> From: Dan Bron
>
> Ric wrote:
> > 10&#.^:_1 (or 10&#.inv ) has always seemed to me to be the "right"
> > way of converting a number to its digits but it is slower and fatter
> > than the other methods:
>
> Pepe wrote:
> > Perhaps, for the reason that you mentioned, there is case for
> > special code for ( 10&#.^:_1 ) .
>
> Agreed.
Seems like a good idea to me.
> > ( ,.&.": ) is indeed a neat phrase
>
> Also agreed! However, I am concerned with (&.":), because it applies
> (":^:_1), which is (". ) which obviously can and will
> execute arbitrary code. And this implication may not be obvious to the
> programmer when he types (&.":). Now, there's no problem
> if he controls (y) but applying these phrases to user strings makes the
> application vulnerable to "J injection" attacks.
>
> This isn't a problem in the current case (because the input is numeric,
> and even if it were a rank-1 string, (".) would never be
> applied to more than one character anyway). But I do have an open
> relevant enhancement request open on it [2], where I recommend
> x&". be made invertible.
Yes I was about to reply to Raul's suggested solution ( >./ 5 */\ "."0 data )
that I had an uneasy feeling using Do (monadic ". ) to convert strings to
numbers, when I realised that was exactly what ,.&.": was doing:
": b. _1
".
I am (was) not so concerned about security vulnerabilities (it would be nice to
have to worry about hordes of J programmers out there waiting to seize on such
opportunities! ;-) ) as I am about conversion errors, although in this
instance there shouldn't be any problems.
". '4 -3 7'
1 _3
_". '4 -3 7'
4 _3 7
". '4 3 7 7e-4'
|ill-formed number
| ".'4 3 7 7e-4'
_". '4 3 7 7e-4'
4 3 7 0.0007
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
