On Feb 12, 4:42 pm, Paul Rubin <http://phr...@[EMAIL
PROTECTED]
> wrote:
....
> What's SML/NJ abstract syntax? Do you mean some internal
> representation that is an artifact of that compiler? Or just
> simply typed lambda calculus?
SNL/NJ has an AbSyn module (structure) which might be unique
to that compiler, but ought to be easily converted into
whatever form MLton uses. It is more than simply typed
lambda calculus due to the signature-structure-functor
module sturcture and maybe some other aspects.
> Ocaml it seems to me is almost as ugly as SML.
Some parts are uglier, IMO.
> I'd look more
> in the direction of Haskell, which has a lot less clutter.
I intend to. No need for clutter, which was partly driven
by inadequate parsing technology at the time those MLs were
defined.
> I don't think this is needed. Is there a practical problem with
> the way Haskell syntax accomodates both styles without a pragma?
Since I have yet to learn just how Haskell does this, I don't
know yet, but I doubt writing the necessary code in the
compiler will be an issue.
To illustrate how far I am willing to go, I'm fairly well
committed to Early's algorithm for parsing: makes handling
syntax error messages much easier, at least for me. I am
considering a parse-a-little, type-a little compilation
strategy since:
(1) more flexible than all surface syntac, then all type
inference and checking;
(2) seems more akin to the way people read hard-to-comprehend
text.
|
