Just to follow up that this limitation/feature of f. is a necessary one. When
using $: as recursion (potential multiple repeats) limiting its scope to just
its (named) self is or at least feels like a prudent way to avoid recursion
errors. Errors that are more painful than most since infinite recursion
requires a "break" to get control back.
On Wednesday, February 9, 2022, 11:29:25 a.m. EST, 'Pascal Jasmin' via
Programming <programm...@jsoftware.com> wrote:
dictionary page for f. mentions "problems" with f. when u includes $:
this example seems ok though:
+ 1 : ' 0 & $: : (u/ % #) f.'
0&$: :(+/ % #)
cut f.
' '&$: :([: -.&a: <;._2@,~)
but:
, 1 : ' u@cut f.'
,@(' '&$: :([: -.&a: <;._2@,~) (1 : 0))
u y
:
x u y
)
adds some "explicit boilerplate"
n&$: is an extremely useful pattern for providing default x argument to dyad.
I still passionately hate the new AC train interpretation. (should be uACv
instead of uACu), but it does offer an elegant-ish workaround.
([: -.&a: <;._2@,~) ((' '&) :)
' '&([: -.&a: <;._2@,~) :([: -.&a: <;._2@,~)
that should be faster as well in that $: is an extra call that probably
includes edge cases that interfere with f.
An even more elegant workaround though is:
mdef =: 2 : 'n&u : u' NB. monad default parameter n to dyad u
cut =: ([: -.&a: <;._2@,~) mdef ' ' NB. same result as tacit AC modifier above.
"explicit junk boilerplate" now removed.
, 1 : ' u@cut f.'
,@(' '&([: -.&a: <;._2@,~) :([: -.&a: <;._2@,~))
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
