Ian Clark wrote:
> Why didn't the blind old boomer simply use "Find..."?
> --they'll say.
I would've left the verb unchanged, and subordinated it into another
function that merely passed the global to it as an argument (assuming you
didn't need to update the global).
Raul wrote:
> niladic=: (0 :) (1 :0)
> assert.-.'m n u v x y'e.&;:m
> 3 :m
> )
>
> This is an adjective. Its argument should be
> a string, a boxed list of strings or 0 (in which
> case it must be followed by a script)
Couple notes:
* Since you're always returning a monadic verb, there is no reason
to check for any local name except y , and you probably don't want to
preclude access to (ill-considered) globals named x u m etc.
* The adverb won't actually work if its input is boxed. You'd
probably want to say something along the lines of 'y' e.&;:S:0 m but I
haven't tested that either.
* Even this doesn't go far enough, because, for example, I could
cheat by passing in ' ''y''~ '' . What we really want is to remove the
possibility of referring to y -- and this is easily achieved by the simple
expedient of removing y . Taking a cue from load_z_ , we could write
niladic=: (0 :) (1 :0)
3 : (' erase ''y'' '; m)
)
Of course, this would cause errors at run time, as opposed to define time
like the assert method. So we could try to get the best of both worlds:
niladic=: (0 :) (1 :0)
assert m e.&;:S:0 'y'
3 : (' erase ''y'' '; m)
)
But again, I haven't tested this*.
-Dan
* And it immediately occurs to me that it might break if the input were
LF-delimited as opposed to boxed sentences or a plain sentence. So maybe we
want to get a little fancier:
niladic=: (0 :) (1 :0)
lv =. 3 : m NB. Hoping this can't ever have side effects.
s =. {:"1 (1) 5!:7 <'lv'
assert. -. 'y' e.&;:S:0 s
3 : s
)
And yet again, this is untested.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
