What should a regex character class be compiled to?

A regex character class is a set of valid character for an expression. It's basically a bunch of or expressions. We could for example have the following:

["a", "e", "i", "o", "u"] = "a" | "e" | "i" | "o" | "u"

These two expressions are equivalent and one could be replaced by the other. Although there are many things that regex character classes make more convenient. For example ranges:

["a" - "z"] = "a" | "b" | "c" | "d" | ... | "z"

Regex character classes also allow to get the complement of a set. For example the following expression matches everything BUT digits:

~["0" - "9"] = ???

In our compilers class they taught us how to make an automata for the or expression. But they never told us anything about character classes.

These are my notes on the or automata.

Most of the character classes stuff could be implemented as a bunch of or expressions, but it's unclear what's the correct translation for the other expressions such as ranges and set complements.

What should be done?

I've read this article about email regex and it has a diagram which shows how character classes COULD BE implemented, so I'm leaving it here as a suggestion: