> > data/metadata being a rank1/list array whose length is +/@:(*/&>) > > boxed/meta shape
> How would this be better than a rank 1 list of boxed arrays? Avoiding fills avoids having to interpret all fills. The advantage of your 1 fill is that it permits item count calculation through */ The disadvantage is that a shape 4 1 might be a frequently enough used shape, when you want to append to it. (inverted tables for an integer field would use an (item,1) shape. The advantage of 0 fill is that an (x,0) shape is not super useful (except i.0 0 can be used to append in table form) A _ fill is another option if 0 could create a limitation. But then boxed format keeps your advantage of */ count, without the downside of cleaning/interpreting fills vs intentional values step(s). On Thursday, March 3, 2022, 03:48:54 p.m. EST, Raul Miller <rauldmil...@gmail.com> wrote: On Thu, Mar 3, 2022 at 3:28 PM 'Pascal Jasmin' via Programming <programm...@jsoftware.com> wrote: > I understand your metashape and data concept to be (using my prior > example): > > (3 2 2 ,: 4 1 1) $ i.16 > > would let code know that there are 12 items in 1st array, and 4 in 2nd. That example matches how I am currently thinking about this, yes. > And then I understand your metarank concept would be the way to tell > if that 2nd array's shape is 4 vs 4 1 1. A list equal to items in > "metashape" might be what you meant? My thinking is that the metashape would be a matrix, which would mean that each row would have to contain the same number of elements. However, it would be a sparse matrix (meaning that the 1s in the shape would be represented implicitly and would appear as fill, in operations on the metashape). > It would seem like boxed shapes would provide the same information with I am sure that a boxed representation could represent the same information, but boxes are irregular -- so it does not make sense to me to introduce a higher degree of regularity using boxes as the core data model. > data/metadata being a rank1/list array whose length is +/@:(*/&>) > boxed/meta shape How would this be better than a rank 1 list of boxed arrays? Thanks, -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
