I think both are sometines useful.   If the result table
representation gets too large, the implementation can use
procedural representation.   Yet, for some implementations,
those functions reserve a room of some optimization.

>From: Brad Lucier <xxxxxx@math.purdue.edu>
Subject: Omit predicate->char-set and char-set-invert[!] ?
Date: Fri, 8 Dec 2000 15:17:12 -0500 (EST)

> The functions predicate->char-set and char-set-invert[!] can take a
> long time if you define char-sets on a Unicode system (which you can do
> on Gambit-C), especially if you follow the suggestion in the body of
> the proposal:
>
> "Large" character types, such as Unicode, should use a sparse representation,
> taking care that the Latin-1 subset continues to be represented with a
> dense 32-byte bit set.