Re: Providing features through the storage class Bradley Lucier 31 Jul 2015 21:39 UTC

On 07/29/2015 05:45 PM, John Cowan wrote:
> (I'm using my suggested terminology here.)
> Being a rectangular array should be orthogonal to being a mutable array,
> which should be orthogonal to being a safe array.  These effects can be
> accomplished by allowing the setter and checker of a storage class to
> be #f rather than a procedure, making the array immutable and unsafe
> respectively.  Furthermore, allow the generalized array creator to
> take an optional setter as well as a domain and a getter.  In that way,
> we wind up with arrays and a single subtype, rectangular arrays.

In this SRFI, basic arrays don't have a "storage class" in your sense.

A "safe" fixed-array checks the correctness of indices as well as values.

And I still don't know what you mean by "rectangular arrays"

> In addition, a standardized sparse-storage-class would be a Good Thing.
> This would use a hash table or similar object as its backing store.

I'll think about how to do this.