Declaration/syntax for sparse arrays

What would be the best way to specify that an array is sparse, now or in the future?

var[sparse] * int32
1000[sparse] * int32

It would be great to see more metadata associated with types, as discussed in the documentation of libndtypes2, but they should also be as standardized as possible to avoid alternative ways of expressing the same semantics.
