[r6rs-discuss] Unicode issues

[William D Clinger]

Ludovic Courtes wrote:
> Right, but these look more complex IMO, whereas memoization is really
> orthogonal to the implementation of the "real" `string-{ref,length}'
> (one could implement memoization and `string-set!' on top of an
> immutable string implementation with O(n) `string-{ref,length}'.)

An orthogonal memoization would perform miserably for
typical access patterns if the underlying representation
were UTF-8 or UTF-16.  With those representations, you
need to memoize the mapping from character index to code
unit index.  To an orthogonal memoization, that mapping
wouldn't even be visible.

Will