SML: What's the difference between using abstype and using a signature to hide the implementation of a structure?

Question

I\'ve done a little work in SML in the past, but I\'m now starting to get to the more interesting parts.

Using the abstype...with...end construct, I can ma

Answer 1

Why would I ever want to use abstype?

Starting with the easiest, you won't. Atleast I can't come up with one good reason.

---

Aren't signatures/structures just a more general version of abstypes?

Well, I guess we have to take a look at the history of SML. The opaque (... :> ...) signature matching was not part of SML '90 as explained on this smlnj document about modules 1.3.9. opaque signature matching :>

... whose purpose was to create an "abstract" instance of the signature SIG. This feature was left out of SML '90 for various reasons, but the need for it was real.

I have no idea about the reasoning for not including it, but as far as I know McQueen was the "farther" of the abstype which was part of SML '90 and for some reason wasn't removed in SML '97 (maybe backwards compatibility?)

There is however a fundamentally difference between them, abstype is part of the core language where modules/signatures/functors are part of the module system.

---

What can I do with abstypes that I can't do with signatures/structures?

I can't come up with anything. However I'm pretty sure it would be easy to construct some example that you can using opaque signature matching and can't do with abstypes.

---

UPDATE

The page Degrade abstype to derived form from the successor-ml wiki actually contains a tiny informal description about abstype being a leftover.

As with many other, they also refer to the sections of the Defects in the Revised Definition of Standard ML paper which contains details about some "minor" errors/defects in the definition of abstype, although their link being dead. The "Revised Definition of Standard ML" is the definition of SML '97.