Thank you for pointing out the distinction between the abstract language and
its
implementation, and for the pointer to validate.ijs.
I have some thinking to do: as Raul pointed out on the first day,
L. E
0
so E =: 0$<1 will not do for an empty set if I use isboxed =: 0 < L. in my
definition of IsSet. In fact, 0 < L. rules out empties in the sense isempty =:
(0 e. $) . A famous point set theorist, R. L. Moore, refused to entertain the
notion of an empty set!
In any case, I agree it is best to program within the abstract language, and
the
evidence now is that "boxed" in the definition of Link ; means 0 < L. .
Thanks again,
Kip
Dan Bron wrote:
>> Here's a concrete example. We
>> know J (the notation) requires
>> arrays to be homogeneous.
>
> Perhaps a more germane example would've been:
>
> five =. {.5 2.3
> five
> 5
>
> Is five an integer?
>
> isInteger =. 4 = 3!:0
> isInteger five
> 0
>
> Not according to. 3!:0 . So maybe that's an inappropriate definition of.
> isInteger? Maybe instead:
>
> isInteger =. =<.
> isInteger five
> 1
>
> The tests in the standard 'validate' script follow this pattern, and take
> pains to avoid implementation details:
>
> NB. http://www.jsoftware.com/svn/base/trunk/main/main/validate.ijs
>
> load'validate'
> isboxed
> 0<L.
>
> Perhaps we can take that as Jsoftware's endorsement of this pattern as a best
> practice.
>
> -Dan
>
>
> Excuse typos; this message was composed on a phone.
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
