. Thanks!!! In k2.8:

$ rlwrap -n ~/k2.8/k
K 2.8 2000-10-10 Copyright (C) 1993-2000 Kx Systems
Evaluation. Not for commercial use. 
\ for help. \\ to exit.

  id:2 2 # 1 1 1 0
  id
(1 1
 1 0)

  5 {id _mul x}/ id
(13 8
 8 5)

  {5 {id _mul x}/ id}[]
(13 8
 8 5)

  {[n] n {id _mul x}/ id}[5]
(13 8
 8 5)

[tavmem]

This is interesting. In kona:

$ rlwrap -n ~/kona/k
kona      \ for help. \\ to exit.

  id:2 2 # 1 1 1 0
  {[n] n {id _mul x}/ id}[5]
(13 8
 8 5)
  \\

but:

$ rlwrap -n ~/kona/k
kona      \ for help. \\ to exit.

  id:2 2 # 1 1 1 0
  5 {id _mul x}/ id
(13 8
 8 5)
  {[n] n {id _mul x}/ id}[5]
type error
x _dot\:y
  ^

[tavmem]

The above is a regression. In the commit 068fc66eb33e8b577965c2b9ba1fa5fa0ae8aa9f of Mar 11, 2014:

$ rlwrap -n ./k
K Console - Enter \ for help

  id:2 2 # 1 1 1 0
(1 1
 1 0)
  5 {id _mul x}/ id
(13 8
 8 5)
  {[n] n {id _mul x}/ id}[5]
(13 8
 8 5)

It no longer works beginning with the commit b887d9005c7de84e42d5c219afd10d8d77ae9ff5 of Mar 13, 2014.

[tavmem]

Just as an FYI: The commit that causes the problem (above) was made to fix a problem in the Burrows Wheeler Transform script:

$ cat bwt.k
bwt:{[s],//1#'|:'{x@<x}(!#a)!\: a:"\0",s} // the BWT
s:{x@<x}
twb:{[e]1_*s(-1+#e){[x]e,'s x}/e} // undo the BWT
twb bwt "dilute dilute ok!"
$ 

Executing the script using the commit of Mar 11, 2014 you get:

$ rlwrap -n ./k ~/nsl/bwt.k
K Console - Enter \ for help

  {[s],//1#'|:'{x@<x}(!#a)!\: a:"\0",s}
  {x@<x}
  {[e]1_*s(-1+#e){[x]e,'s x}/e}
  (;;;;;;;;;;;;;;;;"\000")

Using the commit of Mar 13, 2014 you get the correct result:

$ rlwrap -n ./k ~/nsl/bwt.k
K Console - Enter \ for help

  {[s],//1#'|:'{x@<x}(!#a)!\: a:"\0",s}
  {x@<x}
  {[e]1_*s(-1+#e){[x]e,'s x}/e}
  "dilute dilute ok!"

[spreadLink]

That's quite interesting. So it specifically overrides the argument capture?

[tavmem]

The commit of Mar 13, 2014 was to address a scoping problem described in issues 239 and 240 which was identified by @silentbicycle in his implementation of the Burrows Wheeler Transform. The issue that you identified became a regression caused by fixing the scope problem, but I think it can be fixed in a manner that preserves the proper functioning of the BWT. Still working on it.

[spreadLink]

Oh i see. Thank you very much then :smile: