It's true that the axes of a rank 80 *dense* array must mostly have 0
or 1 elements.

But this constraint does not hold for sparse arrays.

Consider this example:

   example=: 42 (<p:i.60)} 1 $. (60#1e9);(i.60);0

Of course, there are limits:

   0 $. example
|limit error

And, ... it would be good to have some representative use cases
involving such "fat sparse" arrays to guide some implementation
decisions.

If this would require non-trivial changes to the code base, it does
not need to be a priority issue. But that's a different issue.

Thanks,

-- 
Raul

On Wed, Nov 24, 2021 at 9:57 AM Henry Rich <[email protected]> wrote:
>
> No, the rank limitation will remain in the final release.
>
> With rank 80 the axes must mostly have 0 or 1 elements.  The current
> sparse implementation is not efficient for that.  You will need to
> rewrite your sparse application to use dense arrays.  If you have a
> particular application that needs rewriting I will be willing to offer
> suggestions.
>
> Henry Rich
>
> On 11/24/2021 2:14 AM, 'Sergey Kamenev' via Programming wrote:
> > Health to all!
> >
> > I am using the stable version J with level 80 rank in sparse arrays.
> >
> > A large rank can be useful for some purposes (only in sparse arrays,
> > of course).
> >
> > Henry, can you lift the rank restriction in the stable release J 9.03?
> >
> > Nice day!
> > Sergey Kamenev
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
>
>
> --
> This email has been checked for viruses by AVG.
> https://www.avg.com
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to